JP7720771B2 - Image recognition device, image recognition system, and image recognition method - Google Patents

Description

This disclosure relates to image recognition technology.

In order to find abandoned objects, a known image recognition method involves preparing a reference image in advance that does not show any abandoned objects, comparing the reference image with the current image captured by a surveillance camera to detect differences, and then detecting abandoned objects from these differences. In outdoor facilities, this method of comparing with a reference image can result in false positives due to sunlight reflection or shadows if the weather when the reference image was prepared differs from the weather when the current image was captured.

Patent Document 1 discloses a method for setting an image quality measurement area on an image input from an imaging device and issuing a warning if the contrast in the set image quality measurement area decreases.

Japanese Patent Application Laid-Open No. 2001-160146

However, false detections can occur even when image quality, such as contrast, is good due to sunlight reflection or shadows, so the image recognition method described in Patent Document 1 has the problem of being unable to deal with changes in images due to weather.

This disclosure has been made in light of this background, and aims to provide image recognition technology that can reduce false detections that can occur due to changes in weather.

An image recognition device according to an embodiment of the present disclosure includes an image recognition unit that detects objects by comparing an image captured by a surveillance camera with a reference image stored in advance, and a difference non-detection mask setting unit that sets a difference non-detection mask, which is an area in which differences between images are not detected, depending on the weather in the location where the surveillance camera is installed, and the image recognition unit performs the detection using the set difference non-detection mask.

An image recognition device according to an embodiment of the present disclosure can reduce false detections that can occur due to changes in weather.

1 is a diagram illustrating an example of the configuration of an image recognition device and an image recognition system according to a first embodiment; FIG. 2 is a diagram illustrating a detailed configuration example of the image recognition device according to the first embodiment. FIG. 10 is a diagram for explaining differences in an image caused by weather conditions. FIG. 10 is a diagram for explaining differences in an image caused by weather conditions. FIG. 10 is a diagram for explaining differences in an image caused by weather conditions. FIG. 10 is a diagram illustrating an example of application of a difference non-detection mask. FIG. 10 is a diagram illustrating an example of application of a difference non-detection mask. FIG. 10 is a diagram illustrating an example of application of a difference non-detection mask. FIG. 10 is a diagram illustrating an example of application of a difference non-detection mask. FIG. 2 is a diagram illustrating an example of the hardware configuration of an image recognition device. FIG. 2 is a diagram illustrating an example of the hardware configuration of an image recognition device. 4 is a flowchart showing the operation of the image recognition device according to the first embodiment. FIG. 10 is a diagram illustrating a configuration example of an image recognition device according to a second embodiment. FIG. 1 is a diagram showing shadows caused by weather conditions. FIG. 10 is a diagram illustrating an example of application of a difference non-detection mask. FIG. 10 is a diagram illustrating an example of application of a difference non-detection mask. FIG. 10 is a diagram illustrating an example of application of a difference non-detection mask. 10 is a flowchart showing the operation of the image recognition device according to the second embodiment. FIG. 10 is a diagram illustrating an example of the configuration of an image recognition device according to a third embodiment. 10 is a diagram for explaining the operation of a weather information determination unit. FIG. 11 is a flowchart showing the operation of the image recognition device according to the third embodiment. FIG. 10 is a diagram illustrating an example of the configuration of an image recognition device and an image recognition system according to a fourth embodiment. FIG. 10 is a diagram illustrating a display example of a display screen. 10 is a flowchart showing the operation of the image recognition device according to the fourth embodiment. FIG. 13 is a diagram illustrating an example of the configuration of an image recognition device according to a fifth embodiment. 13 is a flowchart showing the operation of the image recognition device according to the fifth embodiment.

Various embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. Note that components with the same or similar reference numerals in the drawings have the same or similar configurations or functions, and redundant descriptions of such components will be omitted.

