JP6542008B2 - Accident occurrence forecasting device and accident occurrence forecasting method - Google Patents

Description

  Embodiments of the present invention relate to an accident occurrence forecasting device and an accident occurrence forecasting method.

  Conventionally, there is known a technique of learning occurrence patterns of traffic accidents using traffic information in the past, and predicting the possibility of occurrence of traffic accidents based on the learning result.

JP 2014-35639 A

  In the above-described techniques, it is desirable if the accuracy of learning and forecasting of a traffic accident occurrence pattern can be further improved.

The accident occurrence prediction apparatus according to the embodiment includes a plurality of learning units, a plurality of prediction units, and a switching unit. The plurality of learning units are based on a self-organizing map that performs learning of traffic accident occurrence patterns using traffic information in the past. The plurality of forecasting units are provided one by one corresponding to each of the plurality of learning units so as to configure a plurality of learning units and a plurality of combinations . Each of the plurality of forecasting units outputs a forecast indicating the possibility of occurrence of a traffic accident based on the learning result by each of the plurality of learning units. The switching unit is the first information included in the traffic information, and executes learning using the first information including the information related to the weather, which can classify the traffic information into a plurality of types according to the occurrence tendency of the traffic accident 1 one of the learning unit and switches the combination of one forecast unit for outputting a forecast among a plurality of combinations of a plurality of the learning unit and a plurality of prediction unit.

FIG. 1 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to the first embodiment. FIG. 2 is an exemplary block diagram showing a hardware configuration of a traffic accident occurrence forecasting device according to the first embodiment. FIG. 3 is an exemplary flowchart showing processing executed by the traffic information management unit according to the first embodiment. FIG. 4 is an exemplary flowchart showing processing executed by the time zone determination unit according to the first embodiment. FIG. 5 is an exemplary flowchart showing processing performed by the accident occurrence pattern learning unit according to the first embodiment. FIG. 6 is an exemplary flowchart showing processing executed by the accident occurrence forecasting unit according to the first embodiment. FIG. 7 is an exemplary block diagram showing a functional configuration of the traffic accident occurrence forecasting device according to the second embodiment. FIG. 8 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to the third embodiment. FIG. 9 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to the fourth embodiment. FIG. 10 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to the fifth embodiment. FIG. 11 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to the sixth embodiment. FIG. 12 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to a seventh embodiment. FIG. 13 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to the eighth embodiment. FIG. 14 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to a ninth embodiment. FIG. 15 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to the tenth embodiment. FIG. 16 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to the eleventh embodiment. FIG. 17 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to a twelfth embodiment. FIG. 18 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to a thirteenth embodiment. FIG. 19 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to a fourteenth embodiment. FIG. 20 is an exemplary block diagram showing a functional configuration of a traffic accident occurrence forecasting device according to a fifteenth embodiment.

  According to the embodiment, there is provided an accident occurrence prediction device including a plurality of learning units, a plurality of prediction units, and a switching unit. The plurality of learning units perform learning of a traffic accident occurrence pattern using traffic information in the past. The plurality of forecasting units are provided to correspond to the plurality of learning units. Each of the plurality of forecasting units outputs a forecast indicating the possibility of occurrence of a traffic accident based on the learning result by each of the plurality of learning units. The switching unit is the first information included in the traffic information, and the learning unit is configured to execute the learning using the first information capable of classifying the traffic information into a plurality of types according to the traffic accident occurrence tendency. The forecasting unit for outputting the forecast is switched among a plurality of forecasting units.

  Hereinafter, embodiments will be described in detail based on the drawings.

First Embodiment
First, with reference to FIG. 1 and FIG. 2, the structure of the traffic accident occurrence forecasting apparatus 100 by 1st Embodiment is demonstrated. In the first embodiment, information on time, more specifically, time zone information (details will be described later) is used as an example of the above-mentioned “first information”.

  As shown in FIG. 1, the traffic accident occurrence forecasting apparatus 100 according to the first embodiment has a traffic information management unit 101, a traffic information DB (database) 102, a time zone determination unit 103, and an accident occurrence as functional components. Pattern learning units 104A and 104B and accident occurrence prediction units 105A and 105B are provided. The accident occurrence pattern learning units 104A and 104B are examples of the “learning unit”, and the accident occurrence prediction units 105A and 105B are examples of the “forecast unit”. The time zone determination unit 103 is an example of the “switching unit”.

  In FIG. 1, the number of accident occurrence pattern learning units 104A and 104B as an example of the “learning unit” and the number of accident occurrence prediction units 105A and 105B as an example of the “forecasting unit” are both two. Although a certain configuration has been shown, the technology of the first embodiment is also applicable to a configuration in which the number of "learning units" and the number of "forecasting units" are both three or more.

  The traffic information management unit 101 is configured to receive input of traffic information, and accumulate and manage the received traffic information in the traffic information DB 102. Sensor information, control information, and time zone information are input to the traffic information management unit 101 according to the first embodiment as traffic information.

  The sensor information is information indicating traffic conditions measured by various sensors (for example, vehicle sensors) installed on the roadside of a road such as an expressway. The sensor information may be, for example, an integrated value of traffic volume indicating the number of vehicles passing the road per unit time, an average speed of the vehicle per unit time, an occupancy rate of the vehicles on the road, and the like.

  Moreover, control information is information used in a traffic control system, and is information that may affect traffic conditions. Such control information may be input by a traffic control system controller. The control information includes, for example, information on measures taken by the controller, information on traffic accidents that have occurred on roads, and information on constructions to be performed on roads. Further, as information on traffic accidents, for example, a flag indicating a factor causing a traffic accident, a time when the traffic accident occurred, a type of traffic accident, and the like can be mentioned.

  In addition, the time zone information is information that may affect the traffic accident occurrence tendency, and the traffic information accumulated in the traffic information DB 102 is a traffic accident occurrence tendency (e.g. It is information that can be classified into a plurality (at least as many as the number of accident occurrence pattern learning units 104A and 104B) according to the type, the scale, and the like. As the time zone information, for example, information indicating different time zones such as midnight, morning, daytime, and evening is set. The time zone information may be input by the traffic control system controller, similar to the above control information.

  The traffic information management unit 101 associates the sensor information, the control information, and the time zone information with each other, stores them in the traffic information DB 102, and stores the stored traffic information as needed (for example, the operator of the traffic accident forecasting device 100) Output to the time zone determination unit 103). The time zone determination unit 103 is configured to be able to output the traffic information input from the traffic information management unit 101 to the accident occurrence pattern learning units 104A and 104B and the accident occurrence prediction units 105A and 105B. There is.

  Here, in the first embodiment, the time zone determination unit 103 uses the time zone information of the traffic information input from the traffic information management unit 101 to output the traffic information to the accident occurrence pattern learning unit 104A. And it is comprised so that it may switch by accident occurrence forecasting part 105A, accident occurrence pattern learning part 104B, and accident occurrence forecasting part 105B. For example, when the time zone information indicates a time zone in which a traffic accident is likely to occur (such as the evening), the time zone determination unit 103 sets the output destination of the traffic information in the accident occurrence pattern learning unit 104A and the accident occurrence forecast unit 105A. When the time zone information indicates a time zone (such as daytime) in which a traffic accident is unlikely to occur, the traffic information output destination is set to the accident occurrence pattern learning unit 104B and the accident occurrence prediction unit 105B. There is.

  The accident occurrence pattern learning units 104A and 104B respectively use the traffic information in the past input from the traffic information management unit 101 via the time zone determination unit 103 to determine the correlation of the traffic information at the time of the traffic accident, etc. By learning, it is comprised so that learning of the occurrence pattern of a traffic accident may be performed. In the first embodiment, as described above, the traffic occurrence information having different time zone information is input to the accident occurrence pattern learning units 104A and 104B. Thus, in the first embodiment, the traffic information to be learned by the accident occurrence pattern learning units 104A and 104B can be made different depending on the likelihood of occurrence of a traffic accident.

  In addition, various methods can be considered as a learning method. As a relatively simple method, for example, a method of holding (accumulating) traffic information at the time of the accident (combination of sensor information, control information, and time zone information) or statistical processing of the combination of the traffic information at the time of the accident And clustering methods to create similar case data. Further, as another method, an algorithm using a self-organizing map as a kind of neural network (for example, see JP-A-2014-35639) and the like can be mentioned. In the first embodiment, as long as it is possible to learn the correlation of traffic information at the time of a traffic accident, other than the method using the self-organizing map, for example, the page network is used. Learning may be performed in the same manner.

