JP6929030B2 - Vehicle recognition device - Google Patents

Description

The present invention relates to a technique for recognizing the existence of a vehicle using a captured image.

There is a need to recognize the existence of a vehicle by using an image (captured image) captured by an imaging device such as a fixed-point camera. For example, if the presence or absence of a vehicle at a specific position at an intersection is known from an image captured by a fixed-point camera, it is not necessary to separately provide an ultrasonic sensor or the like in order to determine the cycle of a traffic light installed at the intersection, which is desirable.

In order to recognize the existence of a vehicle using a captured image, a process of identifying a region in which the vehicle is reflected from the captured image is usually performed by an image recognition technique. At that time, if the vehicles shown in the captured image overlap each other, there is a problem that the individual vehicles are not separated and erroneous recognition is likely to occur.

Techniques for solving the above problems have been proposed. For example, in Patent Document 1, in a device that measures traffic flow using an image taken by a camera installed on a road, when the distance between traveling vehicles is short, the front and rear vehicles cannot be separated on the image and are erroneously recognized. In order to solve the problem that there are cases, an invention is described in which a fish-eye lens camera installed at a position where the traveling vehicle is overlooked at an angle close to perpendicular to the road surface is used for taking a picture.

Japanese Unexamined Patent Publication No. 2001-216598

According to the invention described in Patent Document 1, even if a plurality of vehicles are shown in a captured image, the vehicles do not overlap on the image, so that the individual vehicles are correctly separated in image recognition and are individually separated. The existence of the vehicle is correctly recognized. However, in the invention described in Patent Document 1, it is necessary to install the camera at a position where the traveling vehicle is overlooked at an angle close to perpendicular to the road surface. In order to install such a camera, there is a problem that it is necessary to install a support column for mounting the camera on the side of the road, which increases the cost. Further, if for some reason the support column for attaching the camera cannot be installed on the side of the road, the invention described in Patent Document 1 cannot be carried out.

Further, when the camera is installed at a position where the traveling vehicle is viewed from directly above, the distance between the camera and the vehicle cannot be increased, so that the angle of view is as used in the invention described in Patent Document 1. It is necessary to perform imaging using a wide-angle fisheye lens or the like. Images captured by such a wide-angle lens are strongly distorted, and therefore do not have the versatility of images captured by a general fixed-point camera.

In view of the above background, it is an object of the present invention to provide a means for correctly recognizing the existence of a plurality of vehicles individually and correctly even if a plurality of vehicles are projected to overlap each other in the captured image.

In order to solve the above-mentioned problems, the present invention is based on the result of comparing two images captured at different times in a predetermined imaging direction, and all of one or more regions in which a moving object is reflected in the image. was identified as a vehicle region, characteristic portions having associated vehicles are common, recognized in each of the identified one or more of the vehicle regions, it recognized the feature in each of the identified one or more of the vehicle regions A vehicle recognition device that measures the number of vehicles based on the number of parts is provided as a first aspect.

According to the vehicle recognition device according to the first aspect, the image of the vehicle is imaged by recognizing the characteristic portion commonly possessed by the vehicle in each of the vehicle regions which are one or more regions in which the moving object is reflected. The number of vehicles is measured without the need for separation. Therefore, even if a plurality of vehicles are overlapped with each other in the captured image, the number of those vehicles is correctly measured.
Further, according to the vehicle recognition device according to the first aspect, the area in which the vehicle is reflected can be specified by a simple process as compared with the case where the area in which the vehicle is reflected is recognized by using the feature amount.

In the vehicle recognition device according to the first aspect, a configuration in which data indicating the number of measured vehicles is output may be adopted as the second aspect.

According to the vehicle recognition device according to the second aspect, the number of vehicles is notified to the external device.

Further, in the present invention , based on the result of comparison of two images captured at different times in a predetermined imaging direction, all of one or more regions in which a moving object is reflected in the image is specified as a vehicle region. sensing a vehicle characteristic portions which vehicles have in common, it is recognized in each of the identified one or more of the vehicle regions, based on the recognized the characteristic portion in each of the identified one or more of the vehicle regions The vehicle recognition device to be used is provided as a third aspect.

According to the vehicle recognition device according to the third aspect, the image of the vehicle is imaged by recognizing the characteristic portion commonly possessed by the vehicle in each of the vehicle regions which are one or more regions in which the moving object is reflected. The vehicle is detected without the need for separation. Therefore, even if a plurality of vehicles are overlapped with each other in the captured image, it is possible to detect those vehicles.
Further, according to the vehicle recognition device according to the third aspect, the area in which the vehicle is reflected is specified by a simple process as compared with the case where the area in which the vehicle is reflected is recognized by using the feature amount.

In the vehicle recognition device according to the third aspect, a configuration in which a detection signal is output in response to the detection of the vehicle may be adopted as the fourth aspect.

According to the vehicle recognition device according to the fourth aspect, the presence / absence of the vehicle is notified to the external device.

In the vehicle recognition device according to any one of the first to fourth aspects, the configuration in which the characteristic portion is a window of the vehicle may be adopted as the fifth aspect.

According to the vehicle recognition device according to the fifth aspect, the occupancy ratio in the vehicle is large among the characteristic parts commonly possessed by the vehicle, the image is easily reflected in the image captured from diagonally above the vehicle, and the vehicle is separated from the vehicle body in the image recognition. Since the relatively easy window is the recognition target, the number of vehicles can be measured or the vehicle can be detected with high accuracy.

In the vehicle recognition device according to the first or second aspect, the vehicle moving based on the result of recognition of the feature portion performed for each of the plurality of images captured in the predetermined imaging direction at each of the plurality of imaging times. A configuration in which the arrival at a predetermined position is detected and the number of vehicles passing through the predetermined position is measured based on the result of the detection may be adopted as the sixth aspect.

According to the vehicle recognition device according to the sixth aspect, the number of vehicles that have passed a predetermined position is measured.

