JP5838750B2 - Object recognition system and object recognition apparatus - Google Patents

Description

  The present invention relates to an object recognition system and an object recognition device for recognizing an object existing in a region imaged by an imaging device such as a line sensor using imaging data acquired by the imaging device.

  An object recognition system that recognizes an object such as a vehicle passing through a region imaged by the imaging device using imaging data such as an intensity image obtained by an imaging device such as a line sensor is known (for example, see Patent Document 1). ). In such an object recognition system, template data for various objects is prepared, and an object is recognized by pattern matching between the plurality of template data and the acquired imaging data.

JP 2010-14706 A

  However, in the conventional object recognition system, since the data acquisition rate in the imaging device is constant, the image of the object included in the imaging data is distorted depending on the speed of the object passing through the region, for example, the flatness changes in the case of an axle. . Therefore, the subject that the precision of matching with the template data about the target object which passes at a standard speed fell. Further, when template data corresponding to various speeds is prepared for each object, there is a problem that the number of template data becomes enormous and the processing time for pattern matching processing increases.

  The present invention has been made to solve the above-described problems, and it is an object of the present invention to provide an object recognition system and an object recognition apparatus capable of highly accurate object recognition without increasing processing time. .

An object recognition system according to the present invention includes an imaging unit that outputs imaging data obtained by irradiating a predetermined region through which an object passes, an object speed detection unit that detects a speed at which the object passes, and the imaging The size of the image of the object included in the imaging data output from the means is changed according to the detection speed, and the imaging data after the change is compared with the template data corresponding to the reference speed, or the object or the An object recognizing unit for recognizing a part of the object, and the object recognizing unit includes a speed comparing unit that compares the detected speed with a preset speed threshold, and the detected speed is faster than the speed threshold. In this case, the image size of the object included in the imaging data is changed according to the detection speed, and the changed imaging data is compared with the template data corresponding to the reference speed. Then, when the object or part of the object is recognized and the detection speed is slower than the speed threshold or the speed of the object is not detected, a detection flag is calculated from the histogram calculated for the altitude direction of the imaging data. It generates, characterized that you recognize some of the object or the object from the pattern of the detection flag.

An object recognition apparatus according to the present invention is an object recognition apparatus that recognizes an object or a part of an object by inputting imaging data obtained by irradiating a predetermined region through which the object passes with laser light, and the object passes through the object recognition apparatus. An object speed detecting means for detecting a speed to be changed, and the size of the image of the object included in the imaging data output from the imaging means is changed according to the detection speed, and the changed imaging data and the reference speed are supported. Object recognition means for recognizing the object or part of the object by comparing with the template data to be processed , the object recognition means comprises a speed comparison unit for comparing the detection speed with a preset speed threshold value. And when the detection speed is faster than the speed threshold, the image size of the object included in the imaging data is changed according to the detection speed, and the changed imaging data and reference Compare the template data corresponding to the degree to recognize the object or a part of the object, and if the detection speed is slower than the speed threshold or the speed of the object is not detected, generates a detection flag from the histogram calculated for altitude direction, characterized that you recognize some of the object or the object from the pattern of the detection flag.

  According to the object recognition system and the object recognition device of the present invention, the size of the image of the object included in the imaging data is changed according to the speed of the detected object, and pattern matching processing is performed between the changed imaging data and the template data. Therefore, there is an effect that the object can be recognized with high accuracy without increasing the processing time.