  The accident occurrence prediction units 105A and 105B are provided to correspond to the accident occurrence pattern learning units 104A and 104B, respectively. The accident occurrence forecasting units 105A and 105B are configured to output a forecast indicating the possibility of occurrence of a traffic accident, based on the learning results by the accident occurrence pattern learning units 104A and 104B, respectively.

  For example, accident occurrence forecasting units 105A and 105B compare the similarity between traffic information at the time of forecasting and traffic information at the time of a traffic accident based on the learning result by accident occurrence pattern learning units 104A and 104B, respectively. When the similarity is equal to or higher than the threshold value, it is configured to output as a forecast (accident occurrence tendency) that there is a high possibility that a traffic accident will occur. In FIG. 1, the forecast output from the accident occurrence forecast unit 105A is illustrated as an accident occurrence trend A, and the forecast output from the accident occurrence forecast unit 105B is illustrated as an accident occurrence trend B. The accident occurrence forecasting units 105A and 105B according to the first embodiment calculate numerical values indicating the possibility of occurrence of a traffic accident based on the above similarity, and quantitatively output the possibility of occurrence of a traffic accident. It may be configured as follows.

  The traffic accident occurrence forecasting device 100 according to the embodiment has a hardware configuration using a normal computer 10 as shown in FIG. That is, the traffic accident occurrence forecasting apparatus 100 has an input / output I / F (interface) 11, a storage 12, a central processing unit (CPU) 13, a read only memory (ROM) 14, and a RAM (hardware) as hardware configurations. (Random Access Memory) 15. These hardware are connected to the bus 16.

  The input / output I / F 11 is an interface for connecting an output device such as a display or an input device such as a keyboard. The storage 12 is a storage medium for storing various data, such as a hard disk drive (HDD) or a solid state drive (SSD). The CPU 13 is an arithmetic processing unit configured to perform various arithmetic processes by executing various computer programs. The ROM 14 is a non-volatile memory configured to store various computer programs and the like executed by the CPU 13. The RAM 15 is a memory that functions as a work area for the CPU 13 to execute various computer programs. The functional modules shown in FIG. 1 are generated on the RAM 15 as a result of the CPU 13 reading and executing a predetermined computer program from the ROM 14.

  Next, processing executed by the traffic information management unit 101 according to the first embodiment will be described with reference to FIG.

  In the process flow of FIG. 3, first, in S1, the traffic information management unit 101 receives an input of traffic information. The traffic information management unit 101 according to the first embodiment receives inputs of sensor information, control information, and time zone information as traffic information. Then, the process proceeds to S2.

  In S2, the traffic information management unit 101 stores various information (sensor information, control information, and time zone information) included in the traffic information input in S1 in the traffic information DB 102 in association with each other. Then, the process proceeds to S3.

  In S <b> 3, the traffic information management unit 101 reads traffic information from the traffic information DB 102 and outputs the traffic information to the time zone determination unit 103. The process of S3 is executed, for example, in response to a request from the operator of traffic accident occurrence forecasting apparatus 100 or the like. Then, the process ends.

  Next, processing performed by the time zone determination unit 103 according to the first embodiment will be described with reference to FIG.

  In the process flow of FIG. 4, first, in S11, the time zone determination unit 103 receives an input of traffic information from the traffic information management unit 101. Then, the process proceeds to S12.

  In S12, the time zone determination unit 103 refers to the time zone information of the traffic information input in S11, and according to the time zone information, the traffic information output destination is the accident occurrence pattern learning unit 104A and the accident occurrence The prediction unit 105A is switched to the accident occurrence pattern learning unit 104B and the accident occurrence prediction unit 105B. Then, the process ends.

  Next, processing performed by the accident occurrence pattern learning units 104A and 104B according to the first embodiment will be described with reference to FIG.

  In the process flow of FIG. 5, first, in S21, the accident occurrence pattern learning units 104A and 104B receive the input of the traffic information from the time zone determination unit 103. By the process of S12 in FIG. 4 described above, traffic information having different time zone information is input to the accident occurrence pattern learning units 104A and 104B. Then, the process proceeds to S22.

  In S22, the accident occurrence pattern learning units 104A and 104B learn the occurrence pattern of a traffic accident using the traffic information input in S21. Then, the process ends.

  Next, with reference to FIG. 6, the process performed by the accident occurrence forecasting units 105A and 105B according to the first embodiment will be described.

  In the process flow of FIG. 6, first, in S31, the accident occurrence forecasting units 105A and 105B receive the input of the traffic information at the time of the forecast target. If the time to be forecasted is the future, the predicted value of the traffic information is input to the accident occurrence forecasting units 105A and 105B. Then, the process proceeds to S32.

  In S32, the accident occurrence forecasting units 105A and 105B are similar to the traffic information at the time of the forecast target input in S31 and the learning result by the accident occurrence pattern learning units 104A and 104B in S22 of FIG. Calculate Then, the process proceeds to S33.

  In S33, the accident occurrence prediction units 105A and 105B determine whether the similarity calculated in S32 is equal to or more than a threshold.

  If it is determined in S33 that the similarity is equal to or higher than the threshold value, the process proceeds to S34. Then, in S34, the accident occurrence prediction units 105A and 105B output that the possibility of occurrence of a traffic accident is high. Then, the process ends.

  If it is determined in S33 that the degree of similarity is less than the threshold value, the process as described in S34 is not performed, and the process ends.

  As described above, the traffic accident occurrence forecasting apparatus 100 according to the first embodiment is information included in traffic information, and is an example of information that can classify traffic information into a plurality of types according to the occurrence tendency of traffic accidents. A time zone determination unit 103 is provided, which switches between accident occurrence pattern learning units 104A and 104B for executing learning using time zone information and switches between accident occurrence forecast units 105A and 105B for outputting a forecast. Thereby, according to the first embodiment, the combination of the accident occurrence pattern learning units 104A and 104B and the accident occurrence forecasting units 105A and 105B is appropriately used according to the ease of occurrence of the traffic accident, using the time zone information. Can be selected. As a result, the accuracy of learning by each of the accident occurrence pattern learning units 104A and 104B can be enhanced, and the accuracy of prediction by each of the accident occurrence forecasting units 105A and 105B can be enhanced.

Second Embodiment
Next, with reference to FIG. 7, a traffic accident occurrence forecasting device 200 according to the second embodiment will be described. The second embodiment is the same as the first embodiment in that information on time is used as an example of “first information”, but not time zone information but day information (details will be described later) as information on time ) Differs from the first embodiment in that it is used. In the following, differences between the first embodiment and the second embodiment will be mainly described.

  As shown in FIG. 7, the traffic accident occurrence forecasting apparatus 200 according to the second embodiment has a traffic information management unit 201, a traffic information DB 202, a day of the week decision unit 203, an accident occurrence pattern learning unit 204A and a functional configuration. 204B and accident occurrence prediction units 205A and 205B. The accident occurrence pattern learning units 204A and 204B are an example of the “learning unit”, and the accident occurrence prediction units 205A and 205B are an example of the “forecast unit”. In addition, the day of the week determination unit 203 is an example of the “switching unit”.

  The traffic information management unit 201 is configured to receive inputs of sensor information, control information, and day information as traffic information, and store the received sensor information, control information, and day information in the traffic information DB 202 in association with each other. It is done.

  The day information is information that may affect the occurrence of a traffic accident, and the traffic information accumulated in the traffic information DB 202 is divided into a plurality of traffic information according to the traffic accident tendency (at least the accident occurrence pattern learning unit It is information that can be classified into the same number as the number of 204A and 204B. As the day information, information indicating the name of the day such as Monday, Tuesday,..., Sunday, or information indicating the type of the day such as weekdays and holidays is set.

  Here, in general, it is known that the traffic volume differs depending on the day of the week because the tendency of the destination of the car user is different between weekdays and holidays. In addition, it is known that the traffic volume is different between weekdays and holidays because holidays tend to increase the opportunities for driving even with drivers who do not drive a vehicle on a daily basis. Therefore, it is considered that the tendency of traffic accidents to occur differs depending on the day of the week as well as the traffic volume. For this reason, if it is possible to switch the accident occurrence pattern learning units 204A and 204B for performing the learning according to the day information, it is considered possible to execute the learning according to the tendency of occurrence of the traffic accident.