The figure which showed the state that the vehicle recognition device which concerns on 1st Embodiment is installed. The figure which showed the structure of the vehicle recognition apparatus which concerns on 1st Embodiment. The figure which showed the structure of the vehicle recognition table which concerns on 1st Embodiment. The figure for demonstrating the specific procedure of the window area performed by the vehicle recognition apparatus which concerns on 1st Embodiment. The figure for demonstrating the method of detecting a vehicle which has reached a predetermined position performed by the vehicle recognition device which concerns on 1st Embodiment. The figure which showed the flow of the process performed by the vehicle recognition apparatus which concerns on 1st Embodiment. The figure which showed the structure of the vehicle recognition apparatus which concerns on 2nd Embodiment. The figure which showed the structure of the vehicle type table which concerns on 2nd Embodiment. The figure which showed the structural example of the vehicle position table which concerns on 2nd Embodiment. The figure which illustrated the positional relationship between the front end position and the rear end position of a vehicle, and a window area in the captured image which concerns on 2nd Embodiment. The figure which showed the structure of the vehicle recognition table which concerns on 2nd Embodiment. The figure which showed the flow of the process performed by the vehicle recognition apparatus which concerns on 2nd Embodiment. The figure which illustrated the image image | photographed by the fixed point camera which concerns on 2nd Embodiment. The figure which showed the vehicle area and the window area specified by the vehicle recognition device which concerns on 2nd Embodiment. The figure for demonstrating the procedure which the vehicle recognition device which concerns on 2nd Embodiment identifies a vehicle type. The figure which showed the window area specified by the vehicle recognition apparatus which concerns on 2nd Embodiment, and the front end position and the rear end position of a vehicle. The figure which showed the output timing of the detection signal output by the vehicle recognition device which concerns on 2nd Embodiment. The figure for demonstrating the specific procedure of the window area performed by the vehicle recognition apparatus which concerns on one modification.

[First Embodiment]
The vehicle recognition device according to the first embodiment of the present invention will be described below. FIG. 1 is a diagram showing a state in which the vehicle recognition device according to the first embodiment is installed near the road. In FIG. 1, the direction indicated by the arrow N is north. A pillar 8N is installed on the side of the road near the northwest corner of the intersection 6. Similarly, a support 8E is installed near the northeast corner, a support 8S is installed near the southeast corner, and a support 8W is installed near the southwest corner. Hereinafter, these columns are collectively referred to as "support 8". The support column 8 has an L-shape, and is installed so as to extend vertically upward from the ground and then horizontally extend on the road.

A traffic light 7N is attached to a portion extending in the horizontal direction on the road of the support 8N. Similarly, a traffic light 7E is attached to the support 8E, a signal 7S is attached to the support 8S, and a signal 7W is attached to the support 8W. Hereinafter, these traffic lights are collectively referred to as "traffic light 7".

A fixed-point camera 2N is attached in the vicinity of the traffic light 7N on the support column 8N. Similarly, a fixed point camera 2E is attached to the support column 8E, a fixed point camera 2S is attached to the support column 8S, and a fixed point camera 2W is attached to the support column 8W. Hereinafter, these fixed point cameras are collectively referred to as "fixed point camera 2".

Since the fixed point camera 2 continuously captures still images at sufficiently short time intervals, the obtained series of captured images show moving images. The fixed-point camera 2N captures all vehicles passing through the stop line 61S located in the downward direction in front of the vehicle, that is, vehicles traveling on the routes indicated by arrows X, Y, and Z in FIG. Similarly, the fixed point camera 2E images all vehicles passing through the stop line 61W, the fixed point camera 2S images all vehicles passing through the stop line 61N, and the fixed point camera 2W images all vehicles passing through the stop line 61E.

The vehicle recognition device 1N is installed at a predetermined height of the support column 8N, for example, a portion extending vertically upward from the ground. Similarly, a vehicle recognition device 1E is attached to the support column 8E, a vehicle recognition device 1S is attached to the support column 8S, and a vehicle recognition device 1W is attached to the support column 8W. Hereinafter, these vehicle recognition devices are collectively referred to as "vehicle recognition device 1". Each of these vehicle recognition devices 1 is the vehicle recognition device according to the first embodiment.

The vehicle recognition device 1 is connected to a fixed-point camera 2 attached to the same support column 8 by wire or wirelessly, and continuously receives captured image data representing an image captured by the fixed-point camera 2 from the fixed-point camera 2. .. The vehicle recognition device 1 may be arranged in the same housing as the control device (not shown) that controls the traffic light 7.

FIG. 2 is a diagram showing the configuration of the vehicle recognition device 1. The vehicle recognition device 1 may be configured as a dedicated device, or may be realized by a computer performing processing according to a program. The vehicle recognition device 1 includes a captured image acquisition means 11 that receives captured image data from a fixed-point camera 2, a storage means 12 that stores various data, and an area in which the vehicle is reflected in the image represented by the captured image data (hereinafter, "" An image recognition means 14 that recognizes an image of a front glass (an example of a characteristic portion commonly possessed by a vehicle) in a vehicle area specifying means 13 that specifies (referred to as "vehicle area") and a vehicle area specified by the vehicle area specifying means 13. It was measured by the vehicle number measuring means 15 that measures the number of vehicles that have passed a predetermined position based on the result of image recognition by the image recognition means 14, and the captured image data and the vehicle number measuring means 15 received from the fixed point camera 2. An output means 16 for outputting vehicle number data or the like indicating the number of vehicles is provided.

The captured image acquisition means 11 receives, for example, the captured image data transmitted from the fixed point camera 2 every time a predetermined time elapses, associates the received captured image data with, for example, time data indicating the reception time, and stores the stored image data. Store in 12. The time data is data indicating the imaging time. Therefore, for example, the fixed point camera 2 transmits the time data indicating the imaging time to the vehicle recognition device 1 in association with the captured image data, and the captured image acquisition means 11 receives the data and stores the data in the storage means 12. It may be adopted. In either case, the storage means 12 sequentially stores the captured image data indicating the captured images arranged in time series according to the imaging time.