Embodiment 1.
<Image recognition system configuration>
An image recognition device and an image recognition system according to a first embodiment of the present disclosure will be described with reference to FIGS. 1 to 5 . FIG. 1 is a diagram illustrating an example configuration of the image recognition device and the image recognition system according to the first embodiment. As illustrated in FIG. 1 , an image recognition system 6 a includes, as an example, an image recognition device 100, surveillance cameras 3 a and 3 b, a monitoring terminal 4, and an on-premise network 5. The image recognition system 6 a is a system that uses images from surveillance cameras 3 (3 a, 3 b) installed in a facility to be monitored to detect various events within the facility and notifies facility staff of the detected events via the monitoring terminal 4. Examples of the facility include transportation stations such as train and bus stations, parking lots, amusement parks, and public parks. Images captured by the surveillance cameras 3 are transmitted to the image recognition device 100 via the on-premise network 5 within the facility. The image recognition device 100 analyzes the transmitted images to detect various conditions occurring within the facility. Examples of conditions detected by the image recognition system include the detection of abandoned objects within the facility, people lingering for long periods of time, the bringing in of dangerous objects, and people falling onto or entering the tracks at railway station facilities operated by railway operators. The conditions detected through image analysis by the image recognition device 100 are displayed on the display device of the monitoring terminal 4, and the detected conditions are notified to facility staff.

The image recognition system 6a is connected to the system of a weather information provider 9 via a network 7 and the Internet 8, and obtains weather information 10 indicating the weather from the weather information provider 9. Multiple image recognition systems 6a may be connected to the network 7.

<Configuration of image recognition device>
2 is a diagram showing a detailed configuration example of the image recognition device 100. As shown in FIG. 2, the image recognition device 100 includes an image recognition unit 101 and a weather information analysis unit 102.

(Image recognition section)
The image recognition unit 101 analyzes images captured by the surveillance camera 3 to detect various objects, such as people, in the images. For example, it detects abandoned objects contained in the images. There are various image analysis methods, one of which is to compare a reference image with a current image and detect the difference between them. For example, by comparing the reference image with the current image to detect a difference, if the difference continues to be observed, it is possible to determine that the difference represents an abandoned object and detect it as such. This reference image must not include facility users, such as passengers in the case of a station, so an image taken early in the morning, outside the facility's operating hours, is used. This image taken early in the morning, for example, is stored in a memory (not shown) as the reference image, and the image recognition unit 101 detects objects by comparing the reference image with the current image captured by the surveillance camera 3.

When comparing a reference image with a current image, the cause of false positives when recognizing images of outdoor facilities is the unique image that is displayed due to factors such as weather. This will be explained with reference to Figure 3. As shown in Figure 3, in contrast to the reference image, which is often taken early in the morning (reference image in Figure 3A), there are differences in the shadow of the facility or the reflection of sunlight on sunny days (sunny day image in Figure 3B), and in rainy days there are differences in the reflection of lighting caused by the wetness of the platform floor (rainy day image in Figure 3C). The shadows or reflections in these images are not abandoned objects that should be determined to be objects to be detected, but because they continue to be seen, they are falsely detected as abandoned objects.

One possible way to prevent such false positives is to set a difference non-detection mask. A difference non-detection mask is an area (range) surrounded by an arbitrary shape set within the image analyzed by the image recognition device 100, and even if an image different from the reference image is detected within the set area, the different image is not detected as a difference. As in the example of applying a difference non-detection mask shown in Figure 3D, false positives due to weather can be avoided by applying a difference non-detection mask to all areas where unique images caused by weather, such as the aforementioned shadows or reflections, may be visible. However, when using this method, there is the issue that the range in which differences can be detected is narrowed by setting a large or wide difference non-detection range, which reduces the accuracy of condition detection by the image recognition device.

(Weather Information Analysis Department)
To solve this problem, the image recognition device 100 of the present disclosure adaptively sets a difference non-detection mask according to the climate. To adaptively set a difference non-detection mask according to the climate, the weather information analysis unit 102 includes a weather information acquisition unit 102 a, a weather determination unit 102 b, a difference non-detection mask setting unit 102 c, and a mask information output unit 102 d, as shown in FIG.

(Weather Information Acquisition Department)
The weather information acquisition unit 102a acquires weather information 10 at regular intervals. As an example, the weather information 10 is information provided by an external weather information provider 9 that provides weather information, and is information indicating weather conditions such as sunny, cloudy, rainy, or snowfall that match the area where the target facility, where the surveillance camera 3 is installed, is located. The weather information acquisition unit 102a acquires the weather information 10 from the weather information provider 9 at regular intervals via the connected network 7. Note that the regular interval shown here is assumed to be about one hour, which allows for a period of time to respond to changes in the weather, but can be set arbitrarily by the weather information analysis unit 102.