  Therefore, the day of the week determination unit 203 according to the second embodiment is configured to switch the output destination of the traffic information based on the day of the week information of the traffic information input from the traffic information management unit 201. For example, when the day-of-the-week information indicates the day of the week when a traffic accident is likely to occur, the day-of-week determination unit 203 sets the output destination of the traffic information in the accident occurrence pattern learning unit 204A and the accident occurrence forecasting unit 205A, and the day information indicates traffic When indicating the day of the week when an accident hardly occurs, the output destination of the traffic information is set to the accident occurrence pattern learning unit 204B and the accident occurrence prediction unit 205B.

  In the second embodiment, the same effect as the first embodiment can be obtained by being configured as described above. That is, according to the second embodiment, the accuracy of learning by each of the accident occurrence pattern learning units 204A and 204B can be enhanced using the day of the week information, and the accuracy of the forecast by each of the accident occurrence prediction units 205A and 205B It can be enhanced.

Third Embodiment
Next, a traffic accident occurrence forecasting device 300 according to a third embodiment will be described with reference to FIG. In the third embodiment, unlike the first and second embodiments, information on weather (weather information) is used as an example of the “first information”. In the following, portions of the third embodiment which are different from the first and second embodiments will be mainly described.

  As shown in FIG. 8, the traffic accident occurrence forecasting apparatus 300 according to the third embodiment has a traffic information management unit 301, a traffic information DB 302, a weather determination unit 303, an accident occurrence pattern learning unit 304A and a functional configuration. And an accident occurrence prediction unit 305A and 305B. The accident occurrence pattern learning units 304A and 304B are an example of a "learning unit", and the accident occurrence prediction units 305A and 305B are an example of a "forecasting unit". In addition, the weather determination unit 303 is an example of a “switching unit”.

  The traffic information management unit 301 is configured to receive inputs of sensor information, control information, and weather information as traffic information, and store the received sensor information, control information, and weather information in the traffic information DB 302 in association with each other. It is done.

  The weather information is information that may affect the occurrence of a traffic accident, and the traffic information accumulated in the traffic information DB 302 is divided into a plurality of traffic information according to the tendency of the traffic accident (at least the accident occurrence pattern learning unit Information that can be classified into the same number as the number of 304A and 304B. As the weather information, for example, information indicating the weather such as sunny, rain, fog, and snow is set.

  Here, in general, under conditions where the visibility is blocked, such as rain or fog, there are many drivers that significantly reduce the speed of the vehicle than when it is normal sunny, so traffic disturbance tends to occur. There is. Also, under conditions where the road surface is slippery, such as snow or rain, the braking distance of the vehicle is likely to be longer than when it is sunny, so it is thought that traffic accidents tend to occur more easily than when it is sunny. Therefore, the incidence of traffic accidents is considered to be different depending on the weather on the road. For this reason, if it is possible to switch the accident occurrence pattern learning units 304A and 304B for performing the learning according to the weather information, it is considered possible to perform the learning according to the tendency of the traffic accident to occur.

  Therefore, the weather determination unit 303 according to the third embodiment is configured to switch the output destination of the traffic information based on the weather information of the traffic information input from the traffic information management unit 301. For example, when the weather information indicates that the traffic accident is likely to occur, the weather judgment unit 303 sets the output destination of the traffic information in the accident occurrence pattern learning unit 304A and the accident occurrence prediction unit 305A, and the weather information indicates traffic In the case of showing a weather in which an accident is unlikely to occur, the output destination of the traffic information is set to the accident occurrence pattern learning unit 304B and the accident occurrence prediction unit 305B.

  In the third embodiment, the same effects as the first and second embodiments can be obtained by being configured as described above. That is, according to the third embodiment, it is possible to improve the learning accuracy by each of the accident occurrence pattern learning units 304A and 304B using the weather information, and the prediction accuracy by each of the accident occurrence prediction units 305A and 305B It can be enhanced.

Fourth Embodiment
Next, a traffic accident occurrence prediction device 400 according to a fourth embodiment will be described with reference to FIG. In the fourth embodiment, unlike the first to third embodiments, information on a vehicle traveling on a road, more specifically, large vehicle contamination information (details will be described later) is used as an example of the “first information”. Be In the following, portions of the fourth embodiment which are different from the first to third embodiments will be mainly described.

  As shown in FIG. 9, the traffic accident occurrence forecasting apparatus 400 according to the fourth embodiment has a functional configuration including a traffic information management unit 401, a traffic information DB 402, a large vehicle contamination determination unit 403, and an accident occurrence pattern learning unit 404A and 404B and accident occurrence forecasting units 405A and 405B. The accident occurrence pattern learning units 404A and 404B are an example of the “learning unit”, and the accident occurrence prediction units 405A and 405B are an example of the “forecast unit”. The large vehicle contamination determination unit 403 is an example of the “switching unit”.

  The traffic information management unit 401 receives inputs of sensor information, control information, and large vehicle contamination information as traffic information, and stores the received sensor information, control information, and large vehicle contamination information in the traffic information DB 402 in association with each other. It is configured to

  The large-sized vehicle mixture information is information that may affect the occurrence of traffic accidents, and the traffic information accumulated in the traffic information DB 402 is divided into a plurality of traffic information according to the tendency of traffic accidents (at least an accident occurrence pattern It is information that can be classified into the same number as the number of learning units 404A and 404B. Information such as a numerical value indicating the degree of contamination of a large vehicle (large vehicle) on a road is set as the large vehicle contamination information. A large vehicle is a vehicle such as a truck or bus having a size equal to or greater than a predetermined reference.

  Here, in general, on an expressway or the like, a vehicle group tends to be formed and traveled by a plurality of vehicles while maintaining an inter-vehicle distance with a vehicle traveling ahead ahead at a certain level or more. At that time, it is said that many drivers are often forming a vehicle group while adjusting the speed, paying attention also to the movement of a vehicle traveling two cars ahead. However, when a large vehicle such as a truck or a bus is traveling in front of the eyes, the visibility may be deteriorated, and the necessity of performing a sudden braking operation may be increased. For this reason, it is considered that the tendency of traffic accidents to occur differs depending on the degree of mixing under the situation where large vehicles are mixed. Therefore, if it is possible to switch the accident occurrence pattern learning units 404A and 404B for performing the learning according to the large vehicle mixture information, it is considered possible to perform the learning according to the tendency of the traffic accident to occur.

  Therefore, the large vehicle contamination determination unit 403 according to the fourth embodiment is configured to switch the output destination of the traffic information based on the large vehicle contamination information of the traffic information input from the traffic information management unit 401. . For example, when the mixing degree of a large car based on the large car mixing information indicates a mixing degree in which a traffic accident is likely to occur, the large car mixing determination unit 403 outputs the traffic information to the accident occurrence pattern learning unit 404A and the accident occurrence If it switches to the forecasting unit 405A and the mixing degree of a large vehicle based on large vehicle mixing information indicates a mixing degree in which a traffic accident is hard to occur, the traffic information output destination is the accident occurrence pattern learning unit 404B and the accident occurrence forecasting unit 405B. It is configured to switch.

  In the fourth embodiment, the same effects as in the first to third embodiments can be obtained by being configured as described above. That is, according to the fourth embodiment, it is possible to make the traffic information to be learned by the accident occurrence pattern learning units 404A and 404B different according to the possibility of occurrence of a traffic accident using the large vehicle contamination information. Since it is possible, the accuracy of learning by each of the accident occurrence pattern learning units 404A and 404B can be enhanced. As a result, it is possible to enhance the accuracy of forecasting by each of the accident occurrence forecasting units 405A and 405B corresponding to the accident occurrence pattern learning units 404A and 404B.

Fifth Embodiment
Next, a traffic accident occurrence forecasting device 500 according to the fifth embodiment will be described with reference to FIG. In the fifth embodiment, unlike the first to fourth embodiments, information on the structure, shape, or state of a road, more specifically, road gradient information (details will be described later), as an example of the “first information”. Is used. In the following, portions of the fifth embodiment which are different from the first to fourth embodiments will be mainly described.

  As shown in FIG. 10, the traffic accident occurrence forecasting apparatus 500 according to the fifth embodiment has a traffic information management unit 501, a traffic information DB 502, a road slope determination unit 503, and an accident occurrence pattern learning unit 504A as a functional configuration. And 504B and accident occurrence forecasting units 505A and 505B. The accident occurrence pattern learning units 504A and 504B are examples of the “learning unit”, and the accident occurrence prediction units 505A and 505B are examples of the “forecasting unit”. In addition, the road slope determination unit 503 is an example of the “switching unit”.