The vehicle area specifying means 13 specifies a different area between two adjacent captured images on the time series represented by the captured image data stored in the storage means 12 as a vehicle area. That is, the vehicle area specifying means 13 compares the values of pixels at the same position (or representative values of the values of the pixel group) included in two adjacent captured images in a time series, and the difference between those values is predetermined. If it is equal to or less than the threshold value of, there is no change in the two captured images in the pixel (or pixel group), and if the difference between the values exceeds a predetermined threshold value, the two images are captured in the pixel (or pixel group). It is determined that the image is changing. Subsequently, the vehicle area specifying means 13 specifies the area indicated by the collection of pixels determined to be changing as the vehicle area. The number of vehicle regions specified by the vehicle region specifying means 13 as different regions between the two captured images is not limited to one, and a plurality of vehicle regions may be specified.

The vehicle area specifying means 13 stores the vehicle area data indicating the specified vehicle area in the storage means 12 in association with, for example, the time data indicating the later of the imaging times of the two captured images. As a result, the storage means 12 sequentially stores vehicle area data indicating vehicle areas arranged in time series according to the imaging time.

The area specified by the vehicle area specifying means 13 as the vehicle area shows an object that has moved within the imaging range of the fixed point camera 2 during the period in which the imaging times of the two captured images to be compared are the start and end. The area. Therefore, the vehicle area may include an area in which a moving object other than the vehicle such as a pedestrian is reflected. In addition, the area in which a vehicle stopped due to waiting for a traffic light or the like is shown is not included in the vehicle area. That is, the vehicle area specified by the vehicle area specifying means 13 may be an area in which a moving vehicle may be reflected, and must be exactly the same as the area in which the vehicle is actually reflected. No.

The image recognition means 14 corresponds to the same time data as the time data associated with the captured image data among the captured images represented by the captured image data with respect to each of the captured image data stored in the storage means 12. The image of the windshield is recognized in the vehicle area indicated by the vehicle area data stored in the storage means 12.

For image recognition performed by the image recognition means 14, the storage means 12 stores in advance reference feature amount data prepared in another device. The reference feature amount data is a collection of data indicating the feature amount referred to when the image recognition means 14 recognizes the image of the windshield. The reference feature data includes, for example, data showing the features of various windshield images given as positive samples and the features of various non-windshield images given as negative samples. Includes data indicating.

As the method for the image recognition means 14 to recognize the image of the windshield, any method may be adopted as long as the method can identify the windshield of the vehicle reflected in the captured image from the captured image. For example, a method may be adopted in which the image recognition means 14 recognizes the image of the windshield reflected in the captured image by matching local feature amounts such as Haar-like feature amount and HOG feature amount. In this case, the reference feature data is a collection of data indicating the local features of a large number of positive and negative samples. Then, the image recognition means 14 extracts a large number of local features from the captured image. Subsequently, the image recognition means 14 belongs to the distribution area of the local feature amount of the positive sample indicated by the reference feature amount data and does not belong to the distribution area of the local feature amount of the negative sample from the extracted local feature amount. Search for quantities. The image recognition means 14 recognizes the region of the image showing the local feature amount searched in this way as a region in which the windshield is reflected (hereinafter, referred to as a “window region”).

The number of window regions recognized by the image recognition means 14 in each of the one or a plurality of vehicle regions included in one captured image is not limited to one, and may be 0 or 2 or more. For example, with respect to a vehicle area containing only an image of a pedestrian and not an image of a vehicle, the image recognition means 14 does not recognize any window area. On the other hand, with respect to the vehicle area in which a plurality of vehicles overlapping each other are shown, the image recognition means 14 recognizes two or more window areas.

The image recognition means 14 stores the window area data indicating the recognized window area in the storage means 12 in association with the time data indicating the imaging time of the captured image to be recognized. As a result, the storage means 12 sequentially stores the window area data indicating the window areas arranged in time series according to the imaging time.

The image recognition means 14 identifies the window regions corresponding to the windshields of the same vehicle indicated by the window region data adjacent to each other in the time series based on the similarity of their shapes and positions, and associates them with each other. The image recognition means 14 assigns a new window ID to the newly recognized window area, and then sequentially associates the window area corresponding to the same windshield as the window area with the window ID. Therefore, the groups of window areas that are sequentially associated with each other are distinguished from the groups of other window areas by the window IDs assigned to the groups of those window areas. A group of window regions associated with the same window ID in chronological order indicates the trajectory of the windshield that moves as the vehicle travels.

In the vehicle number measuring means 15, when the window area reaches a predetermined position in the captured image for each of the group of window areas grouped by the window ID, the vehicle provided with the windshield reflected in the window area is the intersection 6 It is determined that the predetermined position of is reached. That is, the vehicle number measuring means 15 detects that the window region in the captured image has reached a predetermined position, so that the actual vehicle shown in the captured image is placed in a predetermined position in the real space (for example, the vehicle recognition device 1N). The vehicle number measuring means 15 detects that it has reached the position of the stop line 61S).

FIG. 3 is a diagram showing a configuration example of a data table (hereinafter, referred to as “vehicle recognition table”) stored in the storage means 12 for use by the vehicle number measuring means 15 for measuring the number of vehicles. The vehicle recognition table is a collection of data records corresponding to each of the windows identified by the window ID, and has data fields [window ID], [arrival time], [captured image], and the like. A window ID that identifies the window area is stored in the data field [window ID]. In the data field [arrival time], time data indicating the time when the window area identified by the window ID reaches a predetermined position in the captured image is stored. In the data field [captured image], a captured image whose imaging time is the time when the window region identified by the window ID reaches a predetermined position in the captured image is stored.

The vehicle number measuring means 15 measures the number of vehicles that have reached a predetermined position using the vehicle recognition table, and associates the vehicle number data indicating the measurement result with the period data indicating the measurement period to display the predetermined position vehicle recognition table. The process of measuring the number of vehicles using the device will be described later.

The vehicle number measuring means 15 counts, for example, the number of vehicles that have reached a predetermined position using a vehicle recognition table every time a predetermined time elapses, and obtains vehicle number data indicating the counting result as period data indicating a measurement period. Is stored in the storage means 12 in association with. The period data and the number of vehicles data stored in the storage means 12 are output to an external device by the output means 16.

FIG. 4 is a diagram for explaining a specific procedure of the window area performed by the vehicle recognition device 1. _{First, the vehicle area specifying means 13 compares the captured images I 1} and the captured images I ₂ whose imaging times are adjacent to each other in time series, and sets different regions between the captured images, for example, the vehicle region C ₁ and the vehicle. Specify as region C _2.

