JP4279790B2 - Car number recognition device - Google Patents

Description

  The present invention relates to a technology of a vehicle number recognition device that recognizes a license plate of a vehicle using a television camera installed on a road or the like.

  2. Description of the Related Art A vehicle number recognition device that installs a TV camera on a road or the like and automatically recognizes a car license plate using an image captured by the TV camera has become widespread. This vehicle number recognition device is used by adjusting the lens of a camera to an appropriate focus, or in the case of a zoom lens, adjusting to an optimum magnification. When adjusting the lens of the TV camera, if the lane is restricted, a stationary vehicle can be taken as an imaging target, so that the lens can be adjusted easily. However, for reasons such as reduction in adjustment costs, lens adjustment is often performed without lane restrictions.

  And when adjusting a lens, without performing lane restrictions, since the vehicle which is drive | working on the road becomes an imaging object, time and effort are needed for focus adjustment. Specifically, when lens adjustment is performed for a moving vehicle as a subject to be photographed, the video at the moment of passing the vehicle is captured as a still image, and whether the subject is in focus by viewing the still image. Adjust the lens while checking.

  2. Description of the Related Art Conventionally, a vehicle number recognition device that captures an image of a traveling vehicle and automatically performs focus adjustment is known (for example, Patent Document 1). According to Patent Document 1, since the apparatus automatically performs the focus adjustment, it is possible to reduce labor compared to the focus adjustment performed by a person. Specifically, in Patent Document 1, a vehicle detection sensor is used to capture an image of the front portion of the vehicle, and the image is used to check the swelling of the character size of the license plate, thereby controlling the focus. In Patent Document 1, since the image capturing position is made constant by the vehicle detection sensor, the focus can be controlled to some extent by the degree of character swelling.

JP-A-10-105873

  However, the technique of Patent Document 1 has the following problems. Specifically, in Patent Document 1, the focus of a camera lens is controlled every time one vehicle is imaged. For this reason, when the data indicating the degree of expansion of characters becomes abnormal for some reason, there arises a problem that the focus is set to a completely different position.

  Further, in recent car number recognition devices, a type of device that automatically detects the front surface of a vehicle by image recognition technology and recognizes a license plate character from the image without using a vehicle detection sensor has become mainstream. ing. This is because the cost of the entire system can be reduced by not using the vehicle detection sensor. In addition, the vehicle detection sensor is difficult to detect a vehicle on a road having a plurality of lanes.

  The vehicle number recognition device that does not use the vehicle detection sensor automatically detects the front portion of the vehicle by image processing technology and recognizes the characters on the license plate from the image. And, in terms of image processing performance, the capture positions for several frames of the TV camera vary. For example, a license plate is photographed at the top of the screen, or a license plate is photographed at the bottom of the screen. That is, when the front surface of the vehicle is detected by the image processing technique, the position of the image to be captured changes every time. For this reason, when the focus is controlled based on the degree of expansion of characters in the captured image as in the technique described in Patent Document 1, the next image may not be focused. For example, when focus adjustment is performed with an image obtained by capturing a license plate at the top of the screen and the next vehicle is captured at the bottom of the screen, there is a high possibility that the next vehicle will not be in focus.

  Furthermore, there are vehicles that are attached with the number plate tilted (tilted forward). With the technique of Patent Document 1, when capturing an image of a mounted vehicle with the license plate tilted (tilted forward), the character height is particularly reduced, and the character size bulge is accurately determined. I can't. In such a case, in Patent Document 1, the focus position may be set to a completely different position.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to optimize the lens position of a camera in a car number recognition device that automatically recognizes a car license plate using an image captured by a television camera. It is to automatically adjust to the correct position.

  In order to solve the above problems, one embodiment of the present invention is applied to a vehicle number recognition device that receives image data from a television camera that captures a vehicle, processes the received image data, and recognizes a vehicle license plate. The Here, the television camera is provided with an electric lens that receives a control signal from the outside and sets the lens position.

  And the said vehicle number recognition apparatus detects a vehicle using the said image data received from the lens control means which transmits a control signal to the said electric lens, and sets the lens position of the said electric lens, and the said television camera, Recognizing means for recognizing the detected license plate of the vehicle, and when the recognizing means recognizes a predetermined number of license plates, the number of vehicles detected while recognizing the predetermined number of license plates; A means for calculating a character recognition rate using the above, a means for storing the calculated character recognition rate as a history, a means for determining whether or not the lens position of the electric lens is in an optimum position, and the lens position Means for setting the lens position of the electric lens to the optimum lens position via the lens control means. The determination means compares the character recognition rate calculated when recognizing the predetermined number of license plates with the history of the stored character recognition rate, and the calculated character recognition rate is the history. If the value is equal to or smaller than the predetermined value, it is determined that the lens position is not in an optimum position.

  According to the present invention, when the vehicle number recognition device recognizes a predetermined number of license plates, the recognition result obtained by the recognition processing of the predetermined number of license plates is compared with the past recognition history, and the recognition result and the recognition history are When the difference is more than a predetermined value, the lens of the camera is automatically adjusted to an optimum state. Therefore, according to this embodiment, the lens of the camera can be automatically adjusted to an optimal position while performing the license plate recognition process in the vehicle number recognition device.

  A vehicle number recognition system to which an embodiment of the present invention is applied will be described below with reference to the drawings.

  FIG. 1 is a functional block diagram of a vehicle number recognition system according to an embodiment of the present invention.

  As shown in the figure, the vehicle number recognition system includes an imaging unit 100 having a television camera 1 and a lens 2, and an image processing device 3. An external device 12 is connected to the image processing apparatus 3. The lens 2 is an electric lens that receives a control signal from the outside and sets the lens position.

  The image processing apparatus 3 includes an image input unit 4, a plate recognition unit 5, a focus analysis unit 6, a zoom analysis unit 7, a lens information analysis unit 8 having an abnormality monitoring unit 13, a lens control unit 9, a lens adjustment schedule management unit 10, And a communication unit 11.

  The image input unit 4 receives an image captured by the television camera 1 and outputs the received image to the plate recognition unit 5. The plate recognition unit 5 receives the image output from the image input unit 4 and automatically processes the received image to recognize the license plate automatically. The plate recognition unit 5 outputs information on the recognized license plate to the focus analysis unit 6 and the zoom analysis unit 7.

  The focus analysis unit 6 uses the information indicating the license plate from the plate recognition unit 5 to analyze the out-of-focus state of the characters of the recognized license plate, and outputs the analysis result to the lens information analysis unit 8. The zoom analysis unit 7 uses the information indicating the license plate from the plate recognition unit 5 to analyze the zoom magnification from the size of the plate and characters, and outputs the analysis result to the lens information analysis unit 8.

  The lens information analysis unit 8 obtains an optimum lens setting value using the analysis result output from the focus analysis unit 6 or the analysis result output from the zoom analysis unit 7. Further, the abnormality monitoring unit 13 provided in the lens information analysis unit 8 monitors whether or not the lens position (zoom position and focus position) of the lens 2 is normal. When it is determined that the lens position is not normal, the lens information analysis unit 8 performs processing for obtaining an optimal lens setting value.

  The lens control unit 9 controls the lens 2 using the lens setting value obtained by the lens information analysis unit 8. Specifically, the lens control unit 9 generates a control signal using the set value of the lens and transmits the control signal to the lens 2. The lens 2 sets the lens position according to a control signal from the lens control unit 9.