  The traffic information management unit 501 receives inputs of sensor information, control information, and road gradient information as traffic information, and stores the received sensor information, control information, and road gradient information in the traffic information DB 502 in association with each other. Is configured.

  The road gradient information is information that may affect the occurrence of traffic accidents, and the traffic information accumulated in the traffic information DB 502 is divided into a plurality of traffic information according to the tendency of traffic accidents (at least accident occurrence pattern learning Information that can be classified into the same number as the number of units 504A and 504B. As the road gradient information, information on the structure or shape of the road, more specifically, information indicating the presence or absence of the uphill and the downhill on the road is set.

  Here, in general, since there are many drivers who drive a long distance on a highway, etc., the traveling speed of the vehicle tends to be kept constant in order to reduce the driving load. However, when traveling with the same amount of acceleration, the speed increases on the downhill and decreases on the uphill. For example, in a place called a sag where there is an uphill after a downhill, the speed drops sharply because it enters the uphill with the accelerator amount narrowed down so that the speed does not go too high on the downhill. Traffic congestion is known to occur easily. Therefore, it is considered that a traffic accident is more likely to occur at a sloped portion of the road than at a flat portion. For this reason, if it is possible to switch the accident occurrence pattern learning units 504A and 504B for performing the learning according to the road gradient information, it is considered possible to perform the learning according to the tendency of the traffic accident to occur.

  Therefore, the road gradient determination unit 503 according to the fifth embodiment is configured to switch the output destination of the traffic information based on the road gradient information in the traffic information input from the traffic information management unit 501. For example, when the road gradient information indicates that there is a gradient on the road where a traffic accident is likely to occur, the road gradient determination unit 503 outputs the traffic information to the accident occurrence pattern learning unit 504A and the accident occurrence prediction unit 505A. If the road gradient information indicates that there is no gradient on the road where a traffic accident is likely to occur, the output destination of the traffic information is set to the accident occurrence pattern learning unit 504B and the accident occurrence prediction unit 505B. It is done.

  In the fifth embodiment, the same effects as in the first to fourth embodiments can be obtained by being configured as described above. That is, according to the fifth embodiment, it is possible to improve the learning accuracy by each of the accident occurrence pattern learning units 504A and 504B using the road gradient information, and the accuracy of the prediction by each of the accident occurrence prediction units 505A and 505B. Can be enhanced.

Sixth Embodiment
Next, a traffic accident occurrence prediction device 600 according to a sixth embodiment will be described with reference to FIG. The sixth embodiment is the same as the fourth embodiment in that information on a vehicle traveling on a road is used as an example of “first information”, but large vehicle contamination information is used as information on a vehicle traveling on a road. The fifth embodiment differs from the fifth embodiment in that low-speed vehicle mixture information (details will be described later) is used. In the following, portions of the sixth embodiment which are different from the fourth embodiment will be mainly described.

  As shown in FIG. 11, the traffic accident occurrence forecasting apparatus 600 according to the sixth embodiment has a traffic information management unit 601, a traffic information DB 602, a low speed vehicle contamination determination unit 603, and an accident occurrence pattern learning unit as functional components. 604A and 604B, and accident occurrence forecasting units 605A and 605B. The accident occurrence pattern learning units 604A and 604B are an example of the “learning unit”, and the accident occurrence prediction units 605A and 605B are an example of the “forecasting unit”. Further, the low speed vehicle mixture determination unit 603 is an example of the “switching unit”.

  The traffic information management unit 601 receives inputs of sensor information, control information, and low speed vehicle mixing information as traffic information, stores the received sensor information, control information, and low speed vehicle mixing information in the traffic information DB 602 in association with each other. It is configured to

  The low-speed vehicle mixture information is information that may affect the occurrence of a traffic accident, and the traffic information accumulated in the traffic information DB 602 is divided into a plurality of traffic information according to the tendency of the traffic accident (at least an accident occurrence pattern It is information that can be classified into the same number as the number of learning units 604A and 604B. As the low speed vehicle mixture information, information such as a numerical value indicating the degree of mixture of low speed vehicles on the road is set. The low speed vehicle is a vehicle traveling on the road at a speed equal to or less than a predetermined value.

  Here, as described above, on an expressway or the like, a plurality of vehicles often travel to form a vehicle group. However, if the head of the car group catches up with a car traveling at a speed lower than the speed of the car group, the speed of the whole car group is suddenly reduced, and the following vehicle must be braked. Become. In this case, it is considered that a driver performing a sudden lane change may occur due to a rapid decrease in speed, and a traffic accident such as a rear-end collision with a following vehicle may easily occur. Therefore, in the situation where low speed vehicles are mixed, it is considered that the tendency of occurrence of traffic accidents differs depending on the degree of mixing. For this reason, if it is possible to switch the accident occurrence pattern learning units 604A and 604B that execute the learning according to the low speed vehicle mixture information, it is considered possible to perform the learning according to the tendency of the traffic accident to occur.

  Therefore, the low-speed-vehicle mixing determination unit 603 according to the sixth embodiment is configured to switch the output destination of the traffic information based on the low-speed-vehicle mixing information of the traffic information input from the traffic information management unit 601. . For example, when the degree of mixing of the low speed vehicle based on the low speed vehicle mixing information indicates the mixing degree with which a traffic accident is likely to occur, the low speed vehicle mixing determination unit 603 outputs the traffic information to the accident occurrence pattern learning unit 604A and the accident occurrence If it switches to the forecasting unit 605A and the degree of mixture of low-speed vehicles based on low-speed vehicle mixture information indicates the degree of mixture where traffic accidents are less likely to occur, the traffic information output destination is the accident occurrence pattern learning unit 604B and the accident occurrence forecasting unit 605B. It is configured to switch.

  In the sixth embodiment, by being configured as described above, the same effects as in the first to fifth embodiments can be obtained. That is, according to the sixth embodiment, the accuracy of learning by each of the accident occurrence pattern learning units 604A and 604B can be enhanced using the low speed vehicle contamination information, and the forecast by each of the accident occurrence forecasting units 605A and 605B Accuracy can be improved.

Seventh Embodiment
Next, a traffic accident occurrence forecasting device 700 according to a seventh embodiment will be described with reference to FIG. In the seventh embodiment, unlike the first to sixth embodiments, information on traffic volume of a road, more specifically, traffic volume information in an interchange (details will be described later) is used as an example of the “first information”. Be In the following, portions of the seventh embodiment which are different from the first to sixth embodiments will be mainly described.

  As shown in FIG. 12, the traffic accident occurrence forecasting apparatus 700 according to the seventh embodiment has a functional configuration including a traffic information management unit 701, a traffic information DB 702, a traffic volume determination unit 703, and an accident occurrence pattern learning unit 704A. And 704B, and an accident occurrence prediction unit 705A and 705B. The accident occurrence pattern learning units 704A and 704B are an example of the “learning unit”, and the accident occurrence prediction units 705A and 705B are an example of the “forecasting unit”. In addition, the traffic amount determination unit 703 is an example of the “switching unit”.

  The traffic information management unit 701 receives inputs of sensor information, control information, and traffic information as traffic information, and stores the received sensor information, control information, and traffic information in the traffic information DB 702 in association with each other. Is configured.

  Traffic volume information is information that may affect the occurrence of traffic accidents, and the traffic information accumulated in the traffic information DB 702 is divided into a plurality of traffic information according to the tendency of traffic accidents (at least accident occurrence pattern learning Information that can be classified into the same number as the number of units 704A and 704B. As traffic information, information indicating the number of vehicles passing through the road interchange and the like is set.

  Here, in general, a rear-end collision tends to occur at the time of traffic congestion. Congestion occurs due to traffic concentration in a sag part on a road, a part where a signal is installed, a branch part, etc. However, when the number of vehicles traveling on the road is small, the congestion hardly occurs. For example, since the expressway is a closed space, the number of vehicles on the expressway is basically determined by the inflow / outflow at the interchange. For this reason, on the expressway, when the number of vehicles flowing in at the interchange is larger than usual, the possibility of traffic congestion on the main line becomes high. Therefore, the tendency of traffic accidents to occur is considered to differ depending on the traffic volume at the interchange. For this reason, if it is possible to switch the accident occurrence pattern learning units 704A and 704B which execute the learning according to the traffic volume information, it is considered possible to execute the learning according to the tendency of the traffic accident to occur.