Subsequently, the image recognition means 14 recognizes an image of the windshield in each of _{the vehicle area C 1} and the vehicle area C ₂ specified by the vehicle area specifying means 13 in the _{captured image I 2.} As a result, _{for example, the window area W 1} , the window area W _2, and the window area W ₃ are specified as the window area in which the windshield is reflected in the _{captured image I 2.} In the example of FIG. 4, the window area W ₁ and the window area W ₂ are the areas of the windshield recognized in the vehicle area C _{1 , and the window area W 3} is the area of the windshield recognized in the _{vehicle area C 2.} The area.

FIG. 5 is a diagram for explaining a method of detecting a vehicle that has reached a predetermined position by the vehicle recognition device 1. The upper figure of _{FIG. 5 shows the positions of the window regions W 1 to} W ₃ recognized in the captured image at the _{imaging time t 1} , and the lower figure of FIG. 5 is recognized in the captured image at _{the imaging time t 2} after the imaging time t _1. The positions of the window areas W _{1 to} W _{3 are shown.} In FIG. 5, line L ₁ is a line indicating a predetermined position. In the example shown in FIG. 5, the number of vehicles measuring means 15 senses that the vehicle in accordance with the window area W ₁ at imaging time t ₂ has reached a predetermined position in the real space (for example, a stop line position).

FIG. 6 is a diagram illustrating a flow of processing performed each time the vehicle recognition device 1 receives and stores captured image data transmitted from the fixed point camera 2. When the captured image data received by the captured image acquisition means 11 is newly stored in the storage means 12, the vehicle area specifying means 13 identifies the vehicle area from the captured images represented by the newly stored captured image data. (Step S101). The vehicle area specifying means 13 stores the vehicle area data indicating the specified vehicle area in the storage means 12 in association with the time data associated with the captured image data (step S102).

Subsequently, the image recognition means 14 has a windshield in a portion of the image represented by the captured image data newly stored in the storage means 12 in the vehicle area indicated by the vehicle area data newly stored in the storage means 12. Recognize the image of (step S103). The image recognition means 14 stores the window area data indicating the window area, which is the area occupied by the recognized windshield in the captured image, in the storage means 12 in association with the time data associated with the captured image data used for recognition. It is stored (step S104). When a plurality of windshield images are recognized in step S103, the plurality of window area data are stored in the storage means 12 in association with the same time data in step S104.

Subsequently, the image recognition means 14 relates to the window area data newly stored in the storage means 12, and the window area data of the previous window area data adjacent to each other in the time series, that is, the window area data of the previous time. It is determined based on the shape and position of the window area indicated by the window area data whether or not there is a window area corresponding to the same front glass (step S105). When the image recognition means 14 determines in step S105 that there is no window area data corresponding to the same windshield in the window area data at the previous time (step S105; No), the window area data Is given a new window ID and stored in the storage means 12 (step S106). Further, in step S106, the image recognition means 14 adds a new data record to the vehicle recognition table (FIG. 3), and stores the newly assigned window ID in the data field [window ID].

On the other hand, when the image recognition means 14 determines in the determination in step S105 that the window area data corresponding to the same windshield is included in the window area data at the previous time (step S105; Yes), the window. The area data is stored in the storage means 12 in association with the window ID assigned to the window area data corresponding to the same windshield at the previous time (step S107).

Subsequently, the vehicle number measuring means 15 moves the window area indicated by the window area data stored in the time order indicated by the time data in association with each of the window IDs stored in the storage means 12. It is determined whether or not the window region has reached a predetermined position in the captured image based on the locus of (step S108).

When the vehicle number measuring means 15 determines that the window area has reached a predetermined position (step S108; Yes), the data record in which the window ID of the window area is stored in the data field [window ID] from the vehicle recognition table. Is searched, and time data indicating the imaging time of the captured image used for recognizing the window area that has reached the predetermined position is stored in the data field [arrival time] of the searched data record (step S109). On the other hand, when it is determined that the window area has not reached the predetermined position (step S108; No), the vehicle number measuring means 15 does not perform the process of step S109. The above is a description of the flow of processing performed each time the vehicle recognition device 1 receives and stores the captured image data transmitted from the fixed point camera 2.

By the above-described processing, individual vehicles passing through the intersection 6 are recognized for each direction of the vehicle entering the intersection 6, and data indicating the recognition result is stored in the vehicle recognition table. As described above, the vehicle number measuring means 15 counts, for example, the number of data records newly stored in the vehicle recognition table every time a predetermined time elapses, and obtains vehicle number data indicating the counted number of vehicles. , It is stored in the storage means 12 in association with the period data indicating the measurement period. Further, when the vehicle number data is newly stored in the storage means 12, the output means 16 outputs the vehicle number data to an external device together with the period data stored in association with the vehicle number data. The above is the description of the processing performed by the vehicle recognition device 1.

In the vehicle recognition device 1 described above, the presence of each vehicle is recognized without the need to separate the image of the vehicle by recognizing the characteristic portion commonly possessed by the vehicle in the area where the vehicle is reflected. Will be done. Therefore, according to the vehicle recognition device 1, even if a plurality of vehicles are overlapped with each other in the captured image, the vehicles are individually and correctly detected correctly.

[Second Embodiment]
The vehicle recognition device 3 according to the second embodiment of the present invention will be described below. The configuration and operation of the vehicle recognition device 3 are common to the configuration and operation of the vehicle recognition device 1 according to the first embodiment in many respects. Therefore, the difference between the vehicle recognition device 3 and the vehicle recognition device 1 will be mainly described below. Further, in the following description, the same reference numerals as those used in the description of the vehicle recognition device 1 are used for the components included in the vehicle recognition device 3 that are common to the components included in the vehicle recognition device 1.