1 is a configuration diagram of an object recognition system according to Embodiment 1. FIG. 4A and 4B are diagrams for explaining the arrangement of a first imaging device and a second imaging device according to Embodiment 1. FIG. FIG. 4 is a diagram illustrating an image of a vehicle that is an object included in imaging data according to Embodiment 1. The figure which looked at the vehicle which concerns on Embodiment 1 from the front. FIG. 3 is a configuration diagram of an object recognition unit according to the first embodiment. The figure for demonstrating the size change of the image of the vehicle contained in the imaging data which concerns on Embodiment 1. FIG. 3 is a flowchart for explaining the operation of the first embodiment. FIG. 3 is a configuration diagram of an object recognition system according to a second embodiment. FIG. 6 is a diagram for explaining the arrangement of a first imaging device and a speedometer according to a second embodiment. 9 is a flowchart for explaining the operation of the second embodiment. FIG. 6 is a configuration diagram of an object recognition unit according to Embodiment 3. FIG. 10 is a diagram for explaining an example of altitude histogram calculation processing according to the third embodiment. FIG. 10 is a diagram for describing an example of a detection flag calculation process for imaging data according to Embodiment 3. 10 is a flowchart for explaining the operation of the third embodiment. FIG. 10 is a diagram for explaining another example of altitude histogram calculation processing according to the third embodiment. FIG. 10 is a diagram for explaining another example of detection flag calculation processing for imaging data in the third embodiment. FIG. 6 is a configuration diagram of an object recognition unit according to a fourth embodiment.

Embodiment 1 FIG.
Embodiment 1 of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of an object recognition system according to the first embodiment. FIG. 2 is a diagram for explaining the arrangement of the first imaging device and the second imaging device in the first embodiment. FIG. 3 is a diagram illustrating an image of a vehicle that is an object included in the imaging data according to the first embodiment. FIG. 4 is a front view of the vehicle in the first embodiment. FIG. 5 is a configuration diagram of the object recognition unit according to the first embodiment. FIG. 6 is a diagram for explaining the size change of the image of the vehicle included in the imaging data in the first embodiment. FIG. 7 is a flowchart for explaining the operation of the first embodiment.

  As shown in FIG. 1, the object recognition system includes a first imaging device (first imaging unit) 1, a second imaging device (second imaging unit) 2, and an object recognition device 3. The Such an object recognition system is applied to, for example, a gate intrusion monitoring system. Here, an object that enters the gate is described as a vehicle. However, the present invention is not limited to this and may be an animal, a person, or the like. The same applies to the following.

  First, the first imaging device 1 and the second imaging device 2 will be described. The first imaging device 1 and the second imaging device 2 correspond to, for example, a reception scanless type laser image measurement device. The first imaging device 1 irradiates the first region and the second imaging device 2 irradiates the second region with laser light, and acquires the intensity value at each point in the irradiated region as intensity data. . Further, the distance value at each point where the region is irradiated is acquired as distance data. The first imaging device 1 and the second imaging device 2 output the acquired intensity data and distance data to the object recognition device 3 as first imaging data and second imaging data, respectively, at predetermined time intervals. To do.

  Note that the region irradiated with the laser beam may be expressed as an observation region or an irradiation region. In the following, description will be made simply as a region (first region, second region).

  As shown in FIG. 2, the first image pickup device 1 and the second image pickup device 2 are installed near a traffic route of the vehicle with a predetermined interval w (m). That is, the second imaging device 2 is arranged at a predetermined interval in the traveling direction (passing direction) of the vehicle with respect to the first imaging device 1. In the example of FIG. 2, the first imaging device 1 and the second imaging device 2 are respectively installed on the opposite sides of the passage route. However, the arrangement is not limited to this arrangement, and may be on the same side. Further, the interval w may be about 0.8 m, for example.

  When the vehicle moves in the traveling direction and passes through the first region where the first imaging device 1 irradiates laser light, the first imaging device 1 acquires intensity data and distance data about the vehicle, and the object recognition device. 3 is output. At this time, the first imaging device 1 may output the intensity data as image data (intensity image). As shown in FIG. 3, the intensity image has an intensity value for each pixel in a plane in which the horizontal axis is defined as the line direction corresponding to the vehicle traveling direction and the vertical axis is defined as the data direction corresponding to the scanning direction of the imaging device (sensor). And an intensity value in the depth direction viewed from the first imaging device 1 may be included.

  Similarly, when the vehicle moves in the traveling direction and the second imaging device 2 passes through the second region irradiated with the laser light, the second imaging device 2 acquires intensity data and distance data about the vehicle. To the object recognition device 3.