  The lens adjustment schedule management unit 10 determines at which timing the lens is adjusted. Specifically, the lens adjustment schedule management unit 10 holds data indicating lens adjustment timing. The lens adjustment schedule management unit 10 causes the focus analysis unit 6, the zoom analysis unit 7, and the lens information analysis unit 8 to perform processing for lens adjustment according to the timing determined by the stored data.

  The communication unit 11 transmits / receives data to / from the external device 12. For example, the communication unit 11 transmits information indicating the recognized license plate to the external device 12. Note that data transmission / reception performed by the communication unit 11 may be wired or wireless.

  Next, the hardware configuration of the image processing apparatus 3 according to the present embodiment will be described with reference to FIG. FIG. 17 is a diagram illustrating a hardware configuration of the image processing apparatus 3 according to the embodiment of this invention.

  As illustrated, the image processing apparatus 3 includes an image processing circuit 200, a CPU (Central Processing Unit) 201, a memory 202, an I / O circuit 203, an A / D conversion circuit 204, and a network interface (network I / F) 205. Have

  The I / O circuit 203 controls data input / output processing performed between the image processing apparatus 3 and the imaging unit 100. The A / D conversion circuit 204 receives analog image data from the television camera 1 via the I / O circuit 203 and converts the received analog image data into digital image data. The network I / F 205 controls data communication performed with the external device 12. The image processing circuit 200 receives the digital image data converted by the A / D conversion circuit 204 and performs image processing on the digital image data. The memory 202 stores a program for realizing the processing performed by the image processing apparatus 3, digital image data to be processed, and the like. The CPU 201 controls predetermined arithmetic processing and operation of the image processing apparatus by executing a program stored in the memory.

  Specifically, the function of the image input unit 4 described above is realized by the I / O circuit 203 and the A / D conversion circuit 204. The function of the plate recognition unit 5 is realized by a program (plate recognition program) executed by the image processing circuit 200 and the CPU 201. The functions of the focus analysis unit 6, zoom analysis unit 7, lens information analysis unit 8, and lens adjustment schedule management unit 10 are the programs executed by the CPU 201 (focus analysis program, zoom analysis program, lens information analysis program, lens adjustment). (Schedule management program). The function of the lens control unit 9 is realized by a program (lens control program) executed by the I / O circuit 203 and the CPU 201. The function of the communication unit 11 is realized by a program (communication program) executed by the network I / F 205 and the CPU 201.

  Here, before specifically describing the function of each part of the image recognition device 3 of the present embodiment, the configuration of the license plate to be recognized and the shooting state of the vehicle will be described with reference to FIGS.

  FIGS. 2A and 2B are diagrams illustrating a license plate of a vehicle. FIG. 2A shows a medium plate and FIG. 2B shows a large plate.

  As shown in the drawing, each of the medium-sized plate and the large-sized plate includes a serial number 20, a land transportation branch office code 21, a vehicle type code 22, and a use code 23. The size of each character constituting the interval of the serial number 20, the serial number 20, the land transportation branch office code 21, the vehicle type code 22, and the application code 23 is uniquely determined. In addition, it is determined that a license plate of a business vehicle has a green background, and a license plate of a private vehicle has a white background plate (light vehicles are determined separately). Then, all characters on the illustrated license plate can be automatically read by the conventional image recognition processing.

  FIG. 3 is a diagram for explaining a shooting state of the vehicle imaged by the vehicle number recognition system of the present embodiment. The camera 1 is installed on the road 31 or on the road side and takes a bird's-eye view of the vehicle 32. In a vehicle number recognition system that uses a vehicle detection sensor separately, an image of a location at the same distance L can always be captured by the sensor. However, in this embodiment, vehicle detection is performed by image recognition technology without using a sensor. Therefore, the distance L from the camera 1 to the place where the vehicle 32 is photographed and the visual field width Wd of the place slightly change.

FIG. 4 is a diagram illustrating an image taken by the camera 1. The visual field width (visual field size Wd) is determined by the distance L between the camera 1 and the license plate 33 of the vehicle 32. Here, the position (plate position) from the image bottom side 34 of the license plate in the captured image is YP. The plate position YP varies depending on the image capturing timing and the mounting height of the plate. When an image is captured by the vehicle detection sensor, the plate is basically photographed at substantially the same position (in this case, YP changes by the change in height at which the plate is attached).
When the camera 1 and the lens 2 are used in a system that captures an image using a vehicle detection sensor, it may be possible to adjust the focus to the distance L before shipping. However, strictly speaking, the height of the position where the camera 1 is installed and the installation direction of the camera change when the camera is installed. For this reason, work for adjusting the focus and zoom occurs on site. In this case, it is difficult to accurately focus because the vehicle is passing. Even if the camera is focused in the daytime, the focus is slightly shifted at night.
Therefore, in this embodiment, as will be described later, by automatically adjusting the focus position and zoom position of the lens while recognizing the license plate, it is possible to reduce the trouble of lens adjustment and improve the recognition rate of the license plate. I have to.
It is also conceivable to use a lens having a general autofocus function in the vehicle number recognition system. The autofocus function adjusts the focus so that the edge of the entire screen or the designated area becomes the largest. However, when the license plate is to be recognized by the autofocus function, the following problems occur. Specifically, when recognizing the license plate, the vehicle passes instantaneously in a state where the road image is initially focused. Therefore, with the autofocus function, the movement of focusing on the position of the license plate is not in time, and the recognition performance cannot be improved. Therefore, in the present embodiment, focus adjustment specialized for license plate recognition is performed.

Based on such preconditions, returning to FIG. 1, the operation of each unit of the image processing apparatus 3 will be described in detail.
The plate recognition unit 5 receives the image data from the camera 1 via the image input unit 4, performs image processing on the image data, and detects the front portion of the vehicle. The plate recognition unit 5 detects the image of the detected license plate on the front surface of the vehicle, and recognizes at least the serial number 20 (see FIG. 2) from the detected license plate image. Further, the plate recognition unit 5 obtains the position of the license plate on the screen (plate position YP) and the position (coordinates) of the serial number 20 using the received image. Further, the plate recognition unit 5 calculates a character recognition rate (number of serial numbers recognized / number of detected vehicles). Note that the recognition processing of the serial number 20, the land transportation branch office code 21, the vehicle type code, and the application code 23 by the plate recognizing unit 5 is realized by a conventional technique (for example, a configuration described in “JP-A-6-2159243”). The

  Then, the plate recognition unit 5 sends the obtained plate position YP, the position of the serial number 20, the character recognition rate, the position of the serial number, and data including the target image to the focus analysis unit 6. Further, the plate recognition unit 5 outputs the obtained plate position YP and the detected license plate image to the zoom analysis unit 7.

  The focus analysis unit 6 evaluates the degree of blurring of the edge of the serial number. Specifically, the focus analysis unit 6 uses the data received from the plate recognition unit 5 to calculate an evaluation value that quantifies the blurring state of the edge of the serial number. Then, the focus analysis unit 6 transmits the calculated evaluation value to the lens information analysis unit 8.

  The zoom analysis unit 7 obtains the pitch of the serial number of the license plate using the target image transmitted from the plate recognition 5 and calculates the visual field size from the pitch. The zoom analysis unit 7 transmits the calculated visual field size and the received plate position YP to the lens information analysis unit 8.