FIG. 7 is a diagram showing the configuration of the vehicle recognition device 3. The vehicle recognition device 3 does not include the vehicle number measuring means 15 included in the vehicle recognition device 1. On the other hand, the vehicle recognition device 3 is a component that the vehicle recognition device 1 does not include, the vehicle type identification means 31 for specifying the vehicle type of the vehicle shown in the captured image, the tip position in the captured image of the vehicle shown in the captured image, and the tip position in the captured image. The vehicle position specifying means 32 for specifying the rear end position and the vehicle detecting means 33 for virtually detecting the existence of the vehicle in a predetermined position in the real space shown in the captured image are provided.

Further, the storage means 12 of the vehicle recognition device 3 stores a vehicle type table and a vehicle position table corresponding to each of the plurality of vehicle types as data tables that are not stored in the storage means 12 of the vehicle recognition device 1. ..

FIG. 8 is a diagram showing a configuration example of a vehicle type table. The vehicle type table is a collection of data records corresponding to each of a plurality of vehicle types, and has data fields [vehicle type], [window area], [vehicle area], and the like. Data indicating the vehicle type (passenger car, bus, etc.) is stored in the data field [vehicle type]. In the data field [window area], data indicating the shape and size of the window area recognized from the captured image of the vehicle of the corresponding vehicle type is stored. In the data field [vehicle area], data showing the shape, size, and positional relationship with the window area of the vehicle area specified from the captured image of the vehicle of the corresponding vehicle type is stored.

FIG. 9 is a diagram showing a configuration example of a vehicle position table. As described above, the storage means 12 stores the vehicle position table corresponding to each of the plurality of vehicle types. The vehicle position table is a collection of data showing the relationship between the front end position of the window region of the vehicle shown in the captured image and the front end position and the rear end position of the vehicle in the captured image. The front end position of the window region and the front end position and the rear end position of the vehicle in the captured image change according to the position in the front-rear direction of the vehicle in the captured image. Therefore, the vehicle position table contains data records corresponding to the front edge positions of the various window areas.

FIG. 10 is a diagram for explaining the positional relationship between the front end position and the rear end position of the vehicle and the window region in the image captured by the fixed point camera 2. FIG. 10 shows a state in which a vehicle imaged from diagonally above right by the fixed point camera 2 is viewed from the right side surface in the traveling direction. In the example of FIG. 10, the vehicle V ₁ and the following vehicle V ₂ are imaged by the fixed point camera 2.

Point F ₁ is a point indicating the front end position of the vehicle V _{1 , and point R 1} is a point indicating the rear end position of the vehicle V _1. The point f ₁ is a point indicating the front end position of the window of the vehicle V _{1 , and the point r 1} is a point indicating the rear end position of the window of the vehicle V _1. Point F ₂ is a point indicating the front end position of the vehicle V _{2 , and point R 2} is a point indicating the rear end position of the vehicle V _2. Point f ₂ is a point indicating the front end position of the window of the vehicle V _{2 , and point r 2} is a point indicating the rear end position of the window of the vehicle V _2. Further, the point L ₂ is a point indicating a predetermined position for detecting the vehicle.

Although the above-mentioned points are actually arranged in the three-dimensional space, they are arranged in the plane in the image captured by the fixed point camera 2. When the image captured by the fixed point camera 2 is _{placed on the plane P, for example, the position of the point F 1} indicating the front end position of the _{vehicle V 1 in the captured image is a straight line connecting the point F 1} and the fixed point camera 2 in the real space. It is the position of the intersection with the plane P. The same applies to other points (point R _{1, etc.).}

Returning to FIG. 7, the description of the configuration of the vehicle recognition device 3 will be continued. The storage means 12 of the vehicle recognition device 3 contains vehicle type data indicating the vehicle type specified by the vehicle type identification means 31 for each vehicle identified by the window ID, and the front end position and the rear of the vehicle specified by the vehicle position identification means 32. Vehicle position data indicating the end position is stored. The storage means 12 of the vehicle recognition device 3 stores a vehicle recognition table having a configuration illustrated in FIG. 11 as a data table for storing these data.

The vehicle recognition table stored in the storage means 12 of the vehicle recognition device 3 is a collection of data records corresponding to each of the windows identified by the window ID, and the data fields [window ID], [vehicle type], and [vehicle front end]. ], [Rear end of vehicle], etc. A window ID that identifies the window area is stored in the data field [window ID]. In the data field [vehicle type], data indicating a vehicle type (passenger car, bus, etc.) specified by the vehicle type specifying means 31 is stored. Data indicating the front end position and the rear end position of the vehicle specified by the vehicle position specifying means 32 are stored in the data fields [vehicle front end] and [vehicle rear end].

Subsequently, the operation of the vehicle recognition device 3 will be described. FIG. 12 is a diagram illustrating a flow of processing performed each time the vehicle recognition device 3 receives and stores captured image data transmitted from the fixed point camera 2. Steps S101 to S107 shown in FIG. 12 are the same processes as steps S101 to S107 performed by the vehicle recognition device 1 shown in FIG.

FIG. 13 is a diagram illustrating an image captured by the fixed point camera 2. In situations where exemplary image I ₃ taken in FIG. 13, shorter than the state in which inter-vehicle distance of the vehicle V ₁ and the vehicle V ₂ is illustrated in FIG. 10, a portion of the vehicle V ₂ is the image I ₃ is the vehicle It is _{blocked by V 1} and is not reflected.

Figure 14 is a vehicle area C ₃ of the vehicle region specifying means 13 is specified in step S101 relates to an image I _3, it relates to an image I ₃ in view the image recognition unit 14 has shown the window area W ₄ and W ₅ recognized in step S103 be.

As described above, in the state where the vehicle area and the window area are specified, the vehicle type specifying means 31 refers to the vehicle type table (FIG. 8) and specifies the vehicle type for each of the window areas (step S201). Vehicle type specifying means 31 first performs a specific model relates the window area W _4. Specifically, vehicle type identification unit 31 converts the shape and size at the standard position the shape and size of the window region W ₄ predetermined conversion formula in accordance (or a conversion table) depending on the position of the window area W _4. Subsequently, vehicle type identification unit 31, from the vehicle type table, the shape and size of the window region indicated by the data stored in the data field window region] is between the shape and size of the window region W ₄ after conversion A data record showing a degree of similarity equal to or higher than a predetermined threshold is extracted.