  Here, the first imaging device 1 and the second imaging device 2 are taken as an example of a line sensor that scans in a line, but the present invention is not limited to this, and at least intensity data and distance data are acquired. Anything is possible. Instead of intensity data, a CCD (Charge Coupled Device) camera or the like that can acquire luminance data composed of luminance values at each point where the region is irradiated may be used. It should be noted that intensity data (or luminance data) and distance data are converted into intensity images (or luminance images), distance images, and image formats, respectively. The first imaging device 1 acquires the first imaging data acquired by irradiating the first region with laser light, and the second imaging device 2 acquires the second imaging device acquired by irradiating the second region with laser light. The imaging data is output to the object recognition device 3.

  Also, the first imaging device 1 and the second imaging device 2 calculate an altitude value (height value) and a depth distance value from the acquired distance data, and output these as imaging data to the object recognition device 3. May be. As shown in FIG. 4, when the depth direction viewed from the imaging device is defined as the x axis and the altitude direction that is the height direction of the imaging device is defined as the z axis, the direction perpendicular to the xz plane is the traveling direction. The second imaging device 1 (second imaging device 2) is installed such that the sensor height H, the beam scanning angle half angle θ, and the sensor off-nadir angle φ are in the z-axis direction. In this case, assuming that the pixel number in the data direction in the distance image is i, the maximum number of pixels is N, and the distance value of each point in the region is Ri, the altitude value and the depth distance value are obtained as in Expression (1).

  The first imaging device 1 and the second imaging device 2 constitute imaging means (first imaging means and second imaging means), respectively. In addition, the imaging unit may be configured by the first imaging device 1 and the second imaging device 2.

  Next, the object recognition device 3 will be described. As shown in FIG. 1, the object recognition device 3 includes an imaging data storage unit 31, an object detection unit 32, a speed detection unit 33, and an object recognition unit 34. The imaging data and the second imaging data are output from the second imaging device 2. The imaging data storage unit 31, the object detection unit 32, the speed detection unit 33, and the object recognition unit 34 respectively include the imaging data storage unit 31, the object detection unit 32, the speed detection unit 33, and the object recognition unit 34. Configure.

  The imaging data storage unit 31 stores the first imaging data and the second imaging data output from the first imaging device 1 and the second imaging device 2. A plurality of template data are stored in advance. This template data is data having a feature amount (feature pattern) of an object passing through a traffic route at a preset reference speed. Examples of the object include various types of vehicles such as trucks, passenger cars, and light cars. Or a person or an animal. Here, how to determine the reference speed may be the maximum vehicle passing speed assumed when performing vehicle recognition, or may be arbitrarily determined by the user. Hereinafter, as an example, the speed Vb at which the vehicle passes through the gate is generally described as a reference speed.

  The object detection unit 32 detects a vehicle existing in the area based on the imaging data acquired from the imaging data storage unit 31. For example, the object detection unit 32 acquires in advance intensity data (background data) in a state where no vehicle exists in the region, and at each point of the background data and first imaging data in the state where the vehicle exists. The intensity difference is read, and if the number of points where the intensity difference exceeds a predetermined threshold is greater than or equal to the set number, it is detected that there is an object vehicle. When the object detection unit 32 detects a vehicle existing in the first region from the first imaging data output to the first imaging device 1, the object detection unit 32 outputs a first detection signal representing detection of the vehicle, When a vehicle existing in the second region is detected from the second imaging data output to the second imaging device 2, a second detection signal representing the detection of the vehicle is output.

  The speed detection unit 33 calculates the speed at which the vehicle passes from the time difference between the input of the first detection signal and the second detection signal input from the object detection unit 32. That is, the speed detection unit 33 has information about the distance w between the first imaging device 1 and the second imaging device 2 in advance, so that the first imaging data output from the first imaging device 1 and If the timing difference of the output of the second imaging data output from the second device 2 can be grasped, the passing speed of the vehicle can be calculated. More specifically, if the data acquisition rate of the first imaging device 1 and the second imaging device 2 is fs (Hz), and the number of data acquisitions in the line direction in which the vehicle is detected in each imaging device is S, The speed V is obtained as shown in Equation (2). The speed detection unit 33 outputs the detected speed to the object recognition unit 34.