  Then, the lens information analysis unit 8 uses the evaluation value received from the focus analysis unit 6 to obtain an optimum focus position, and sets the lens 2 to the obtained focus position via the lens control unit 9. The lens information analysis unit 8 obtains an optimum zoom position using the field size and plate position YP received from the zoom analysis unit 7, and obtains the lens 2 via the lens control unit 9. Set to.

  Next, focus control processing according to the present embodiment will be described.

First, the plate recognition unit 5 detects a license plate using an image from the camera 1, and cuts out the serial number 20 of the detected license plate. The plate recognition unit 5 obtains a rectangular area for each character of the cut out serial number 20. The plate recognition unit 5 transmits the “coordinates” and “target image” of the obtained rectangular area to the focus analysis unit 6. The focus analysis unit 6 evaluates the degree of blur of the serial number using the “coordinates” and “target image” of the transmitted rectangular area.
Specifically, the focus analysis unit 6 receives the “coordinates” and “target image” of the rectangular area transmitted by the plate recognition unit 5. The focus analysis unit 6 uses the received “coordinates” and “target image” to obtain a density projection distribution 41 on the section AA ′ plane of the region 40 for one character of the serial number 20 as shown in FIG. . The focus analysis unit 6 performs a differentiation process on the obtained density distribution 41 to obtain a contour waveform 42a of the character. For this contour waveform 42a, the width (d (1), d (2), d (3) to d (n)) larger than the threshold value th is obtained, and the average value is calculated to obtain the character width. Quantify the blur condition.

  FIG. 6 is a diagram illustrating a state where the “target image” in the rectangular area is in focus. When the “target image” in the rectangular area is in focus, the contour waveform 42 b is sharp. Therefore, the width (d (1), d (2), d (3) to d (n)) of the character outline waveform 42b obtained from the “target image” when the image is in focus is in focus. The value is smaller than the width (d (1), d (2), d (3) to d (n)) of the contour waveform 42a of the character obtained from the “target image” in the absence.

FIG. 5 shows the waveform of the AA ′ cross section obtained by cutting the processing area 40 along the line AA ′. For example, the contour waveform 42 of the cross section for N lines in the area 40 is obtained. (N is an integer of 1, 2, 3,... N). Also in this case, the plate recognizing unit 5 obtains a peak width (d (1), d (2), d (3) to d (n)) larger than the threshold value th for each contour waveform 42. , Average the values.
When an image of one vehicle is processed, since there are four serial numbers, edge information for four characters (the number of characters is known by the image recognition processing of the plate recognition unit 5) is obtained. The focus analysis unit 6 averages the edge information for four characters. The obtained average value of the width d (i) is “1, 2, 3,..., 5, 6”, which is a value that increases as the focus shifts. The focus analysis unit 6 normalizes the average value of the obtained width d (i), and sets the value obtained by normalization as the edge sharpness E. The focus analysis unit 6 normalizes the average value of the obtained width d (i) to “0 to 1” using, for example, (Equation 1) shown below. When the value “E” obtained by (Equation 1) is 0 or less, the value of “E” is set to 0. Also, β shown in (Equation 1) is a preset constant, and uses a numerical value indicating the edge width (d (i)) when the focus is greatly shifted (for example, β is about 10). ).
E = (β−average value of edge width d (i)) / β (Expression 1)
The edge sharpness E obtained by (Equation 1) approaches 1 when the subject is in focus, and approaches 0 when the subject is not in focus. The focus analysis unit 6 sends the edge sharpness E, which is the evaluation value, to the lens information analysis unit 8. The focus analysis unit 6 also sends the plate position YP received from the plate recognition unit 5 to the lens information analysis unit 8 when transmitting the edge sharpness E to the lens information analysis unit 8.
The lens information analysis unit 8 uses the plurality of edge sharpness E output from the plate recognition unit 5 and the plate position YP of the license plate to obtain the edge sharpness corresponding to the center of the screen (hereinafter, the edge sharpness corresponding to the center of the screen). The degree is referred to as “Ec”). Thus, the reason why the edge sharpness Ec corresponding to the center of the screen is calculated is that it is considered that the capturing position changes according to the timing of photographing the vehicle, and the edge width also changes accordingly.

Specifically, the lens information analysis unit 8 receives a plurality of plate positions YP and edge sharpness E output from the focus analysis unit 6. The lens information analysis unit 8 uses the received data to obtain a relational expression (linear regression y = ax + b or “linear position y = ax + b”, or “plate position YP” as “x” coordinates and “edge sharpness E” as “y” coordinates). A quadratic curve regression y = ax ² + bx + c) is obtained. For example, the lens information analysis unit 8 obtains the linear regression (y = ax + b) 70 illustrated in FIG. Then, the lens information analysis unit 8 uses the obtained linear regression (y = ax + b) 70 to calculate “edge sharpness Ec” where “plate position YP” is the center of the screen. The plurality of units described above is about 10 units (usually can be collected in about 1 to 3 minutes). If the number of license plates that can be recognized is small, it cannot be obtained by linear regression. In that case, the simple average value of the sample data may be the edge sharpness Ec with respect to the center of the screen.
The vehicle number recognition system of this embodiment calculates the edge sharpness Ec at the center of the screen for each focus position while moving the focus position of the lens 2. That is, the edge sharpness Ec at the center of the screen is obtained when the lens 2 is set at a certain focus position.

  Here, a case where the focus position of the lens 2 can be moved to 10 different positions (positions 0 to 9) will be described as an example. When the focus position moves from 0 to 9 different positions, the character recognition rate and edge sharpness Ec at the focus position 0 are obtained, and the same processing is performed at the next focus position. This operation is repeated at 0 to 9 focus positions. The lens information analysis unit 8 stores the edge sharpness Ec obtained at the focus position in association with each focus position.

  Further, the lens information analysis unit 8 stores the character recognition rate of the serial number at the focus position in association with each focus position. And since the character recognition rate is at least a predetermined threshold value (for example, 90%) at a focused position, the focus position with the highest edge sharpness is specified within the range. The lens information analysis unit 8 sets the focus position of the lens 2 to the specified focus position via the lens control unit 9.

  Here, the process of determining the focus position of the lens information analysis unit 8 will be described with reference to FIG. FIG. 8 is a diagram for explaining a focus position selection process performed by the lens information analysis unit 8.

  The lens information analysis unit 8 holds “data in which the edge sharpness E is associated with each focus position” and “data in which the character recognition rate is associated with each focus position”. Then, the lens information analysis unit 8 uses the “data in which the character recognition rate is associated with each focus position” to identify a focus position with a character recognition rate equal to or greater than a threshold value. In the illustrated example, four focus positions F (3) to F (6) are specified. Next, the lens information analysis unit 8 uses the “data in which the edge sharpness E is associated with each focus position” to associate the edge sharpness Ec with the highest value among the four specified focus positions 80. Specify the focus position. The lens information analysis unit 8 determines the identified focus position as the optimum focus position. In the illustrated example, the focus position of F (5) is selected as the optimum focus position.