Subsequently, the vehicle type specifying means 31 has a predetermined ratio in which the vehicle area indicated by the data stored in the data field [vehicle area] of the extracted data record _{overlaps with the vehicle area C 3 specified by the vehicle position specifying means 32.} Determine if it is greater than or equal to the threshold. In FIG. 15, _{the vehicle area Y 1} having the shape, size, and position shown in the data field [vehicle area] of the data record extracted from the vehicle type table with respect to _{the window area W 4} is specified by the vehicle area specifying means 13. is a diagram which is displayed over the vehicle area C _3. The vehicle type specifying means 31 selects a data record for storing data indicating the _{vehicle area Y 1} having the highest ratio of the portion overlapping the _{vehicle area C 3 in the data field [vehicle area].} The vehicle type indicated by the data stored in the data field [vehicle type] of the data record selected in this way is the vehicle type specified with respect _{to the window area W 4 by the vehicle type specifying means 31.}

Vehicle type specifying means 31, subsequently, relates window region W _5, performs a specific vehicle type as well as the window area W _4. The data indicating the vehicle type specified by the vehicle type specifying means 31 is stored in the data field [vehicle type] of the data record corresponding to _{the window areas W 4} and W _{5 of the vehicle recognition table (FIG. 11).}

When the vehicle type identification by the vehicle type identification means 31 is completed, the vehicle position identification means 32 refers to the vehicle position table (FIG. 9) according to the specified vehicle type for each of the _{window areas W 4} and W _{5, and the vehicle position.} Is specified (step S202). Specifically, in the vehicle position specifying means 32, first, _{from the data records included in the vehicle position table corresponding to the vehicle type in the window area W 4} , the data stored in the data field [window front end] is displayed in the window. The data record indicating the _{front end position f 1} in the image of the area W _{4 is searched.} The positions indicated by the data stored in the data fields [vehicle front end] and [vehicle rear end] of the data record searched in this way are the front end position and the rear end position of the vehicle specified by the vehicle position specifying means 32. ..

Vehicle position specifying means 32, subsequently, relates window region W _5, performs a specific front end position and rear end position of the vehicle as well as the window area W _4. The data indicating the front end position and the rear end position of the vehicle specified by the vehicle position specifying means 32 are the data fields [vehicle front end] of the data record corresponding to _{the window areas W 4} and W _{5 of the vehicle recognition table (FIG. 11).} It is stored in the [rear end of the vehicle].

FIG. 16 shows _{the front end position F 1} and the rear end position R _{1 of} the vehicle specified by the vehicle position specifying means 32 with respect to _{the window area W 4,} and the front end position of the vehicle specified by the vehicle position identifying means 32 with respect to the window area W _5. F ₂ and the rear end position R ₂ are shown together with the front end position f ₁ of the window area W _{4 and} the front end position f ₂ of the window area W _5. The range from the front end position F ₁ to the rear end position R ₁ is the range occupied by the vehicle corresponding to the _{window area W 4 , and the range from the front end position F 2} to the rear end position R ₂ corresponds to the window area W _5. It is the range occupied by the vehicle.

Subsequently, the vehicle sensing means 33, a predetermined position indicated by the line L ₂ in FIG. 16 (a predetermined position in the longitudinal direction of the vehicle) is between the front end position and rear end position of any vehicle that is specified in step S202 (Step S203).

In the determination of step S203, when it is determined that the _{predetermined position indicated by the line L 2} is between the front end position and the rear end position of any vehicle (step S203; Yes), the vehicle sensing means 33 detects the vehicle. (Step S204). On the other hand, in the determination of step S203, when it is determined that the _{predetermined position indicated by the line L 2} is not between the front end position and the rear end position of any vehicle (step S203; No), the vehicle sensing means 33 determines the vehicle. Not detected (step S205).

Subsequently, the output means 16 outputs a detection signal corresponding to the detection of the vehicle performed in step S204 or S205 to the external device (step S206). The above is a description of the processing performed by the vehicle recognition device 3.

FIG. 17 shows the position of the actual vehicle in motion, the positions of the windows and the front and rear ends of the vehicle in the captured image, and the output timing of the sensing signal output by the vehicle recognition device 3 (sensing and non-sensing indicated by the sensing signal). It is a figure which showed the relationship with the switching timing). In FIG. 17, the horizontal axis is the time axis, and time elapses from left to right.

The figure in the first column of FIG. 17 is a view from the right side of an actual vehicle traveling near a predetermined position indicated by line L _{2 at each of time t 1} , t ₂ , t ₃ , and t _4. Is. In the figure of the first row of FIG. 17, the vehicle is traveling from left to right. The time t ₁ is the time at which the leading end position of the vehicle V ₁ is reached the line L _2, the time t ₂ is the time at which the rear end position of the vehicle V ₁ is reached the line L _2, time t ₃ Is the time when the tip position _{of the vehicle V 3 reaches the line L 2} , and time t ₄ is the time when the rear end position of the _{vehicle V 2} reaches the _{line L 2.}

The second column of FIG. 17 is a diagram showing a window region recognized in the image captured by the fixed point camera 2 and the front end position and the rear end position of the vehicle specified based on the window region. The window area W ₄ is the window area of the vehicle V ₁ , and the window area W ₅ is the window area of the vehicle V ₂ .

The figure in the third column of FIG. 17 is a diagram showing the output timing of the sensing signal output from the output means 16. As shown in FIG. 17, _{while the line L 2} is between the front end position and the rear end position of the vehicle specified based _{on the window area W 4} , and the front end position of the vehicle specified based on the window area W _{5. A sensing signal indicating that there is a vehicle is output while the line L 2} is between the rear end positions, and a sensing signal indicating that there is no vehicle is output during other periods. As a result, the vehicle recognition device 3 outputs a detection signal indicating that the vehicle is present while one of the vehicles is in _{a predetermined position according to the line L 2 in the real space, and there is no vehicle during the other period.} A sensing signal indicating is output.

According to the vehicle recognition device 3, an external device (for example, a control device for a traffic light) that receives an input of a detection signal from the vehicle recognition device 3 is an occupancy rate (of the time when the vehicle is detected) at a predetermined position on the road by the vehicle. Ratio) can be specified.