(Weather Judgment Department)
The weather determination unit 102b determines the weather at the target facility from the weather information 10 acquired by the weather information acquisition unit 102a. As described above, the weather information 10 is acquired at regular intervals, so the weather determination unit 102b can determine the current climate for a certain time interval.

(Difference non-detection mask setting section)
The difference non-detection mask setting unit 102c dynamically sets the difference non-detection mask. That is, the difference non-detection mask setting unit 102c sets the difference non-detection mask according to the current weather determined by the weather determination unit 102b. The difference non-detection mask setting unit 102c sets multiple difference non-detection masks within one detection target image and independently specifies whether each mask is valid or invalid. This will be described with reference to FIG. 4.

As shown in FIG. 4A, the difference non-detection mask setting unit 102c sets multiple difference non-detection masks M1 to M4 at the respective positions of unique differences that occur due to weather. On sunny days, as shown in FIG. 4B, the difference non-detection mask setting unit 102c enables difference non-detection masks M1, M3, and M4 to mask sunlight reflections and equipment shadows. On rainy days, as shown in FIG. 4C, the difference non-detection mask setting unit 102c enables difference non-detection mask M2 to mask lighting reflections caused by wetness on the platform floor. In this way, the difference non-detection mask setting unit 102c sets multiple difference non-detection masks within a single detection target image and independently specifies whether each mask is enabled or disabled.

(Mask information output unit)
The mask information output unit 102d outputs the setting information of the difference non-detection mask set by the difference non-detection mask setting unit 102c to the image recognition unit 101 as mask information.

The image recognition unit 101 performs image analysis using the enabled difference non-detection mask indicated by the mask information output from the mask information output unit 102d. In other words, the image recognition unit 101 refers to the enabled difference non-detection mask and performs image analysis without detecting, as differences, fluctuations in areas masked by the difference non-detection mask in the image captured by the surveillance camera 3.

Next, the hardware configuration of the image recognition device 100 will be described with reference to Figures 5A and 5B. As an example, as shown in Figure 5A, the image recognition device 100 is realized by a processing circuit 100a. The functions of the functional units of the image recognition device 100 are realized by the processing circuit 100a. The processing circuit 100a is, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination of these. The functions of each functional unit of the image recognition device 100 may be realized by separate processing circuits, or these functions may be realized together by a single processing circuit.

As another example, as shown in FIG. 5B, the image recognition device 100 is implemented by a processor 100b and a memory 100c. The functions of the functional units of the image recognition device 100 are implemented by the processor 100b reading and executing a program stored in the memory 100c. The program is implemented as software, firmware, or a combination of software and firmware. Examples of the memory 100c include non-volatile or volatile semiconductor memory such as RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory), and EEPROM (Electrically-EPROM), as well as magnetic disks, flexible disks, optical disks, compact disks, minidisks, and DVDs.

<Operation>
Next, the operation of the image recognition device 100 will be described with reference to Fig. 6. In step S51, the weather information acquisition unit 102a acquires weather information 10 from an external weather information provider.

In step S52, the weather determination unit 102b determines the current weather indicated by the weather information 10.

In step S53, the difference non-detection mask setting unit 102c sets the difference non-detection mask assigned to the image specific to each weather to be valid, depending on the determined weather.

In step S54, the mask information output unit 102d outputs information indicating the enabled difference non-detection mask to the image recognition unit 101 as mask information. The image recognition unit 101 performs image recognition using the enabled difference non-detection mask. When comparing the current image with the reference image, the range of the enabled difference non-detection mask is not detected as a difference even if there is a change, so erroneous detection due to weather-specific images can be prevented.

Then, in step S55, after a certain period of time has elapsed, processing returns to step S51.

This operation performs a judgment at regular intervals, allowing you to set a difference non-detection mask that is appropriate for that time. This prevents you from setting an unnecessary difference non-detection mask and unnecessarily narrowing the detection range of image recognition.

In addition, this operation can be used for all content that can be detected using the difference from a reference image, and in addition to detecting abandoned objects, it can also be used to detect people staying behind, and in station facilities, people falling onto or entering the tracks.

Embodiment 2.
<Configuration>
Next, an image recognition device 100A according to a second embodiment of the present disclosure will be described with reference to Figures 7 and 8. Figure 7 is a diagram illustrating an example configuration of the image recognition device 100A. The image recognition device 100A includes a weather information analysis unit 102A, which additionally includes a shadow position calculation unit 102e that calculates the position of a shadow of equipment, in addition to the components of the weather information analysis unit 102 of the first embodiment.