  Therefore, the traffic volume determination unit 703 according to the seventh embodiment is configured to switch the output destination of the traffic information based on the traffic volume information of the traffic information input from the traffic information management unit 701. For example, when the number of vehicles based on traffic information indicates a value that is likely to cause a traffic accident, traffic volume determination unit 703 sets the output destination of traffic information in accident occurrence pattern learning unit 704A and accident occurrence forecasting unit 705A. When the number of vehicles based on traffic information indicates a value that is unlikely to cause a traffic accident, the output destination of the traffic information is set to the accident occurrence pattern learning unit 704B and the accident occurrence prediction unit 705B. .

  In the seventh embodiment, the same effects as in the first to sixth embodiments can be obtained by being configured as described above. That is, according to the seventh embodiment, it is possible to improve the learning accuracy by each of the accident occurrence pattern learning units 704A and 704B using traffic volume information, and the accuracy of the prediction by each of the accident occurrence prediction units 705A and 705B. Can be enhanced.

Eighth Embodiment
Next, a traffic accident occurrence forecasting device 800 according to the eighth embodiment will be described with reference to FIG. The eighth embodiment is the same as the fifth embodiment in that information on the road structure, shape, or state is used as an example of the "first information", but information on the road structure, shape, or state The fourth embodiment differs from the fifth embodiment in that speed limit information (details will be described later) is used instead of road slope information. In the following, portions different from the fifth embodiment of the eighth embodiment will be mainly described.

  As shown in FIG. 13, the traffic accident occurrence forecasting apparatus 800 according to the eighth embodiment has a traffic information management unit 801, a traffic information DB 802, a speed limit determination unit 803, and an accident occurrence pattern learning unit 804A as a functional configuration. And 804B and accident occurrence forecasting units 805A and 805B. The accident occurrence pattern learning units 804A and 804B are examples of the “learning unit”, and the accident occurrence prediction units 805A and 805B are examples of the “forecast unit”. Further, the speed limit determination unit 803 is an example of the “switching unit”.

  The traffic information management unit 801 receives inputs of sensor information, control information, and speed limit information as traffic information, and stores the received sensor information, control information, and speed limit information in the traffic information DB 802 in association with each other. Is configured.

  The speed limit information is information that may affect the occurrence of a traffic accident, and the traffic information stored in the traffic information DB 802 is divided into a plurality of traffic information according to the tendency of the traffic accident (at least accident occurrence pattern learning Information which can be classified in the same number as the number of units 804A and 804B. As the speed limit information, information on the state of the road, more specifically, information such as a numerical value indicating the speed limit set for the road is set.

  Here, in general, the patterns of traffic accidents that can occur differ depending on whether the speed limit is high or low. For example, if the speed limit is high, the size of the traffic accident may be larger than if the speed limit is low. Therefore, the tendency of traffic accidents to occur is considered to be different depending on the speed limit. For this reason, if it is possible to switch the accident occurrence pattern learning units 804A and 804B for performing the learning according to the speed limit information, it is considered possible to execute the learning according to the tendency of the traffic accident to occur.

  Therefore, the speed limit determination unit 803 according to the eighth embodiment is configured to switch the output destination of the traffic information based on the speed limit information of the traffic information input from the traffic information management unit 801. For example, when the speed limit information indicates a value that is likely to cause a traffic accident, the speed limit determination unit 803 sets the output destination of traffic information in the accident occurrence pattern learning unit 804A and the accident occurrence forecast unit 805A, and the speed limit information However, when indicating a numerical value that is unlikely to cause a traffic accident, the output destination of the traffic information is set to the accident occurrence pattern learning unit 804B and the accident occurrence prediction unit 805B.

  In the eighth embodiment, by being configured as described above, the same effects as those of the first to seventh embodiments can be obtained. That is, according to the eighth embodiment, the accuracy of learning by each of the accident occurrence pattern learning units 804A and 804B can be enhanced using the speed limit information, and the accuracy of the prediction by each of the accident occurrence prediction units 805A and 805B. Can be enhanced.

The ninth embodiment
Next, a traffic accident occurrence prediction device 900 according to a ninth embodiment will be described with reference to FIG. In the ninth embodiment, unlike the first to eighth embodiments, information on the condition around the road, more specifically, event information (details will be described later) is used as an example of the “first information”. In the following, portions of the ninth embodiment which are different from the first to eighth embodiments will be mainly described.

  As shown in FIG. 14, the traffic accident occurrence forecasting apparatus 900 according to the ninth embodiment has a traffic information management unit 901, a traffic information DB 902, an event determination unit 903, an accident occurrence pattern learning unit 904A and a functional configuration. And an accident occurrence prediction unit 905A and 905B. The accident occurrence pattern learning units 904A and 904B are examples of the “learning unit”, and the accident occurrence prediction units 905A and 905B are an example of the “forecasting unit”. Also, the event determination unit 203 is an example of the “switching unit”.

  The traffic information management unit 901 is configured to receive inputs of sensor information, control information, and event information as traffic information, and store the received sensor information, control information, and event information in the traffic information DB 902 in association with each other. It is done.

  The event information is information that may affect the occurrence of a traffic accident, and the traffic information accumulated in the traffic information DB 902 is divided into a plurality of traffic information according to the tendency of the traffic accident (at least the accident occurrence pattern learning unit It is information that can be classified into the same number as the number of 904A and 904B. As the event information, information indicating the holding status of the event around the road is set.

  Here, in general, when an event in which many people move around a road to be forecasted is held, it is conceivable that traffic congestion may occur due to traffic concentration around the road. In this case, compared with the traffic congestion that may occur when there is no event, the time zone in which the traffic congestion occurs, the type of vehicle that fits in the traffic congestion, etc. may differ, and many people traveling around the road may gather. Is considered. Therefore, the occurrence tendency of traffic accidents is considered to be different depending on the event situation in the vicinity of the road. For this reason, if it is possible to switch the accident occurrence pattern learning units 904A and 904B for performing the learning according to the event information, it is considered possible to execute the learning according to the tendency of the traffic accident to occur.

  Therefore, the event determination unit 903 according to the ninth embodiment is configured to switch the output destination of the traffic information based on the event information of the traffic information input from the traffic information management unit 901. For example, when the held status of the event based on the event information indicates that a traffic accident is likely to occur, the event determination unit 903 sets the traffic information output destination in the accident occurrence pattern learning unit 904A and the accident occurrence forecasting unit 905A. When the held status of the event based on the event information indicates a situation where a traffic accident is unlikely to occur, the output destination of the traffic information is set to the accident occurrence pattern learning unit 904B and the accident occurrence forecasting unit 905B.

  In the ninth embodiment, by being configured as described above, the same effects as those of the first to eighth embodiments can be obtained. That is, according to the ninth embodiment, the accuracy of learning by each of the accident occurrence pattern learning units 904A and 904B can be enhanced using event information, and the prediction accuracy by each of the accident occurrence prediction units 905A and 905B It can be enhanced.

Tenth Embodiment
Next, a traffic accident occurrence prediction device 1000 according to a tenth embodiment will be described with reference to FIG. The tenth embodiment is the same as the fifth and eighth embodiments in that information on the structure, shape, or state of a road is used as an example of the “first information”, but speed limit information is also available for road gradient information. However, it differs from the fifth and eighth embodiments in that hazard information (details will be described later) is used. In the following, portions of the tenth embodiment which are different from the fifth and eighth embodiments will be mainly described.

  As shown in FIG. 15, the traffic accident occurrence forecasting apparatus 1000 according to the tenth embodiment has, as a functional configuration, a traffic information management unit 1001, a traffic information DB 1002, a hazard determination unit 1003, an accident occurrence pattern learning unit 1004A and 1004B and accident occurrence forecasting units 1005A and 1005B. The accident occurrence pattern learning units 1004A and 1004B are examples of the “learning unit”, and the accident occurrence prediction units 1005A and 1005B are examples of the “forecast unit”. In addition, the hazard determination unit 1003 is an example of the “switching unit”.

  The traffic information management unit 1001 is configured to receive inputs of sensor information, control information, and hazard information as traffic information, and store the received sensor information, control information, and hazard information in the traffic information DB 1002 in association with each other. It is done.

  Hazard information is information that may affect the occurrence of a traffic accident, and the traffic information accumulated in the traffic information DB 1002 is divided into a plurality of traffic information according to the tendency of the traffic accident (at least the accident occurrence pattern learning unit It is information that can be classified into the same number as the number of 1004A and 1004B. As the hazard information, information on the state of the road based on a hazard map or the like, that is, information indicating the degree of danger of the road is set.