When the character recognition rate is equal to or less than a predetermined value, it is not necessary to obtain the edge sharpness Ec. Therefore, the lens information analysis unit 8 may not calculate the edge sharpness Ec when the recognition rate is equal to or less than a predetermined value. In this way, it is possible to move quickly to the next focus position. Since the edge sharpness E changes depending on the size of the character, it is possible to improve the accuracy by judging the type of large and medium plates and using the edge information of only the medium plates.
In this embodiment, as shown in FIG. 5, the out-of-focus state (in-focus state) is evaluated by the sharpness E of the vertical edge of the serial number of the license plate, but this is particularly limited. It is not a thing. For example, a horizontal edge may be used instead of the vertical edge. In addition, pattern matching and neural networks are often used for character recognition, but it is also possible to evaluate whether or not the image is in focus using the degree of matching and the certainty obtained during character recognition.
Here, an example of using the certainty factor obtained at the time of character recognition will be described with reference to FIG. FIG. 9 is a diagram for explaining the certainty of character recognition when the image of the land transportation branch office code of the license plate is cut out. The certainty of character recognition is data indicating the recognition accuracy of “0 to 1” output when the plate recognition unit 5 performs character recognition processing, and is set to “1” when the focus is on. Get closer. The certainty factor approaches “0” when the focus is not achieved. 9A shows a state where the “target image” is in focus, and FIG. 9B shows a state where the “target image” is not in focus.

  For example, when the “target image” is in focus as shown in FIG. 9A, the certainty factor is a high value of “0.95”. Also, as shown in FIG. 9B, when the “target image” is not focused (when the focus is out of focus), the value is lower than when the focus is focused as “0.75”. The certainty of Replacing this certainty factor with the edge sharpness degree E shown in FIG. 8 makes it possible to select the optimum focus position as described above.

  In addition, the name of the license plate changes in quality as the quality (dirt, bend, etc.) changes. For this reason, the certainty factor cannot be used as it is for the focus evaluation value. Therefore, for example, when a camera is installed in Mito City, Ibaraki Prefecture, the number of vehicles that pass is “Mito”, so it is stable by averaging using the confidence level when outputting “Mito”. It can be used as a focus evaluation value. Further, when the certainty factor is used as the focus evaluation value, only the certainty factor when the specific character of the serial number, for example, the numeral 5 is recognized, may be averaged instead of the land transportation branch office code. Further, the focus evaluation value may be a multiplication value of at least two certainty factors among the serial number of the license plate, the land transportation branch office code, the vehicle type code, and the application code. For example, you may make it use the multiplication value of the reliability of the above-mentioned land transportation branch office code and the reliability of a serial number (applicable to FIG. 8 if normalized to a value of 0 to 1). Further, the certainty factors of all characters may be summed and averaged, and the average value may be used as the focus evaluation value.

  In addition, although it is possible to use the use code of a license plate for calculation of a certainty factor, use of a use code is not preferable for the following reasons. Specifically, since the use code is almost different for each passing car, it is considered that it takes too much time to accumulate the data of the predetermined character and the adjustment time increases.

  Subsequently, the flow of the adjustment process of the focus position of the lens 2 will be described.

  FIG. 10 is a diagram for explaining a flow of processing for adjusting the focus position of the lens 2 performed by the present embodiment.

  First, a table used for statistical processing held by each unit of the image processing apparatus 3 is initialized (S50). In addition, the lens information analysis unit 8 sets the focus position of the lens 2 to the initial value “cn = 0” via the lens control unit 9 (S51).

  In this state, the plate recognition unit 5 receives the image data from the camera 1 via the image input unit 4. The plate recognition unit 5 detects the vehicle from the received image and recognizes the detected license plate of the vehicle. Moreover, the plate recognition part 5 calculates plate position YP of the detected license plate of the vehicle (S52-53). In S52, the plate recognition unit 5 counts the number of vehicles detected for each focus position and the number of times the number plate has been successfully recognized, and holds the value in association with the focus position.

  Subsequently, the focus analysis unit 6 acquires the license plate image of the vehicle recognized from the plate recognition unit 5 and the calculated plate position YP. The focus analysis unit 6 uses the acquired license plate image of the vehicle to obtain the edge sharpness E of the serial number according to the above-described procedure (S54). Further, the focus analysis unit 6 transmits the acquired plate position YP and the obtained edge sharpness E to the lens information analysis unit 8.

  Thereafter, when the lens information analysis unit 8 receives the plate position YP and the edge sharpness E output from the focus analysis unit 6, whether or not the serial number recognition processing of a predetermined number of vehicles (for example, 10 vehicles) has been performed. If it is confirmed, the process proceeds to S56. On the other hand, when the lens information analysis unit 8 determines that the serial number of the predetermined number is not recognized (that is, when the predetermined number of license plates are not recognized), the process returns to S52 (S55). .

  The confirmation as to whether or not the recognition processing for a predetermined number of vehicles has been performed may be confirmed by, for example, acquiring the number of times of successful recognition of the license plate held by the plate recognition unit 5.

  In S <b> 56, the lens information analysis unit 8 causes the plate recognition unit 5 to calculate the character recognition rate, and acquires the character recognition rate calculated by the plate recognition unit 5 via the focus analysis unit 6. Here, the character recognition rate calculated by the plate recognition unit 5 will be described with reference to the case where the license plate in FIG. 2A is recognized.

  Even if the plate recognition unit 5 can detect the vehicle according to the state of the image received from the camera 1, the character recognition of the serial number 20 of the license plate may fail. The case where the character recognition of the serial number 20 of the license plate fails refers to the case where unrecognizable characters are included in the serial number 20 recognized from the detected vehicle image data. For example, in the example of FIG. 2, a case where data indicating a character that cannot be recognized is included in the recognition result, such as “? 34?” (“?” Indicates a character that cannot be recognized). On the other hand, the case where the character recognition of the serial number 20 is successful refers to the case where all the serial numbers 20 (“2345” in the example of FIG. 2) of the license plate can be recognized from the vehicle image detected by the image processing.

In accordance with a request from the lens information analysis unit 8, the plate recognition unit 5 calculates a character recognition rate using the number of detected vehicles held and the number of times of successful recognition of the serial number 20 of the license plate. . For example, the plate recognition unit 5 divides “the number of successful recognitions of the character of the serial number 20” by “the number of detected vehicles” as shown in (Equation 2) below, thereby recognizing characters for each focus position. The rate R (cn) is obtained. The plate recognition unit 5 outputs the obtained character recognition rate to the lens information analysis unit 8 via the focus analysis unit 6.
R (cn) = number of successful serial number recognition / number of detected vehicles (Equation 2)
Returning to FIG. When the lens information analysis unit 8 acquires the recognition rate from the plate recognition unit 5, the lens information analysis unit 8 stores the acquired character recognition rate in association with the focus position, and proceeds to the process of S57.

  In S57, the lens information analysis unit 8 obtains the edge sharpness Ec at the center of the screen according to the procedure described in FIG. 7, stores the obtained edge sharpened Ec in association with the focus position, and S58. Proceed to the process. If the character recognition rate acquired in S56 is equal to or less than a predetermined threshold (threshold used in S61 described later), the lens information analysis unit 8 may proceed to S58 without obtaining the edge sharpened Ec. Good.

  In S58, the lens information analysis unit 8 confirms whether or not the focus movement has been completed. If not, the focus position is moved via the lens control unit 9, and the process returns to S52 (S59 to 60). ). On the other hand, when all the movements of the focus are completed by the lens information analysis unit 8, the process proceeds to S61.

  In S61, the lens information analysis unit 8 uses the stored “data in which the character recognition rate is associated with each focus position” to store the focus position associated with the character recognition rate equal to or higher than a predetermined threshold. Select. Further, the lens information analysis unit 8 uses the stored “data in which the edge sharpness Ec is associated with each focus position” to select a focus position having the largest edge sharpness Ec from among the selected focus positions. Specify (S61).