(Shadow position calculation section)
The shadow position calculation unit 102e stores location information indicating the latitude and longitude of the facility to be detected in a memory (not shown), and calculates the position of the shadow for each hour in the image of the image recognition target based on the position of the sun determined by the date and time and the location information of the facility. As shown in Figure 8A, the position of the shadow is different between 10:00 on XX/XX/XX and 16:00 on XX/XX/XX. The shadow position calculation unit 102e calculates the position of the shadow of the target facility according to the relevant date and time.

In the second embodiment, the difference non-detection mask setting unit 102c prepares multiple difference non-detection masks with different setting positions. For example, as shown in the difference non-detection mask setting example of FIG. 8B, multiple difference non-detection masks such as difference non-detection masks M11 and M12 with different setting positions are prepared. The difference non-detection mask setting unit 102c activates the difference non-detection mask that masks the area closest to the calculated shadow position. For example, if it is 10:00 on XX/XX/XX/Xth, difference non-detection mask M11 is activated as shown in FIG. 8C. Also, if it is 16:00 on XX/XX/XX/Xth, difference non-detection mask M12 is activated as shown in FIG. 8D. It is determined that the masked area of difference non-detection mask M11 is closer to the shadow at 11:00 on XX/XX/XX/Xth than the masked area of difference non-detection mask M12, and difference non-detection mask M11 is activated. The mask area of difference non-detection mask M12 is determined to be closer to the shadow of 15:00 on XX/XX/XX date than the mask area of difference non-detection mask M11, and difference non-detection mask M12 is set to be valid.

This allows you to set the most effective difference non-detection mask for equipment shadows, which change depending on the season and time of day. For example, if it's a sunny day but there are no equipment shadows on the screen, you can avoid setting an unnecessary difference non-detection mask to be active, preventing the image recognition detection range from being narrowed more than necessary.

<Operation>
Next, the operation of the image recognition device 100A will be described with reference to Fig. 9. Fig. 9 is a flowchart showing the operation of the image recognition device 100A. The flowchart in Fig. 9 is a flowchart obtained by adding steps S83 to S85 for dynamically changing the mask position on clear skies to the flowchart in Fig. 6 according to the first embodiment.

As in the case of Figure 6, in step S81, the weather information acquisition unit 102a obtains weather information, and in step S82, the weather determination unit 102b determines the weather.

If the current weather is determined to be sunny, in step S84, the shadow position calculation unit 102e calculates the position of the shadow of the target equipment in the facility on the relevant date and time.

In step S85, the difference non-detection mask setting unit 102c, based on the shadow position calculated in step S84, sets the difference non-detection mask that masks the location closest to the calculated shadow position to be active, from among the multiple difference non-detection masks prepared and stored (for example, the setting example for XX year, X month, X day, 10:00 in Figure 8C, and the setting example for XX year, X month, X day, 16:00 in Figure 8D).

In step S87, the mask information output unit 102d outputs information indicating the enabled difference non-detection mask to the image recognition unit 101 as mask information. The image recognition unit 101 performs image recognition using the enabled difference non-detection mask. This makes it possible to set the most effective difference non-detection mask for facility shadows, which change depending on the season and time of day. Additionally, if the weather is clear but the shadow of the facility does not exist on the screen due to the position of the sun, an unnecessary difference non-detection mask will not be set, thereby preventing a reduction in the detection range of image recognition.

If a result other than sunny weather is determined in step S83, in step S86, the difference non-detection mask setting unit 102c sets the difference non-detection mask to be active according to weather other than sunny weather, as in embodiment 1.

After that, as in embodiment 1, after a certain period of time has elapsed, processing returns to step S81 (step S88).

By including the shadow position calculation unit 101e as in the image recognition device 100A of embodiment 2, it is possible to set a difference non-detection mask more accurately, thereby enabling more accurate object detection.

Embodiment 3.
<Configuration>
Next, an image recognition device 100B according to a third embodiment of the present disclosure will be described with reference to Figures 10 and 11. Figure 10 is a diagram illustrating an example configuration of the image recognition device 100B according to the third embodiment. The image recognition device 100B includes a weather information analysis unit 102B, which includes a weather information determination unit 102f that determines the weather conditions of the facility where the surveillance cameras 3 (3a, 3b) are installed, instead of the weather information acquisition unit 102a of the first embodiment.