  Here, in general, in a portion where a rapid speed change of a vehicle is likely to occur, such as a curve or a sag portion provided on a road, a rear-end collision is more likely than a portion where speed change is less likely to occur. The probability of occurrence is high. For this reason, when forecasting at a location where speed change is likely to occur, it may be desirable to statistically analyze the number of accidents and the like and to consider hazard information in which dangerous locations are summarized. For this reason, if it is possible to switch the accident occurrence pattern learning units 1004A and 1004B for performing the learning according to the hazard information, it is considered possible to perform the learning according to the tendency of occurrence of the traffic accident.

  Therefore, the hazard determination unit 1003 according to the tenth embodiment is configured to switch the output destination of the traffic information based on the hazard information of the traffic information input from the traffic information management unit 1001. For example, when the degree of danger of the road based on the hazard information indicates the degree of danger that a traffic accident is likely to occur, the hazard determination unit 1003 sets the output destination of the traffic information to the accident occurrence pattern learning unit 1004A and the accident occurrence prediction unit 1005A. If the degree of risk of the road based on the hazard information indicates the degree of risk that a traffic accident is unlikely to occur, the output destination of the traffic information is set to the accident occurrence pattern learning unit 1004B and the accident occurrence forecasting unit 1005B. There is.

  In the tenth embodiment, by being configured as described above, the same effects as in the first to ninth embodiments can be obtained. That is, according to the tenth embodiment, the hazard information can be used to improve the learning accuracy of each of the accident occurrence pattern learning units 1004A and 1004B, and the prediction accuracy of each of the accident occurrence prediction units 1005A and 1005B It can be enhanced.

Eleventh Embodiment
Next, a traffic accident occurrence forecasting device 1100 according to the eleventh embodiment will be described with reference to FIG. The eleventh embodiment is the same as the fourth and sixth embodiments in that information on a vehicle traveling on a road is used as an example of the “first information”, but a large size information is used as information on a vehicle traveling on a road. The fourth embodiment is different from the fourth and sixth embodiments in that neither the vehicle mixing information nor the low-speed vehicle mixing information is used, and two-wheeled vehicle mixing information (details will be described later) is used. In the following, portions of the eleventh embodiment which are different from the fourth and sixth embodiments will be mainly described.

  As shown in FIG. 16, the traffic accident occurrence forecasting apparatus 1100 according to the eleventh embodiment has a traffic information management unit 1101, a traffic information DB 1102, a two-wheeled vehicle mixing determination unit 1103, and an accident occurrence pattern learning unit 1104A as a functional configuration. And 1104B and accident occurrence forecasting units 1105A and 1105B. The accident occurrence pattern learning units 1104A and 1104B are examples of the “learning unit”, and the accident occurrence prediction units 1105A and 1105B are examples of the “forecasting unit”. The two-wheeled vehicle mixture determination unit 1103 is an example of the “switching unit”.

  The traffic information management unit 1101 receives inputs of sensor information, control information, and motorcycle mixing information as traffic information, and stores the received sensor information, control information, and motorcycle mixing information in the traffic information DB 1102 in association with each other. Is configured as.

  The two-wheeled vehicle mixing information is information that may affect the occurrence of a traffic accident, and the traffic information accumulated in the traffic information DB 1102 is divided into a plurality of traffic information according to the tendency of the traffic accident (at least accident occurrence pattern learning Information which can be classified in the same number as the number of units 1104A and 1104B. As the two-wheeled vehicle mixing information, information on two-wheeled vehicles, more specifically, information such as a numerical value indicating the degree of mixing of two-wheeled vehicles on the road is set. The motorcycle mixture information may be created based on traffic information in the past or measurement results of an image-type vehicle sensor, or the controller may display an image of an ITV (Industrial Television) camera or the past. It may be input based on experience and the like.

  Here, vehicle sensors installed on roads are often of the ultrasonic type. Generally, unlike an image-type vehicle sensor, an ultrasonic-type vehicle sensor can not determine the vehicle type of a vehicle traveling on the road, so when a two-wheeled vehicle mixes in a vehicle traveling on the road, The number of motorcycles can not be measured. In other words, in the case of a combination of an ordinary car and a two-wheeler, the ultrasonic type vehicle sensor can measure only the number of normal cars, the vehicle speed, the vehicle density, etc. Deviation from the situation occurs. Therefore, it is desirable to consider the degree of mixing of the two-wheeled vehicle in learning the occurrence pattern of traffic accidents. For this reason, if the accident occurrence pattern learning units 1104A and 1104B that execute learning can be switched according to the two-wheeled vehicle mixture information, it is possible to perform learning with higher accuracy.

  Therefore, the two-wheeled vehicle mixing determination unit 1103 according to the eleventh embodiment is configured to switch the output destination of the traffic information based on the two-wheeled vehicle mixing information in the traffic information input from the traffic information management unit 1101. For example, when the mixing degree of the two-wheeled vehicle based on the two-wheeled vehicle mixing information indicates a value that a traffic accident is likely to occur, the two-wheeled vehicle mixing determination unit 1103 outputs the traffic information to the accident occurrence pattern learning unit 1104A and the accident occurrence forecasting unit 1105A. It is configured to set the output destination of the traffic information to the accident occurrence pattern learning unit 1104B and the accident occurrence forecasting unit 1105B when the mixing degree of the two-wheeled vehicle based on the two-wheeled vehicle mixing information indicates a value that a traffic accident hardly occurs. ing.

  In the eleventh embodiment, by being configured as described above, the same effects as in the first to tenth embodiments can be obtained. That is, according to the eleventh embodiment, the accuracy of learning by each of the accident occurrence pattern learning units 1104A and 1104B can be enhanced using the motorcycle mixture information, and the accuracy of the prediction by each of the accident occurrence prediction units 1105A and 1105B Can be enhanced.

(Twelfth embodiment)
Next, a traffic accident occurrence forecasting device 1200 according to a twelfth embodiment will be described with reference to FIG. The twelfth embodiment is similar to the fifth, eighth, and tenth embodiments in that information on the structure, shape, or state of a road is used as an example of the “first information”, but the structure of the road is In the fifth, eighth, and tenth embodiments and in that road maintenance information (details will be described later) which is not any of road gradient information, speed limit information, and hazard information is used as information on shape, shape, or state. Is different. In the following, portions of the twelfth embodiment which are different from the fifth, eighth, and tenth embodiments will be mainly described.

  As shown in FIG. 17, the traffic accident occurrence forecasting apparatus 1200 according to the twelfth embodiment has a traffic information management unit 1201, a traffic information DB 1202, a road maintenance determination unit 1203, and an accident occurrence pattern learning unit 1204 A as a functional configuration. And 1204B and accident occurrence forecasting units 1205A and 1205B. The accident occurrence pattern learning units 1204A and 1204B are examples of the “learning unit”, and the accident occurrence prediction units 1205A and 1205B are examples of the “forecasting unit”. In addition, the road maintenance determination unit 1203 is an example of the “switching unit”.

  The traffic information management unit 1201 receives inputs of sensor information, control information, and road maintenance information as traffic information, and stores the received sensor information, control information, and road maintenance information in the traffic information DB 1202 in association with each other. Is configured as.

  The road maintenance information is information that may affect the occurrence of traffic accidents, and the traffic information accumulated in the traffic information DB 1202 is divided into a plurality of traffic information according to the tendency of traffic accidents (at least accident occurrence pattern learning Information which can be classified into the same number as the number of units 1204A and 1204B. As the road maintenance information, information indicating the state of the road caused by road maintenance such as construction performed on the road, that is, the state of the road maintenance is set.

  Here, in general, on a route where it is often possible to drive only a driving lane or an overtaking lane for road maintenance, when the number of lanes decreases, the traffic volume measured by the vehicle detector is the number of lanes. It is different from normal time without decrease. For example, in the case of a road with two lanes on one side, when only the overtaking lane can be traveled, the vehicle running in the traveling lane changes lanes to the overtaking lane, and drivers with different traveling modes are mixed on one lane It will be done. In this case, since the method of setting the distance between vehicles and the traveling speed and the like are different from usual, the occurrence pattern of a rear-end collision and the like also differs. For this reason, if the accident occurrence pattern learning units 1204A and 1204B that execute learning can be switched according to the road maintenance information, it becomes possible to execute learning according to the tendency of occurrence of traffic accidents.