  Thereafter, the lens information analysis unit 8 moves the focus position of the lens 2 so as to be the focus position specified via the lens control unit 9 (S62), and ends the process. When the certainty of character recognition is used instead of the edge sharpness Ec, the certainty of a predetermined character (place name or number) is stored in S54, and the average of the certainty is obtained in S57.

  As described above, in the present embodiment, the license plates of a predetermined number of vehicles are recognized for each focus position, and it is determined whether or not the focus is achieved using a plurality of license plates recognized for each focus position. An evaluation value (for example, edge sharpness Ec) is obtained. Therefore, in the present embodiment, an evaluation value indicating that the focus is not achieved is obtained by recognizing a license plate in a special state (for example, a vehicle with the license plate tilted and attached). Can reduce the effect. That is, in this embodiment, even if the evaluation value recognized from one vehicle is abnormal for some reason, the possibility that the focus is set to a completely different position can be reduced.

  Next, automatic adjustment for zooming of the lens 2 performed by the present embodiment will be described.

The zoom analysis unit 7 obtains the pitch of the serial number from the processing result of the plate recognition unit 5, and calculates the visual field size Wd based on the size. As shown in FIG. 2A, the size 25 of the first digit and the fourth digit from the left of the serial number of the medium plate is 195 mm. Moreover, as shown in FIG.2 (b), the size 26 of the 1st digit and the 4th digit from the left of a large sized plate is 272.5 mm. At the time of plate recognition, it is determined whether the recognized license plate is a large plate or a medium plate. Then, the zoom analysis unit 7 obtains the visual field size Wd using the following (formula 3) in the case of the medium size. Further, the zoom analysis unit 7 obtains the visual field size Wd using the following (formula 4) in the case of a large size.
Field size = number of horizontal pixels of screen × (195 / XSIZE) (Equation 3)
Field size = number of horizontal pixels of screen × (272.5 / XSIZE) (Formula 4)
Here, the number of horizontal pixels of the screen in (Equation 3) and (Equation 4) is the number of pixels of the entire screen, which is about 640 pixels in the case of an NTSC-compliant camera. XSIZE is the number of pixels corresponding to the center interval of the first and fourth digits from the left in the region coordinates obtained when the serial number is extracted. Large / medium size is determined by the height of the character. Specifically, the height of the characters of the medium number plate serial number 20 is “80 mm”, and the height of the characters of the serial number 20 of the large license plate is “120 mm”. Since the height of the character differs between “large” and “medium” by “1.5 times”, it can be easily determined by obtaining this height. If the serial number 20 is 3 digits, 2 digits, or 1 digit, the same calculation may be executed at intervals with the dot pattern. The zoom analysis unit 7 sends the visual field size Wd thus obtained and the plate position YP received from the plate recognition unit 5 to the lens information analysis unit 8.
The visual field size Wd varies depending on the capture position. Therefore, the lens information analysis unit 8 stores the visual field size Wd together with the plate position YP as shown in FIG. 11, and estimates the visual field size Wd at the center of the screen from the processing results of a plurality of units.
FIG. 12 is a diagram in which the relationship between the obtained visual field size Wd and the plate position YP is plotted for a plurality of units.

  The lens information analysis unit 8 uses the received data to obtain a relational expression (regression line y = ax + b) where “plate position YP” is “x” coordinates and “field size Wd” is “y” coordinates. . The field-of-view size at the center of the screen can be obtained by substituting the coordinates “YC” at the center of the screen for the “x” coordinates of the obtained regression line “y = ax + b”. It is assumed that the lens information analysis unit 8 holds in advance the optimum visual field size corresponding to the center of the screen. The lens information analysis unit 8 obtains the zoom movement amount from the difference “d_w” between the held optimum visual field size and the obtained visual field size at the center of the screen, and moves the zoom to the optimum position.

For example, if “d_w” is different by “50 mm”, it is possible to move the zoom position of the lens 2 to an optimum position if “d_w” is obtained by determining in advance how many zooms to move. In the field size calculation, the field size changes as the height from the road surface to the plate changes. In particular, since the mounting height of large plates varies greatly, the accuracy can be improved by not including the visual field size of the large plates in the statistics. In the above description, the zoom magnification is adjusted based on the field size. However, it is equivalent to replace with the resolution as in (Expression 5) and (Expression 6) below.
Resolution = (195 / XSIZE) (Formula 5)
Resolution = (272.5 / XSIZE) (Expression 6)
Next, the flow of zoom adjustment processing for the lens 2 will be described.

  FIG. 13 is a diagram for explaining a flow of zoom adjustment processing of the lens 2 performed by the present embodiment.

  First, a table used for statistical processing held by each unit of the image processing apparatus 3 is initialized (S70). Subsequently, the plate recognition unit 5 performs plate recognition processing (S71). Specifically, the plate recognition unit 5 receives the image data from the camera 1, detects the vehicle from the received image data, and recognizes the detected license plate of the vehicle. Further, the plate recognition unit 5 obtains the plate plate position YP of the license plate (S72). The plate recognition unit 5 outputs the recognized license plate image data and plate position YP to the zoom analysis unit 7.

  The zoom analysis unit 7 calculates the field size Wd according to the above-described procedure, and outputs the calculated field size Wd and the plate position YP received from the plate recognition unit 5 to the lens information analysis unit 8 (S73). When the lens information analysis unit 8 receives the visual field size Wd and the plate position YP output from the zoom analysis unit 7, the lens information analysis unit 8 determines whether or not a predetermined number of visual field sizes have been acquired (S74).

  The lens information analysis unit 8 proceeds to the process of S75 when the visual field size for a predetermined number of images can be acquired. On the other hand, when the lens information analysis unit 8 cannot acquire the visual field size for the predetermined number, the process returns to S71.

  In S75, the lens information analysis unit 8 estimates the visual field size “Wd” at the center of the screen from the relationship between the visual field size Wd and the plate position YP (S75). The lens information analysis unit 8 obtains the zoom movement amount “d_w” from the difference between the estimated visual field size “Wd” and the optimum visual field size held in advance, and determines the zoom position of the lens 2 via the lens control unit 9. The determined zoom movement amount “d_w” is moved (S76 to 77).

  The processing flow of FIG. 13 shows the zoom adjustment processing when the zoom magnification is deviated within a range where the license plate can be recognized. However, if the zoom is extremely deviated, the license plate cannot be recognized at all, and the control is not performed. Cannot be obtained (because the visual field size Wd cannot be obtained). Therefore, in the present embodiment, when the license plate cannot be recognized, similarly to the focus adjustment process described above, the process of evaluating the visual field size while changing the zoom magnification (lens zoom position) and searching for the optimal zoom position I do. Hereinafter, the flow of the zoom adjustment processing of the lens 2 performed when the license plate cannot be recognized will be described.

  FIG. 14 is a diagram for explaining the flow of the zoom adjustment processing of the lens 2 performed when the present embodiment cannot recognize the plate.

  First, as in S70 described above, a table used for statistical processing held by each unit of the image processing apparatus 3 is initialized (S80). Subsequently, the lens information analysis unit 8 sets the zoom position of the lens 2 to the initial value “cn = 0” via the lens control unit 9 (S81).

  In this state, the plate recognition unit 5 receives the image data from the camera 1 via the image input unit 4. The plate recognition unit 5 detects the vehicle from the received image data and recognizes the detected license plate of the vehicle (S82). The plate recognition unit 5 extracts the detected plate position YP of the license plate of the vehicle (S83). The plate recognition unit 5 outputs the recognized license plate image data and plate position YP to the zoom analysis unit 7.