(Weather Information Judgment Department)
The weather information determination unit 102f acquires outdoor images of the facility where the monitoring cameras 3 (3a, 3b) are installed from the monitoring cameras 3 (3a, 3b). The number of monitoring cameras capturing outdoor images may be one or more. As shown in FIG. 11 , the weather information determination unit 102f compares the current image acquired from the monitoring camera 3 with multiple weather images pre-stored in the weather information analysis unit 102B or the image recognition device 100B to determine the current weather. FIG. 11 illustrates an example in which the current image is compared with stored images including images of sunny weather, cloudy weather, rainy weather, and stormy weather, and the current weather is determined to be "cloudy." Weather can be determined based on, for example, the presence or absence of shadows on a sunny day, the presence or absence of sunlight reflection in a specific location, or color changes due to wetness of the floor or ground on a rainy day, but other factors may also be taken into consideration.

Information indicating the weather determined by the weather information determination unit 102f is output as weather information to the weather determination unit 102b. Based on this weather information, the weather determination unit 102b, difference non-detection mask setting unit 102c, and mask information output unit 102d function in the same way as in embodiment 1.

<Operation>
Next, the operation of the image recognition device 100B will be described with reference to Fig. 12. Fig. 12 is a flowchart showing the operation of the image recognition device 100B. Fig. 12 differs from Fig. 6 in that step S111 is provided instead of step S51 in Fig. 6 according to the first embodiment. In Fig. 12, instead of acquiring weather information as in step S51 in Fig. 6, the weather information determination unit 102f determines the weather from the image and uses it as weather information.

Steps S112 to S115 in Figure 12 are the same as steps S52 to S55 in Figure 6, and the difference non-detection mask is enabled or disabled depending on the weather, and the enabled difference non-detection mask is set in the image recognition unit 101.

In addition, in embodiment 3, if the weather is determined to be clear, it is possible to add the shadow position calculation unit 102e shown in embodiment 2 and dynamically change the position of the shadow of the facility.

It is also possible to combine embodiment 3 and embodiment 1 to comprehensively determine weather information from weather information obtained from an external weather information provider 9 and weather information determined from video footage of the outside of the facility.

Embodiment 4.
<Configuration>
Next, an image recognition system 6b according to a fourth embodiment of the present disclosure will be described with reference to Figures 13 and 14. Figure 13 is a diagram illustrating an example configuration of the image recognition system 6b according to the fourth embodiment. As shown in Figure 13, the image recognition system 6b includes a monitoring terminal 4A, which includes a weather information input unit 13 that accepts user input of weather information indicating the weather. In the image recognition system 6a according to the first embodiment, the weather information 10 is obtained from an external weather information provider 9, but the image recognition system 6b according to the fourth embodiment obtains weather information from the weather information input unit 13.

Figure 14 shows an example of a display screen displayed on the display device of the monitoring terminal 4A. The display screen includes, for example, an outdoor image display screen that displays outdoor images to the facility employee (user), a weather information output screen that displays current weather information determined by the weather information input system, and a weather information input screen on which the facility employee can input weather information. Corresponding to this weather information input screen, the monitoring terminal 4A is equipped with a weather information input unit 13. The facility employee determines the current weather by checking the outdoor image and inputs weather information via the weather information input screen. If the employee determines that the weather displayed on the weather information output screen is the same as the current weather, they can choose not to input weather information. This input can be made at any time. In embodiment 4, the difference non-detection mask setting unit 102c sets the difference non-detection mask according to the weather indicated by the weather information input by the user. The hardware configuration of the monitoring terminal 4 (4A) can be the same as that shown in Figure 5A or Figure 5B.

<Operation>
Next, the operation of the image recognition system 6b will be described with reference to FIG. 15. FIG. 15 is a flowchart showing the operation of the image recognition system 6b. Steps S141 to S144 in FIG. 15 are substantially the same as steps S51 to S54 in FIG. 6 according to the first embodiment. The only difference is where the weather information is acquired. That is, in step S141 in FIG. 15, the weather information acquisition unit 102a acquires weather information from the weather information input unit 13.