[Modification example]
The above-described embodiment can be modified in various ways. Examples of these modifications are shown below. The above-described embodiment and the modifications shown below may be combined as appropriate.

(1) In the first embodiment described above, the vehicle recognition device 1 measures the number of vehicles that have passed a predetermined position. Instead of this, a configuration may be adopted in which the vehicle recognition device 1 measures the number of vehicles shown in the captured image at a specific time.

FIG. 18 is a diagram for explaining a specific procedure of a window region performed by the vehicle recognition device 1 according to this modification. In this modification, the fixed point camera 2 images the entire parking lot. Although the captured image of the parking lot of the parking space for three cars is illustrated in FIG. 18, the number of parking spaces included in the imaging range is not limited to three.

The vehicle recognition device 1 stores in the storage means 12 in advance a _{captured image I 0} , which is an image captured by the fixed point camera 2 in a state where no vehicle is parked in the parking lot, as a reference image.

First, the vehicle area specifying means 13 compares the reference captured image I ₀ with the captured image I ₄ at a certain imaging time, and identifies different regions between the captured images as, for example, the vehicle region C ₄ .

Subsequently, the image recognition means 14 recognizes the image of the windshield in _{the vehicle area C 4} specified by the vehicle area identification means 13 in the _{captured image I 4.} As a result, for example, the window area W ₆ , the window area W _7, and the window area W ₈ are specified.

The vehicle number measuring means 15 specifies the number of the window regions specified as described above as the number of vehicles shown in the _{captured image I 4.} In this case, the number of vehicles is specified as three.

(2) In the above-described embodiment, the fixed-point camera 2 images a vehicle approaching the fixed-point camera 2. The traveling direction of the vehicle captured by the fixed point camera 2 may be arbitrarily changed. For example, when the fixed-point camera 2 images a vehicle moving away from the fixed-point camera 2, the image recognition means 14 may recognize the rear glass in the vehicle region instead of the windshield. Further, when the fixed-point camera 2 images a vehicle crossing the front of the fixed-point camera 2, the image recognition means 14 may recognize the side glass in the vehicle region instead of the windshield.

(3) In the above-described embodiment, the image recognition means 14 recognizes the windshield. The type of the feature portion recognized by the image recognition means 14 is not limited as long as it is a feature portion commonly possessed by the vehicle. For example, an outdoor rear-view mirror (fender mirror or door mirror), an indoor rear-view mirror, or the like may be recognized by the image recognition means 14. Further, in the present invention, the feature portion commonly possessed by the vehicle means the feature portion commonly included in the image of the vehicle. Therefore, when measuring the number of traveling vehicles, the driver who is always reflected in the driver's seat is also an example of the characteristic portion that the vehicles have in common.

It is desirable that the feature portion recognized by the image recognition means 14 is a feature portion provided by each vehicle, but the number is limited to one if each vehicle has the same number of feature portions. No. For example, since the outdoor rearview mirror (fender mirror or door mirror) is a characteristic part provided by each vehicle, the vehicle number measuring means 15 uses 1/2 of the number of rearview mirrors recognized by the image recognition means 14 of the vehicle. It will be counted as the number of units.

(4) In the first embodiment described above, the target vehicles for which the vehicle recognition device 1 measures the number of vehicles need not be all types of vehicles, and the number of only specific types of vehicles is measured by the vehicle recognition device 1. May be done. For example, in Japan, company name indicators are currently installed on the roofs of almost all taxis. Therefore, the number of taxis may be measured by the image recognition means 14 recognizing the company name indicator light.

(5) In the above-described embodiment, the image captured by the fixed-point camera 2 changes greatly depending on the sunlight, the weather, and the like. Therefore, when the image recognition means 14 performs image recognition, a configuration may be adopted in which different reference feature amount data are used depending on the time zone, the presence or absence of rainfall, the illuminance, and the like. According to this modification, the accuracy of image recognition is improved, and as a result, the accuracy of the number of vehicles measured is improved.

For example, when the image recognition means 14 uses reference feature data in which the image recognition means 14 differs depending on the presence or absence of rainfall, the vehicle recognition device 1 or the vehicle recognition device 3 acquires data indicating a rainfall detection result by a rainfall sensor attached to a support column 8 or the like. Then, depending on the presence or absence of rainfall indicated by the data, it is determined whether to use the reference feature data for fine weather or rainy weather.

Further, when the image recognition means 14 uses different reference feature amount data depending on the illuminance, the vehicle recognition device 1 or the vehicle recognition device 3 acquires data indicating the measurement result of the illuminance by the illuminance sensor attached to the support column 8 or the like. Then, the reference feature amount data corresponding to the illuminance band including the illuminance indicated by the data is used.

(6) In the above-described embodiment, the vehicle recognition device 1 or the vehicle recognition device 3 is arranged near the fixed point camera 2. The place where the vehicle recognition device 1 or the vehicle recognition device 3 is arranged is not limited, and may be arranged in, for example, a monitoring center remote from the fixed point camera 2. In this case, the fixed point camera 2 is connected to a transmission device that transmits data to the vehicle recognition device 1 or the vehicle recognition device 3, and the fixed point camera 2 transmits the captured image data to the vehicle recognition device 1 or the vehicle recognition device 3 via the transmission device. Send to.

(7) In the above-described embodiment, the vehicle recognition device 1 or the vehicle recognition device 3 measures or senses the number of vehicles using the captured image data received from one fixed point camera 2. Instead, the vehicle recognition device 1 or the vehicle recognition device 3 receives captured image data from each of the two or more fixed-point cameras 2, and uses the received captured image data to each of the two or more fixed-point cameras 2. A configuration for measuring or sensing the number of vehicles reflected in the image captured by the fixed-point camera 2 may be adopted.

(8) In the above-described embodiment, the vehicle area is specified by the vehicle area specifying means 13 based on the result of comparison of the two images. Instead, the vehicle area may be identified by image recognition by the image recognition means 14. In this case, the storage means 12 stores in advance reference feature amount data indicating the feature amount of the outer edge portion of the vehicle image, and the image recognition means 14 is surrounded by the outer edge of the vehicle recognized using the reference feature amount data. The area to be specified is specified as a vehicle area.