  Subsequently, the zoom analysis unit 7 performs the same processing as in S73 described above, calculates the field size Wd, and outputs the calculated field size Wd and the plate position YP received from the plate recognition unit 5 to the lens information analysis unit 8. (S84).

  The lens information analysis unit 8 performs the same processing as the above S74 to 75, and estimates the visual field size Wd corresponding to the center of the screen using the visual field size Wd and the plate position Yp for a predetermined number (S85 to 86). The lens information analysis unit 8 stores the visual field size Wd estimated in association with the zoom position.

  Subsequently, the lens information analysis unit 8 causes the plate recognition unit 5 to calculate the character recognition rate R (cn) (S87), and acquires the character recognition rate R (cn) calculated from the plate recognition unit 5. When the zoom is not at all suitable, the character recognition rate R (cn) becomes 0 or the vehicle detection process does not operate normally. Therefore, when the acquired character recognition rate R (cn) is 0, the lens information analysis unit 8 ends the process. In S85, if the vehicle recognition process does not end even after a predetermined time has elapsed, the lens information analysis unit 8 sets the character recognition rate R (cn) to 0 and forcibly ends. When the zoom movement is completed in all cases, an optimal zoom position is specified (S91), and the zoom is moved. Specifically, the lens information analysis unit 8 determines the zoom position at which the visual field size Wd having a value closest to a predetermined visual field size is obtained from the estimated visual field sizes Wd for each stored zoom position. Is identified. Then, the lens information analysis unit 8 moves to the zoom position that specifies the zoom position of the lens 2 via the lens control unit 9.

By such processing, it is possible to return to the normal position even when the zoom is extremely moved. Note that the focus state basically does not change even when the zoom position is changed.
In addition, since the character size tolerance is provided in the license plate character recognition process, it is possible to recognize a character size falling within that range. For this reason, a slight zoom shift can be allowed, but if the zoom is largely shifted, for example, it may be recognized only when the image is taken at the lower side of the screen. In this case, the collected visual field size data as shown in FIG. 15 is concentrated at a specific location, and the visual field size at the center of the screen cannot be estimated. For this reason, the optimal visual field size is stored in advance as a straight line as well as the center of the screen, and when the obtained sample data is biased, the average visual field size “wd_av” of the sample data and the center coordinate “YP_av” of the plate position Yp ”Is obtained, and a shift amount“ d_w ”of the field size is obtained from the relationship between the center point and the optimum field straight line, and the zoom is moved. In this way, even when only a part of the recognition result is obtained, it is possible to set an optimum zoom magnification.

  By the way, while performing the focus adjustment and zoom adjustment described above, the lens position is moved to a position where the license plate cannot be recognized. Therefore, it is desirable to reduce the adjustment frequency of the lens 2.

  In this embodiment, the lens adjustment schedule management unit 10 holds data for determining the timing for performing lens adjustment. Then, the lens adjustment schedule management unit 10 specifies the lens adjustment time using the held data, and sends the lens to the focus analysis unit 6 (or zoom analysis unit 7) and the lens information analysis unit 8. Instruct the adjustment. The timing for adjusting the lens 2 is not particularly limited, but may be determined as shown in (1) to (5) below, for example.

  For example, the vehicle number recognition device (1) performs lens adjustment on the night of the installation date of the camera 1. In this case, a timer is provided in the lens adjustment schedule management unit 10, and the date and time for lens adjustment is set in the timer. The lens adjustment schedule management unit 10 instructs the focus analysis unit 6, zoom analysis unit 7, and lens information analysis unit 8 to adjust the lens when the set date / time is reached. The time set for the timer may be, for example, around 19:00 to 20:00. This is because if the set time is set to midnight, the number of vehicles passing through decreases, and data necessary for lens adjustment may not be acquired.

  For example, the car number recognition device adjusts the lens 2 at a rate of (2) once a week. Also in this case, as in (1) above, a timer is provided in the lens adjustment schedule management unit 10 to set the date and time of lens adjustment to be performed periodically (for example, 19:00 on Sunday). The lens adjustment schedule management unit 10 instructs the focus analysis unit 6, zoom analysis unit 7, and lens information analysis unit 8 to adjust the lens when the set date / time is reached.

  For example, the car number recognition device may perform lens adjustment when (3) the character recognition rate of the license plate is lower than the past character recognition rate history. In this case, for example, the lens adjustment schedule management unit 10 acquires the character recognition rate calculated by the plate recognition unit 5 (character recognition rate when a serial number of license plates for a predetermined number of units is recognized), and the acquired character The recognition rate is kept as a history. Then, when the character adjustment rate is acquired from the plate recognition unit 5, the lens adjustment schedule management unit 10 compares the history (for example, average value) of the past character recognition rate held with the acquired character recognition rate. . The lens adjustment schedule management unit 10 adjusts the lens adjustment to the focus analysis unit 6, zoom analysis unit 7, and lens information analysis unit 8 when the acquired character recognition rate is lower than a predetermined value compared to the past character recognition rate history. The instructions are given. If the character recognition rate is reduced, it cannot be determined whether the cause is the focus of the lens 2 or the zoom. Therefore, in this embodiment, when the character recognition rate is lowered, both focus and zoom are adjusted.

  Further, the vehicle number recognition device may (4) instruct the lens adjustment when the passing speed of the vehicle to be monitored largely fluctuates compared to the passing speed of the past vehicle. Specifically, a function for calculating the passing speed of the vehicle is provided in the plate recognition unit 5. And the lens adjustment schedule management part 10 acquires the passing speed of a vehicle from the plate recognition part 5, and hold | maintains the acquired passing speed as a log | history. The lens adjustment schedule management unit 10 calculates an average value when a predetermined number of passage speeds are acquired, and the calculated average value fluctuates by a predetermined value or more compared to an average of past passage speeds held as a history. If so, instruct lens adjustment. The passing speed of the vehicle can be obtained by a conventional technique (for example, JP-A-8-83393). Since the passing speed of the vehicle is calculated by moving pixels, the speed changes when the zoom is shifted.

  Further, the vehicle number recognition device may (5) always monitor the focus and zoom states, and adjust the lens when the predetermined range deviates from the optimum state (an example of always monitoring the focus and zoom states). Will be described in detail in FIG. 16).

  Note that data for determining the timing held by the lens adjustment schedule management unit 10 (for example, a date and time for performing lens adjustment and a predetermined value for determining whether or not the character recognition rate has decreased) is communicated from the external device 12. It can be changed via the part 11.

  Next, the monitoring process in the case of constantly monitoring the focus state and zoom state (5) described above will be described with reference to FIG.

  FIG. 16 is a diagram for explaining the flow of the lens adjustment process of the present embodiment. The lens adjustment schedule management unit 10 is provided with a zoom adjustment flag (the flag can be set to either “ON” or “OFF”) for determining whether or not to perform zoom adjustment. The lens adjustment schedule management unit 10 is provided with a focus adjustment flag (the flag can be set to either “ON” or “OFF”) for determining whether or not to perform focus adjustment.

  First, the lens adjustment schedule management unit 10 refers to the zoom adjustment flag. When the zoom adjustment flag is set to “ON”, the lens adjustment schedule management unit 10 proceeds to the process of S101. When the zoom adjustment flag is set to “OFF”, the lens adjustment schedule management unit 10 proceeds to S102. Proceed to processing.