In Figure 15, weather information is judged each time it is input via the weather information input unit 13, without waiting for a certain period of time to elapse. In other words, the flowchart in Figure 15 does not include a step corresponding to step 55 in Figure 6.

In addition, in embodiment 4, if the weather is determined to be clear, it is possible to add the shadow position calculation unit 102e shown in embodiment 2 and dynamically change the position of the facility's shadow.

It is also possible to combine embodiment 4 with embodiments 1 and 2. In this case, priority is given to input by facility staff via the weather information input unit 13, and if there is no input for a certain period of time, the current weather is determined to be external weather information, weather information determined from video footage of the outside of the facility, or weather information determined comprehensively from both of these pieces of information.

Embodiment 5.
<Configuration>
Next, an image recognition device 100C according to a fifth embodiment of the present disclosure will be described with reference to Fig. 16. Fig. 16 is a diagram illustrating an example configuration of the image recognition device 100C. As shown in Fig. 16, the image recognition device 100C includes a weather information analysis unit 102C, which additionally includes a stop instruction unit 102g that instructs the weather information analysis unit 102 of the first embodiment to stop the image recognition process (difference detection).

(Stop instruction section)
The stop instruction unit 102g instructs the image recognition unit 101 to stop difference detection when the weather determined by the weather determination unit 102b is a rainstorm.
<Operation>
FIG. 17 is a flowchart showing the operation of the fifth embodiment. FIG. 17 is a diagram in which processing for when the weather is a rainstorm is added to FIG. 6 for the first embodiment. In the case of a rainstorm caused by a typhoon or the like, rain may fall on a surveillance camera capturing outdoor footage, making it impossible to compare the image with the reference image. To deal with such a case, when the weather is a rainstorm, difference detection itself is stopped rather than setting a difference non-detection mask. This makes it possible to prevent false detections that occur when comparison with the reference image is impossible.

The operation will be described with reference to Figure 17. If the weather determined in step S152 is a rainstorm, rather than enabling or disabling the difference non-detection mask, the image recognition unit 101 is caused to stop difference detection itself in step S154. The other steps are the same as in Figure 6.

Storm information may be obtained from typhoon information or heavy rain warnings provided by an external weather information provider as in embodiment 1, from judgments based on external video as in embodiment 2, or from input from personnel as in embodiment 3.

<Additional Notes>
Some aspects of the various embodiments described above are summarized below.

(Appendix 1)
The image recognition device (100; 100A; 100B; 100C) of Appendix 1 comprises an image recognition unit (101) that detects objects by comparing an image captured by a surveillance camera with a reference image stored in advance, and a difference non-detection mask setting unit (102c) that sets a difference non-detection mask, which is an area where differences between images are not detected, according to the current weather at a location where the surveillance camera is installed, and the image recognition unit performs the detection using the set difference non-detection mask.

(Appendix 2)
The image recognition device (100; 100A; 100C) of Supplementary Note 2 is the image recognition device of Supplementary Note 1, further comprising a weather information acquisition unit (102a) that acquires weather information indicating the weather, the weather information being information provided by a weather information provider via a network, and the difference non-detection mask setting unit sets the difference non-detection mask according to the weather indicated by the acquired weather information.

(Appendix 3)
The image recognition device (100B) of Appendix 3 is the image recognition device of Appendix 1, further comprising a weather information determination unit (102f) that determines weather information indicating the weather from an image captured by a surveillance camera, and the difference non-detection mask setting unit sets the difference non-detection mask according to the weather indicated by the weather information determined from the image.

(Appendix 4)
The image recognition device (100; 100A; 100B; 100C) of Supplementary Note 4 is the image recognition device of any one of Supplementary Notes 1 to 3, and further includes a shadow position calculation unit (102e) that calculates the position of a shadow of the facility from position information of latitude and longitude where the facility is located and the position of the sun based on date and time, and the difference non-detection mask setting unit changes the position of the difference non-detection mask in accordance with the calculated position of the shadow when the weather is fine.

(Appendix 5)
The image recognition device (100; 100A; 100B; 100C) of Supplementary Note 5 is any one of the image recognition devices of Supplementary Note 1 to 4, and further includes a stop instruction unit (102g) that instructs the image recognition unit to stop detecting differences when the weather is a heavy rainstorm.