In this modification, the image recognition means 14 only needs to recognize the outer edge of an image in which a plurality of vehicles are overlapped, and does not need to recognize each vehicle separately. Therefore, the number of vehicles is measured or sensed with higher accuracy than when individual vehicles are recognized separately.

(9) In the description of the above-described embodiment, it is assumed that the image captured by the fixed-point camera 2 shows only the vehicle traveling in one lane, but the image captured by the fixed-point camera 2 includes a vehicle traveling in a plurality of lanes. It may be reflected. In this case, the vehicle recognition device 1 or the vehicle recognition device 3 measures or senses the number of vehicles passing in each lane based on the position where the movement locus of each window region specified by the image recognition means 14 passes. You may go.

(10) In the above-described embodiment, the fixed-point camera 2 is assumed to be an optical camera that captures images with a sensor that senses visible light. Alternatively or additionally, the fixed point camera 2 may perform imaging with a sensor that senses electromagnetic waves other than visible light such as infrared rays. For example, the fixed-point camera 2 may perform imaging with visible light during the day and with infrared light at night.

(11) In the first embodiment described above, in the vehicle number measuring means 15, when the lowest point of the window region reaches a predetermined position shown by a broken line in FIG. 5, the vehicle corresponding to the window region is in the actual space. The configuration is adopted in which it is assumed that the predetermined position has been reached. Instead of this, for example, a configuration may be adopted in which the vehicle corresponding to the window area reaches the predetermined position in the real space when the highest point, the center of gravity, or the like of the window area reaches the predetermined position. ..

(12) In the first embodiment described above, the vehicle recognition device 1 outputs vehicle number data indicating the number of detected vehicles to an external device every time a predetermined time elapses. The length of the predetermined time is not limited, and may be, for example, 5 minutes, 1 day, 1 week, 1 month, or the like. Further, the timing at which the vehicle recognition device 1 outputs data indicating the detection result of the vehicle to the external device is not limited to each predetermined time. For example, the vehicle recognition device 1 may promptly output data indicating a detection result to an external device each time it detects that the vehicle has reached a predetermined position.

For example, every time time data is newly stored in the data field [arrival time] of the vehicle recognition table (FIG. 3), the output means 16 outputs the data record in which the time data is stored to an external device. May be adopted. In this configuration, the vehicle recognition device 1 does not need to count the number of vehicles, and an external device that receives data indicating the detection result of the vehicle from the vehicle recognition device 1 counts the number of vehicles as needed. I do.

(13) In the first embodiment described above, the vehicle recognition device 1 stores the detection result of the vehicle in a format in which data indicating the time when the vehicle reaches a predetermined position is stored in the vehicle recognition table (FIG. 3). The format in which the vehicle recognition device 1 stores the detection result of the vehicle is not limited to this. For example, the detection result of the vehicle is stored in the format of a counter indicating the number of vehicles detected in the period from the reference time to the present. May be good. In this modification, when the vehicle number measuring means 15 senses that the window area has reached a predetermined position, the counter is incremented by 1. The counter used in this modification may be either a counter that is reset at a predetermined time (for example, a timing when the date changes), a ring counter that is automatically reset when the maximum value is reached, or the like.

(14) In the second embodiment described above, the vehicle position specifying means 32 specifies the front end position and the rear end position of the vehicle with reference to the front end position of the window region. Instead of this, the vehicle position specifying means 32 specifies the front end position and the rear end position of the vehicle with reference to the positions of feature points other than the front end position, such as the rear end position of the window region and the intermediate position in the front-rear direction. May be adopted. In that case, the vehicle position table (FIG. 9) stores data indicating the positions of those feature points instead of the data indicating the front end positions of the window area.

(15) In the second embodiment described above, the vehicle position specifying means 32 specifies the front end position and the rear end position of the vehicle according to the front end position of the window region with reference to the vehicle position table (FIG. 9). Instead of this, even if the vehicle position specifying means 32 specifies the front end position and the rear end position of the vehicle according to a functional expression in which the front end position (or the position of the feature point instead) of the window area is used as a variable. good.

(16) In the second embodiment described above, the mode of the sensing signal output by the vehicle recognition device 3 is not limited to that described above. For example, a mode may be adopted in which a signal indicating the start of detection is output at the start timing of the detection of the vehicle and a signal indicating the end of detection is output at the end timing of the detection of the vehicle.

1 ... Vehicle recognition device, 2 ... Fixed point camera, 3 ... Vehicle recognition device, 6 ... Intersection, 7 ... Traffic light, 8 ... Support, 11 ... Captured image acquisition means, 12 ... Storage means, 13 ... Vehicle area identification means, 14 ... Image recognition means, 15 ... Vehicle number measuring means, 31 ... Vehicle type identifying means, 32 ... Vehicle position identifying means, 33 ... Vehicle detecting means

Claims (6)

Based on the results of comparison of the two images captured at different times at a predetermined imaging direction, all of one or more regions where those who move in the image are reflected identified as the vehicle area, vehicles are commonly A vehicle that recognizes a characteristic portion having one or more in each of the specified vehicle regions and measures the number of vehicles based on the number of the characteristic portions recognized in each of the specified one or more vehicle regions. Recognition device. The vehicle recognition device according to claim 1, which outputs data indicating the number of measured vehicles. Based on the results of comparison of the two images captured at different times at a predetermined imaging direction, all of one or more regions where those who move in the image are reflected identified as the vehicle area, vehicles are commonly A vehicle recognition device that recognizes a feature portion having a feature portion in each of the specified one or more vehicle regions and senses a vehicle based on the feature portion recognized in each of the specified one or more vehicle regions. The vehicle recognition device according to claim 3, which outputs a detection signal in response to the detection of the vehicle. The vehicle recognition device according to any one of claims 1 to 4, wherein the characteristic portion is a window of a vehicle. Based on the result of recognition of the feature portion performed for each of the plurality of images captured in the predetermined imaging direction at each of the plurality of imaging times, it is detected that the moving vehicle has reached a predetermined position, and the result of the detection. The vehicle recognition device according to claim 1 or 2, which measures the number of vehicles that have passed the predetermined position based on the above.

Images