  In step S <b> 101, the lens adjustment schedule management unit 10 instructs the zoom analysis unit 7 and the lens information analysis unit 8 to adjust the zoom of the lens 2. Then, the zoom analysis unit 7 and the lens information analysis unit 8 perform zoom adjustment according to the processing procedure of FIG. 13 or FIG. When the zoom adjustment is completed, the lens information analysis unit 8 transmits data indicating that to the lens adjustment schedule management unit 10. When the lens adjustment schedule management unit 10 receives data indicating that the zoom adjustment has been completed, the lens adjustment schedule management unit 10 sets the zoom adjustment flag to “OFF”, and the process returns to S100.

  Next, the processing of S1102 that proceeds when the zoom adjustment flag is “OFF” in S100 will be described.

  In S102, the lens adjustment schedule management unit 10 refers to the focus adjustment flag. If the focus adjustment flag is set to “ON”, the process proceeds to S103. If the focus adjustment flag is set to “OFF”, the process proceeds to S104. Proceed to processing.

  In S <b> 103, the lens adjustment schedule management unit 10 instructs the focus analysis unit 6 and the lens information analysis unit 8 to adjust the focus of the lens 2. Then, the focus analysis unit 6 and the lens information analysis unit 8 adjust the focus of the lens 2 in the above-described procedure of FIG. When the focus adjustment is completed, the lens information analysis unit 8 transmits data indicating that to the lens adjustment schedule management unit 10. When the lens adjustment schedule management unit 10 receives data indicating that the focus adjustment has been completed, the lens adjustment schedule management unit 10 sets the focus adjustment flag to “OFF” and returns to the process of S100.

  Subsequently, the processing from S104 onward when the focus adjustment flag is “OFF” in S102 will be described.

  In S104, the plate recognition unit 5 performs a normal license plate recognition process. Specifically, the plate recognition unit 5 performs the same processing as S52 and S53 in FIG.

  In S <b> 105, the zoom analysis unit 7 calculates a zoom evaluation value, and outputs the calculated zoom evaluation value to the lens information analysis unit 8. The zoom evaluation value is not particularly limited, and for example, the visual field size described in S84 of FIG. 14 can be used. In this case, similarly to S84, the zoom analysis unit 7 acquires the license plate image and the plate position YP from the plate recognition unit 5, and calculates the visual field size Wd from the acquired image. The zoom analysis unit 7 associates the calculated visual field size Wd with the acquired plate position YP and outputs it to the lens information analysis unit 8.

  Subsequently, in S <b> 106, the focus analysis unit 6 calculates a focus evaluation value, and outputs the calculated evaluation value to the lens information analysis unit 8. Although the focus evaluation value is not particularly limited, for example, the edge sharpness E of the serial number described in S54 of FIG. 10 can be used. In this case, the focus analysis unit 5 calculates the edge sharpness E in the same procedure as in S54, and outputs the calculated edge sharpness E and the plate position YP to the lens information analysis unit 8 in association with each other.

  When the lens information analysis unit 8 receives the zoom evaluation value (here, the visual field size Wd) and the focus evaluation value (here, the edge sharpness E) corresponding to the predetermined number, the lens information analysis unit 8 proceeds to the processing of S108 and proceeds to the predetermined number. When the zoom evaluation value and the focus evaluation value are not received, the process returns to S100.

  In S108, the abnormality monitoring unit 8 of the lens information analysis unit 8 determines whether or not zoom adjustment is necessary using the received zoom evaluation values Wd for a predetermined number of units. Specifically, the abnormality monitoring unit 8 holds data indicating the optimum visual field size in advance. The abnormality monitoring unit 8 obtains a difference “d_w” between the respective zoom evaluation values for the predetermined number and the held optimum visual field size, and if the difference “d_w” is equal to or larger than the predetermined value, the process proceeds to S109. If the difference “d_w” is smaller than the predetermined value, the process proceeds to S110. Note that the abnormality monitoring unit 8 may obtain the field-of-view size Wd at the center of the screen as shown in FIG. 12 using the evaluation values (field-of-view size) and the plate positions YP for a predetermined number. The abnormality monitoring unit 8 compares the obtained visual field size Wd with the optimum visual field size and determines whether or not to perform zoom adjustment.

  In S109, the abnormality monitoring unit 8 sets the lens adjustment flag held by the lens adjustment schedule management unit 10 to “ON”, and proceeds to the process of S111.

  Next, the process of S110 that is performed when it is determined in S108 that the zoom adjustment is unnecessary is described. In S110, the abnormality monitoring unit 8 of the lens information analysis unit 8 determines whether focus adjustment is necessary by using the received focus evaluation values (edge sharpness E) for a predetermined number of units.

  Specifically, the abnormality monitoring unit 8 has a “threshold value” for determining the focus evaluation value. Although the threshold value is not specifically limited, when the edge sharpness E is a value between “0 to 1”, for example, “0.5” is set. The abnormality monitoring unit 8 obtains the “edge sharpness Ec” at the center of the screen by using the received focus evaluation values (edge sharpness) and the plate position YP for the predetermined number. The abnormality monitoring unit 8 determines that it is necessary to perform focus adjustment when the “edge sharpness Ec” is lower than the “threshold value”. The abnormality monitoring unit 8 proceeds to S111 when it is determined that the focus adjustment is necessary, and returns to the process of S100 when it is determined that the focus adjustment is not performed.

  In S111, the abnormality monitoring unit 8 sets the focus adjustment flag held by the lens adjustment schedule management unit 10 to “ON” and returns to the process of S100.

  As described above, according to the present embodiment, while the characters on the license plate are recognized, the degree of blurring of the characters is evaluated and adjusted to the optimum focus position. Therefore, according to the present embodiment, the adjustment time at the time of installation is greatly reduced, and the operation can be performed in the most focused state, so that the recognition accuracy can be improved. In this embodiment, the visual field size is calculated from the recognized character plate spacing, and if the size changes compared to the initial value (optimal visual field size) held in advance, the zoom magnification is adjusted to the optimum zoom magnification. I am doing so. Therefore, in this embodiment, since it can be operated in a state where the zoom is most suitable, the recognition accuracy can be improved.

  The present invention is not limited to the embodiment described above, and various modifications can be made within the scope of the gist of the present invention. For example, in the above-described embodiment, the serial number of the license plate is used when obtaining the character recognition rate, but the present invention is not particularly limited to this. For example, all characters on the license plate may be used, or characters other than the serial number may be used.