  Therefore, the road maintenance determination unit 1203 according to the twelfth embodiment is configured to switch the output destination of the traffic information based on the road maintenance information of the traffic information input from the traffic information management unit 1201. For example, when the road maintenance status based on the road maintenance information indicates a traffic accident is likely to occur, the road maintenance judging unit 1203 sends the traffic information output destination to the accident occurrence pattern learning unit 1204A and the accident occurrence forecasting unit 1205A. When the maintenance status of the road based on the road maintenance information indicates that traffic accidents are unlikely to occur, the traffic information output destination is set to the accident occurrence pattern learning unit 1204B and the accident occurrence forecasting unit 1205B. ing.

  In the twelfth embodiment, by being configured as described above, the same effects as in the first to eleventh embodiments can be obtained. That is, according to the twelfth embodiment, the accuracy of learning by each of the accident occurrence pattern learning units 1204A and 1204B can be enhanced using the road maintenance information, and the accuracy of the prediction by each of the accident occurrence prediction units 1205A and 1205B. Can be enhanced.

(13th Embodiment)
Next, a traffic accident occurrence forecasting device 1300 according to the thirteenth embodiment will be described with reference to FIG. The thirteenth embodiment is similar to the fifth, eighth, tenth and twelfth embodiments in that information on the structure, shape or state of the road is used as an example of the "first information". The fifth, the eighth, the eighth in that the stop information (details will be described later) which is not any of the road slope information, the speed limit information, the hazard information and the road maintenance information is used as the information regarding the structure, shape or state of This is different from the tenth and twelfth embodiments. In the following, portions of the thirteenth embodiment which are different from the fifth, eighth, tenth and twelfth embodiments will be mainly described.

  As shown in FIG. 18, the traffic accident occurrence forecasting device 1300 according to the thirteenth embodiment has a traffic information management unit 1301, a traffic information DB 1302, a blocking decision unit 1303, an accident occurrence pattern learning unit 1304 A and a functional configuration. And an accident occurrence forecasting unit 1305A and 1305B. The accident occurrence pattern learning units 1304A and 1304B are examples of the “learning unit”, and the accident occurrence prediction units 1305A and 1305B are examples of the “forecasting unit”. Further, the closing judgment unit 1303 is an example of the “switching unit”.

  The traffic information management unit 1301 is configured to receive inputs of sensor information, control information, and stop information as traffic information, and store the received sensor information, control information, and stop information in the traffic information DB 1302 in association with each other. It is done.

  Deadlock information is information that may affect the occurrence of a traffic accident, and the traffic information accumulated in the traffic information DB 1302 is divided into a plurality of traffic information according to the traffic accident tendency (at least the accident occurrence pattern learning unit It is information that can be classified into the same number as the number of 1304A and 1304B. As the closing information, information indicating the state of the road due to a sudden closing caused by an accident or equipment trouble or the like, that is, the status of the closing is set.

  Here, generally, on the road, traffic congestion often occurs in a time zone such as morning or evening. This traffic congestion is a natural traffic congestion that occurs due to traffic concentration caused by going to work or returning home, and when there is only such a natural traffic congestion, changes in traffic conditions result in traffic volumes gradually increasing and exceeding traffic volumes It will be more or less regular, as congestion will occur when the quantity is reached. However, in the case of a sudden traffic jam caused by a sudden traffic closure caused by an accident or equipment trouble, etc., unlike natural traffic congestion, changes in traffic conditions are often abrupt. For this reason, it is desirable to consider the presence or absence of sudden traffic closure etc in learning of the occurrence pattern of traffic accidents. That is, if it is possible to switch the accident occurrence pattern learning units 1304A and 1304B that execute learning in accordance with the closing information, it is possible to perform learning with higher accuracy.

  Therefore, the closing judgment unit 1303 according to the thirteenth embodiment is configured to switch the output destination of the traffic information based on the closing information of the traffic information input from the traffic information management unit 1301. For example, when the sudden stop condition based on the stop information indicates a state in which a traffic accident is likely to occur, the stop decision unit 1303 outputs the traffic information to the accident occurrence pattern learning unit 1304A and the accident occurrence forecast unit 1305A. If the situation of sudden closure based on the closure information indicates a situation where traffic accidents are unlikely to occur, the output destination of traffic information is set to the accident occurrence pattern learning unit 1304B and the accident occurrence prediction unit 1305B. It is done.

  In the thirteenth embodiment, the same effects as in the first to twelfth embodiments can be obtained by being configured as described above. That is, according to the thirteenth embodiment, the accuracy of learning by each of the accident occurrence pattern learning units 1304A and 1304B can be enhanced using the closing information, and the accuracy of the prediction by each of the accident occurrence prediction units 1305A and 1305B It can be enhanced.

Fourteenth Embodiment
Next, a traffic accident occurrence forecasting device 1400 according to a fourteenth embodiment will be described with reference to FIG. The fourteenth embodiment is the same as the fifth, eighth, tenth, twelfth, and thirteenth embodiments in that information on the road structure, shape, or state is used as an example of the "first information". However, road surface condition information (details will be described later), which is not any of road gradient information, speed limit information, hazard information, road maintenance information, and stoppage information, is used as information on the structure, shape, or state of the road. It differs from the fifth, eighth, tenth, twelfth and thirteenth embodiments. In the following, portions of the fourteenth embodiment which are different from the fifth, eighth, tenth, twelfth and thirteenth embodiments will be mainly described.

  As shown in FIG. 19, the traffic accident occurrence forecasting apparatus 1400 according to the fourteenth embodiment has a traffic information management unit 1401, a traffic information DB 1402, a road surface condition determination unit 1403, and an accident occurrence pattern learning unit 1404A as a functional configuration. And 1404B and accident occurrence forecasting units 1405A and 1405B. The road surface state determination unit 1403 is an example of a “switching unit”.

  The traffic information management unit 1401 receives inputs of sensor information, control information, and road surface state information as traffic information, associates the received sensor information, control information, and road surface state information with one another, and stores them in the traffic information DB 1402 Is configured.

  The road surface state information is information that may affect the occurrence of a traffic accident, and the traffic information accumulated in the traffic information DB 1402 is divided into a plurality of traffic information according to the tendency of the traffic accident (at least accident occurrence pattern learning Information which can be classified into the same number as the number of units 1404A and 1404B. As the road surface state information, information indicating the state of the road surface (road) such as whether the road surface (road) is dry, wet, or frozen is set.

  Here, in general, the braking distance of the vehicle changes in accordance with the state of the road surface. For example, when the road surface is wet due to rainfall or the like, it is generally known that the braking distance is longer than when the road surface is dry, so the driver tends to travel at a slower speed with the inter-vehicle distance taken. There is. Therefore, when the road surface is wet, a traffic accident may occur under a different traffic condition than when the road surface is dry. For this reason, it is desirable to consider the condition of the road surface in learning the occurrence pattern of traffic accidents. That is, if the accident occurrence pattern learning units 1404A and 1404B that execute learning can be switched according to the road surface state information, it is possible to perform learning with higher accuracy.

  Therefore, the road surface condition determination unit 1403 according to the fourteenth embodiment is configured to switch the output destination of the traffic information based on the road surface condition information of the traffic information input from the traffic information management unit 1401. For example, when the road surface state based on the road surface state information indicates a state in which a traffic accident is likely to occur, the road surface state determining portion 1403 sets the output destination of the traffic information to the accident occurrence pattern learning portion 1404A and the accident occurrence forecasting portion 1405A. When the road surface state based on the road surface state information indicates a state in which a traffic accident is unlikely to occur, the output destination of the traffic information is configured to be set in the accident occurrence pattern learning unit 1404B and the accident occurrence forecasting unit 1405B. .

  In the fourteenth embodiment, the same effects as in the first to thirteenth embodiments can be obtained by being configured as described above. That is, according to the fourteenth embodiment, the accuracy of learning by each of the accident occurrence pattern learning units 1404A and 1404B can be enhanced using road surface state information, and the accuracy of the prediction by each of the accident occurrence forecasting units 1405A and 1405B. Can be enhanced.

(Fifteenth embodiment)
Next, a traffic accident occurrence forecasting device 1500 according to the fifteenth embodiment will be described with reference to FIG. The fifteenth embodiment is the same as the seventh embodiment in that information on traffic is used as an example of the “first information”, but it is not traffic information in the interchange but traffic jam information as information on traffic. The seventh embodiment differs from the seventh embodiment in that (the details will be described later) is used. In the following, portions different from the seventh embodiment of the fifteenth embodiment will be mainly described.