(Appendix 6)
The image recognition system (6a; 6b) of Supplementary Note 6 comprises a monitoring terminal (4A) having a weather information input unit (13) that accepts user input of weather information indicating the weather, and any one of the image recognition devices of Supplementary Note 1 to 5, and the difference non-detection mask setting unit sets the difference non-detection mask according to the weather indicated by the weather information input by the user.

(Appendix 7)
The image recognition method of Appendix 7 is an image recognition method using an image recognition device that has a difference non-detection mask setting unit and an image recognition unit, and includes a step (step S53) in which the difference non-detection mask setting unit sets a difference non-detection mask, which is an area in which differences between images are not detected, in accordance with the current weather, and a step (step S54) in which the image recognition unit uses the set difference non-detection mask to compare an image captured by a surveillance camera with a reference image stored in advance to detect an object, and the difference non-detection mask is a difference non-detection mask that is set in accordance with the current weather at a location where the surveillance camera is installed.

It is possible to combine embodiments, and to modify or omit each embodiment as appropriate.

The image recognition device or image recognition system disclosed herein can be used in systems that recognize images taken by surveillance cameras of outdoor facilities such as train or bus transportation stations, parking lots, amusement parks, and public parks.

3 (3a, 3b) Surveillance camera, 4 (4A) Surveillance terminal, 5 In-house network, 6 (6a, 6b) Image recognition system, 7 Network, 9 Weather information provider, 10 Weather information, 13 Weather information input unit, 100 (100A, 100B, 100C) Image recognition device, 100a Processing circuit, 100b Processor, 100c Memory, 101 Image recognition unit, 102 (102A, 102B, 102C) Weather information analysis unit, 102a Weather information acquisition unit, 102b Weather determination unit, 102c Difference non-detection mask setting unit, 102d Mask information output unit, 102e Shadow position calculation unit, 102f Weather information determination unit, 102g Stop instruction unit, M (M1, M11, M12, M2, M3, M4) Difference non-detection mask.

Claims (6)

an image recognition unit that detects an object by comparing an image captured by a surveillance camera with a reference image stored in advance;
a difference non-detection mask setting unit that sets a difference non-detection mask, which is an area in which differences between images are not detected, in accordance with the current weather at a location where the surveillance camera is installed;
Equipped with
the image recognition unit performs the detection using a set difference non-detection mask;
An image recognition device ,
a shadow position calculation unit that calculates the position of a shadow of the facility based on the latitude and longitude position information of the facility and the position of the sun according to date and time;
the difference non-detection mask setting unit changes the position of the difference non-detection mask in accordance with the calculated position of the shadow when the weather is fine.
Image recognition device . A weather information acquisition unit that acquires weather information indicating weather conditions is further provided,
The weather information is provided by a weather information provider via a network,
the difference non-detection mask setting unit sets the difference non-detection mask according to the weather indicated by the acquired weather information;
2. The image recognition device according to claim 1. The device further includes a weather information determination unit that determines weather information from an image captured by the monitoring camera,
the difference non-detection mask setting unit sets the difference non-detection mask in accordance with the weather indicated by the weather information determined from the image;
2. The image recognition device according to claim 1. a stop instruction unit that instructs the image recognition unit to stop detecting the difference when the weather is a heavy rainstorm;
4. The image recognition device according to claim 1. a monitoring terminal including a weather information input unit that accepts user input of weather information indicating the weather;
An image recognition device according to any one of claims 1 to 4 ;
Equipped with
the difference non-detection mask setting unit sets the difference non-detection mask in accordance with the weather indicated by the weather information input by the user;
Image recognition system. An image recognition method using an image recognition device including a shadow position calculation unit, a difference non-detection mask setting unit, and an image recognition unit,
a step in which the shadow position calculation unit calculates the position of the shadow of the facility from the position information of latitude and longitude where the facility is located and the position of the sun based on date and time;
a step in which the difference non-detection mask setting unit sets a difference non-detection mask, which is an area in which a difference between images is not detected, in accordance with current weather, and when the weather is fine, changes the position of the difference non-detection mask in accordance with the calculated position of the shadow ;
the image recognition unit uses the set difference non-detection mask to compare the image captured by the surveillance camera with a reference image stored in advance to detect an object;
An image recognition method comprising:
The difference non-detection mask is a difference non-detection mask set according to the current weather at a location where the surveillance camera is installed.
Image recognition methods.