It is a functional block diagram of the vehicle number recognition system of the embodiment of the present invention. It is the figure which illustrated the Japanese license plate. It is a figure explaining the photography state of the vehicles which the car number recognition system of this embodiment picturizes. 2 is a diagram illustrating an image taken by a camera 1. FIG. It is a figure for demonstrating the method of quantifying the blurring state of the character of a license plate. It is the figure which illustrated the state where the character of a naver plate is in focus. It is a figure for demonstrating the procedure which calculates | requires the edge sharpness of a screen center from the edge sharpness obtained by processing the image of a several vehicle. It is a figure for demonstrating the selection process of the focus position which the vehicle number recognition system of this embodiment performs. It is a figure for demonstrating the certainty of character recognition at the time of cutting out the image of the land transportation branch office code of a license plate. It is a figure for demonstrating the flow of the process which adjusts the focus position which the vehicle number recognition system of this embodiment performs. It is a figure for demonstrating a visual field size. It is the figure which plotted the relationship between the visual field size wd obtained by processing the image of several vehicles, and plate position YP. It is a figure for demonstrating the flow of the zoom adjustment process of the lens which the vehicle number recognition system of this embodiment performs. It is a figure for demonstrating the flow of the zoom adjustment process of a lens performed when the vehicle number recognition system of this embodiment cannot recognize a license plate. It is a figure for demonstrating the process which estimates the visual field size of the center of a screen performed when a visual field size collects only in one part. It is a figure for demonstrating the flow of the lens adjustment process which the vehicle number recognition system of this embodiment performs. It is the figure which illustrated the hardware constitutions of the image processing apparatus of embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Camera, 2 ... Lens, 3 ... Image processing apparatus, 4 ... Image input part, 5 ... Plate recognition part, 6 ... Focus analysis part, 7 ... Zoom analysis part, 8 ... Lens information analysis part, 9 ... Lens control part DESCRIPTION OF SYMBOLS 10 ... Lens adjustment schedule management part, 11 ... Communication part, 12 ... External apparatus, 13 ... Abnormality monitoring part, 100 ... Imaging part, 200 ... Image processing circuit 201 ... CPU, 202 ... Memory, 203 ... I / O circuit, 204: A / D conversion circuit, 205: Network interface

Claims (10)

In a vehicle number recognition device that receives image data from a television camera that captures a vehicle, processes the received image data, and recognizes a vehicle license plate,
The television camera is provided with an electric lens that receives a control signal from the outside and sets the focus position of the lens,
A lens control means for transmitting a control signal to the electric lens and setting a focus position of the lens of the electric lens;
Recognizing means for detecting a vehicle using the image data received from the television camera and recognizing a license plate of the detected vehicle;
Means for calculating a character recognition rate using the number of vehicles detected by the recognition means in a predetermined period and the number of vehicles that have recognized the license plate among the detected vehicles ;
Recognition history management means for storing the calculated character recognition rate as a history;
Means for calculating the edge sharpness at the center of the screen using images of a plurality of license plates recognized by the recognition means;
Focus abnormality determination means for determining whether or not the focus position is abnormal ;
If the focus position is determined to be abnormal, through the lens control unit, anda means for setting the optimum position of the focus position,
The focus abnormality determination means includes
When the calculated character recognition rate is compared with the average value of the character recognition rates stored in the recognition history management means, and the difference between the calculated character recognition rate and the average value is a predetermined value or more , or When the calculated edge sharpness is equal to or less than a predetermined value, the focus position is determined to be abnormal . In a vehicle number recognition device that receives image data from a television camera that captures a vehicle, processes the received image data, and recognizes a vehicle license plate,
The television camera is provided with an electric lens that receives a control signal from the outside and sets the zoom position of the lens,
A lens control means for transmitting a control signal to the electric lens and setting a zoom position of the lens of the electric lens;
Recognizing means for detecting a vehicle using the image data received from the television camera and recognizing a license plate of the detected vehicle;
Means for calculating the detected passing speed of the vehicle;
Speed history management means for storing the calculated passing speed as a speed history ;
Means for calculating the visual field size of the screen using the image of the license plate recognized by the recognition means;
Zoom abnormality determining means for determining whether or not the zoom position is abnormal ;
If the zoom position is determined to be abnormal, through the lens control unit, anda means for setting the optimum position of the zoom position,
The zoom abnormality determining means includes
The calculated passing speed is compared with the average value of the passing speeds stored in the speed history management means , and the calculated passing speed is larger than the average value by a predetermined value or the calculated view size When the difference between the predetermined optimum visual field size is a predetermined value or more, the zoom position is determined to be abnormal . In a vehicle number recognition device that receives image data from a television camera that captures a vehicle, processes the received image data, and recognizes a vehicle license plate,
The television camera is provided with an electric lens that receives a control signal from the outside and sets a focus position.
The license plate image is detected from the image data received from the television camera, and the license plate image is recognized using the detected license plate image, and the position of the license plate on the screen (plate position) Plate recognition means for
Evaluation means for obtaining an evaluation value quantifying the out-of-focus state of the detected license plate;
A control signal is transmitted to the motorized lens, the focus position of the motorized lens is sequentially moved to a plurality of predetermined focus positions, the plate recognition unit and the evaluation unit are controlled, and for each focus position moved. The plate positions and the evaluation values of a plurality of vehicles are obtained, a representative evaluation value for each focus position is calculated using the obtained plate positions and evaluation values, and the focus is calculated from the calculated representative evaluation values. A vehicle number recognition apparatus comprising: means for specifying a focus position at which a representative evaluation value of the most suitable value is obtained, and setting the focus position of the electric lens to the specified focus position. In the vehicle number recognition device according to claim 3 ,
The evaluation means includes
Using the license plate image detected by the plate recognition means, the serial number of the license plate is cut out, the blurring state of the character outline is quantified for each character area of the cut out serial number, and the quantified value of A vehicle number recognition device characterized in that an average value is an evaluation value of the license plate. In the vehicle number recognition device according to claim 3 ,
The plate recognizing means calculates a certainty factor that is a value representing recognition accuracy when recognizing the characters of the license plate,
The evaluation means includes
An evaluation value of the license plate is calculated using at least one confidence factor of a serial number of the license plate calculated by the plate recognition means, a land transportation branch office code, a vehicle type code, and a use code. Recognition device. The vehicle number recognition device according to claim 3 ,
The plate recognizing means calculates a certainty factor that is a value representing recognition accuracy when recognizing the characters of the license plate,
The evaluation unit is
The evaluation value of the license plate is calculated using a multiplication value of at least two certainty factors among the serial number of the license plate calculated by the plate recognition means, the land transportation branch office code, the vehicle type code, and the use code. Car number recognition device. The vehicle number recognition device according to claim 3 ,
The plate recognizing means calculates a certainty factor that is a value representing recognition accuracy when recognizing the characters of the license plate,
The evaluation unit is
A vehicle number recognition device characterized in that the evaluation value of the license plate is calculated using the certainty of the recognition result of a predetermined land transportation branch code from the certainty output from the plate recognition means. The vehicle number recognition device according to claim 3 ,
The evaluation means includes
An evaluation value is calculated using data of a medium-sized plate among license plates recognized by the plate recognition means. In a vehicle number recognition device that receives image data from a television camera that captures a vehicle, processes the received image data, and recognizes a vehicle license plate,
The television camera is provided with an electric lens for receiving a control signal from the outside and setting a zoom position,
The license plate image is detected from the image data received from the television camera, and the license plate image is recognized using the detected license plate image, and the position of the license plate on the screen (plate position) Plate recognition means for
Visual field size calculation means for calculating the visual field size of the screen from the character spacing of the license plate, using the detected license plate image;
A control signal is transmitted to the electric lens, the zoom position of the electric lens is sequentially moved to a plurality of predetermined zoom positions, and the plate recognition unit and the field size calculation unit are controlled for each of the moved zoom positions. Then, the plate position and the field size of a plurality of vehicles are obtained, a representative field size for each zoom position is calculated using the obtained plate position and field size, and the representative field size is calculated in advance from the calculated representative field sizes. A zoom position where a representative visual field size having a value closest to a predetermined visual field size is determined, and a zoom position of the electric lens is set to the specified zoom position. Car number recognition device. The vehicle number recognition device according to claim 9 ,
The visual field size calculation means includes
A vehicle number recognizing device, wherein a visual field size is calculated using data of a medium-sized license plate from among license plates recognized by the plate recognizing means.

Images