  As shown in FIG. 20, the traffic accident occurrence forecasting apparatus 1500 according to the fifteenth embodiment has a traffic information management unit 1501, a traffic information DB 1502, a traffic congestion determination unit 1503, and an accident occurrence pattern learning unit 1504A as a functional configuration. 1504 C and accident occurrence prediction units 1505 A to 1505 C. The accident occurrence pattern learning units 1504A to 1504C are examples of the “learning unit”, and the accident occurrence prediction units 1505A to 1505C are an example of the “forecasting unit”. The traffic congestion determination unit 1503 is an example of the “switching unit”.

  The traffic information management unit 1501 is configured to receive inputs of sensor information, control information, and traffic jam information as traffic information, and store the received sensor information, control information, and traffic jam information in the traffic information DB 1502 in association with each other. It is done.

  The traffic jam information is information that may affect the occurrence of traffic accidents, and the traffic information accumulated in the traffic information DB 1502 is divided into a plurality of traffic information according to the traffic accident tendency (at least the accident occurrence pattern learning unit 1504A to 1504C) can be classified. As the traffic jam information, information is set that indicates the degree of traffic jam, such as the length (small or large) of the traffic jam. Congestion is related to the traffic volume on the road, the density of vehicles on the road, the speed of vehicles traveling on the road, and the like.

  Here, in general, on highways, vehicle detectors are often installed every 1 to 2 km, and traffic congestion that occurs in the morning, in the evening, and long holidays is divided into a plurality of sections where vehicle detectors are installed. It will be straddling. When an accident occurrence forecast is performed for each section in which a vehicle detector is installed, congestion etc. generated in each section can be taken into consideration, but in a situation where traffic congestion occurs across multiple sections When performing an accident occurrence forecast of the target section, it is conceivable that the downstream section is affected by the traffic congestion. Therefore, when there is a large traffic jam that spans a plurality of sections, it is desirable to forecast the occurrence of an accident not on every section but on the entire sections. For this reason, it is desirable to consider the degree of congestion, etc. in learning of the occurrence pattern of traffic accidents. That is, if the accident occurrence pattern learning units 1504A to 1504C that execute learning can be switched according to the traffic congestion information, learning with higher accuracy can be performed.

  Therefore, the traffic congestion determination unit 1503 according to the fifteenth embodiment is configured to switch the output destination of the traffic information based on the traffic congestion information of the traffic information input from the traffic information management unit 1501. For example, when the traffic jam information indicates that long traffic congestion is occurring, the traffic congestion judging unit 1503 sets the output destination of the traffic information in the accident occurrence pattern learning unit 1504A and the accident occurrence forecasting unit 1505A, and the traffic congestion information is about medium. If the traffic information output destination is set in the accident occurrence pattern learning unit 1504B and the accident occurrence forecasting unit 1505B, and the traffic congestion information indicates that traffic congestion is occurring, The output destination of the information is configured to be set in the accident occurrence pattern learning unit 1504C and the accident occurrence prediction unit 1505C.

  In the fifteenth embodiment, the same effects as in the first to fourteenth embodiments can be obtained by being configured as described above. That is, according to the fifteenth embodiment, it is possible to improve the learning accuracy by each of the accident occurrence pattern learning units 1504A to 1504C using the traffic congestion information, and the prediction accuracy by each of the accident occurrence prediction units 1505A to 1505C It can be enhanced.

  The above first to fifteenth embodiments can also be combined. That is, two or more of the 15 different "first information" exemplified in the above first to fifteenth embodiments are combined, and a plurality of "one or more" are selected according to the two or more "first information". The learning unit "and a plurality of" forecasting units "may be switched.

  The computer program executed in the first to fifteenth embodiments described above is provided as a computer program product in an installable format or an executable format. That is, the computer program executed in the first to fifteenth embodiments is non-temporary and computer readable such as CD-ROM, flexible disk (FD), CD-R, DVD (Digital Versatile Disk), etc. Provided in a state included in a computer program product having a recording medium.

  In addition, the computer programs executed in the above first to fifteenth embodiments may be provided or distributed via a network while being stored in a computer connected to a network such as the Internet. The computer program executed in the first to fifteenth embodiments may be provided in a state of being incorporated in advance in a ROM or the like.

  As mentioned above, although embodiment of this invention was described, the said embodiment is an example to the last, and limiting the scope of invention is not intended. The above embodiments can be implemented in various forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. The embodiments described above are included in the scope and the gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500 Traffic accident occurrence forecasting device (accident occurrence forecasting device)
104A, 104B, 204B, 304A, 304B, 404A, 404B, 504A, 504B, 604A, 604B, 704A, 704B, 804A, 904B, 904B, 1004A, 1004B, 1104B, 1104A, 1204B, 1204A, 1304. 1304B, 1404A, 1404B, 1504A, 1504B, 1504C accident occurrence pattern learning unit (learning unit)
105A, 105B, 205A, 205B, 305A, 305B, 405A, 405B, 505A, 505B, 605A, 605B, 705A, 705B, 805A, 805B, 905A, 905B, 1005A, 1005B, 1105A, 1105B, 1205A, 1205B, 1305A, 1305B, 1405A, 1405B, 1505A, 1505B, 1505C Accident occurrence forecasting unit (forecasting unit)
103 Time zone determination unit (switching unit)
203 Day of the week judgment unit (switching unit)
303 Weather judgment unit (switching unit)
403 Large-sized car mixture judgment unit (switching unit)
503 Road slope judgment unit (switching unit)
603 Low-speed Vehicle Contamination Determination Unit (Switching Unit)
703 Traffic Determination Unit (Switching Unit)
803 Speed limit judgment unit (switching unit)
903 Event determination unit (switching unit)
1003 Hazard judgment unit (switching unit)
1103 Motorcycle Contamination Determination Unit (Switching Unit)
1203 Road Maintenance Determination Department (Switching Department)
1303 Dead end judgment unit (switching unit)
1403 Road Surface Condition Determination Unit (Switching Unit)
1503 Traffic congestion judgment unit (switching unit)

Claims (12)

Multiple learning units based on a self-organizing map that performs learning of traffic accident occurrence patterns using past traffic information;
A plurality of forecasting units provided one by one corresponding to each of the plurality of learning units so as to configure the plurality of learning units and a plurality of combinations, each of the plurality of forecasting units being the plurality of the plurality of forecasting units A plurality of forecasting units for outputting a forecast indicating the possibility of occurrence of a traffic accident based on the learning result by each of the learning units of
1 is the first information included in the traffic information, and the learning can be performed using the first information including the information related to the weather, which can classify the traffic information into a plurality of types according to the occurrence tendency of the traffic accident 1 one of the learning unit and provided with said switch a combination of one forecast unit to output the forecast between said plurality of combinations of the plurality of the learning unit and the plurality of prediction unit switching unit, accident prediction means. The accident occurrence prediction device according to claim 1, wherein the first information further includes information on time. The accident occurrence prediction device according to claim 1, wherein the first information further includes information on a vehicle traveling on a road. The accident occurrence prediction device according to claim 1, wherein the first information further includes information on a road structure, a shape, or a state. The accident occurrence prediction device according to claim 1, wherein the first information further includes information on traffic volume of a road. The accident occurrence forecasting device according to claim 1, wherein the first information further includes information on a situation around a road. Performing learning of a traffic accident occurrence pattern using past traffic information by a plurality of learning units based on a self-organizing map ;
By each of a plurality of forecasting units provided one by one corresponding to each of the plurality of learning units so as to configure the plurality of learning units and a plurality of combinations, the learning result by each of the plurality of learning units is Outputting a forecast indicating the possibility of occurrence of a traffic accident based on
A first information included in the traffic information, which can be classified into a plurality in accordance with the traffic information generating trend of the traffic accident by using the first information including information on weather, 1 to execute the learning one of the learning unit and and a switching the combination of one forecast unit for outputting the forecast between said plurality of combinations of the plurality of the learning unit and the plurality of prediction unit, accident forecasting method. Further using information on time as the first information, the one of the learning unit and switches the combination of the one forecast unit, accident forecasting method according to claim 7. Information about a vehicle traveling on a road further used as the first information, the one of the learning unit and switches the combination of the one forecast unit, accident forecasting method according to claim 7. Structure of the road, the shape or the information about the state further be used as the first information, said one of the learning unit and switches the combination of the one forecast unit, accident forecasting method according to claim 7. Information about traffic road further used as the first information, the one of the learning unit and switches the combination of the one forecast unit, accident forecasting method according to claim 7. Information about the status of the peripheral roads further used as the first information, the one of the learning unit and switches the combination of the one forecast unit, accident forecasting method according to claim 7.

Images