  The object detection unit 32 and the speed detection unit 33 constitute an object speed detection unit.

  The object recognition unit 34 changes the image size of the vehicle included in the imaging data in accordance with the detection speed V, and compares the changed imaging data with the template data corresponding to the reference speed Vb, thereby including the imaging data. Recognize the vehicle or part of the vehicle. The recognition result is output to a higher-level device such as a PC (Personal Computer) (not shown in FIG. 4). As illustrated in FIG. 5, the object recognition unit 34 includes an image size changing unit 341, a determination unit 342, and a recognition processing unit 343.

  The image size changing unit 341 changes the size of the vehicle image included in the imaging data acquired from the imaging data storage unit 31 based on the speed V detected by the speed detection unit 33. Specifically, the image size changing unit 341 performs bilinear interpolation, bicubic interpolation, or the like so that the length of the vehicle line direction, that is, the passing direction (lateral direction) is changed with respect to the vehicle image included in the imaging data. To change the size. Note that the process of changing the size of the image may be expressed as correcting the image.

  The extent to which the size of the image in the line direction is changed may be changed in proportion to the ratio of the detection speed V to the reference speed Vb, for example. As shown in FIG. 6, when the length in the line direction of the vehicle passing at the detection speed V is A, the length in the line direction of the vehicle after the size change is A × V / Vb. When the detection speed V is faster than the reference speed Vb, the image of the vehicle included in the imaging data is an image contracted in the line direction than the image corresponding to the reference speed, and extends in the line direction when the size is changed. It becomes like this. On the other hand, when the detection speed V is slower than the reference speed Vb, the image of the vehicle included in the imaging data is an image that extends in the line direction from the image corresponding to the reference speed, and the line direction is changed by resizing the image. To shrink. If the image is an axle, the flatness of the axle after size conversion will approach zero.

  The length in the line direction after the size change corresponds to the length in the line direction of the vehicle when the vehicle passes at the reference speed Vb. Therefore, regardless of the speed at which the vehicle passes, the distortion of the image due to the passing speed is corrected to match the image size when passing at the reference speed Vb.

  Note that the size change includes those in which the aspect ratio is further changed after the size is changed so that the length in the line direction of the image is increased. Further, instead of increasing (shortening) the length in the line direction, the length in the data direction (vertical direction) may be changed to be short (longer), or both the lengths in the line direction and data direction may be changed. You may change size by changing.

  The determination unit 342 receives the imaging data whose image size has been changed by the image size changing unit 341, and performs pattern matching processing on the imaging data and the template data stored in advance in the imaging data storage unit 31. The determination unit 342 performs comparison by moving the feature pattern of the target object included in the template data in the xy direction on the search target image that is an image of the vehicle included in the captured image data. The feature pattern of an object is detected. That is, the determination unit 342 determines whether or not the image of the vehicle that is an object included in the imaging data has the feature pattern of the target included in the template data. The determination result is output to the recognition processing unit 343.

  Here, the image included in the imaging data corresponds to an intensity image, a luminance image, or a distance image. Further, matching between an image having altitude value and depth distance value data and a template image may be used, or three-dimensional shape matching based on a distance value in the line direction calculated using the detection speed V may be used. . Assuming that the relative number of lines of scanning for detection is j, the distance value Yj in the line direction is obtained as shown in Equation (3).

  As a pattern matching method, template matching processing or texture matching processing may be used. The template data is stored in the imaging data storage unit 31, but may be stored in another storage device (memory) in advance, or the determination unit 342 has an internal memory and stores it in advance. You may keep it. Further, the template data has been described as having the feature amount of the object, but it may be the image data of the object itself.

  The recognition processing unit 343 obtains the recognition result of the object based on the determination result from the determination unit 342, and outputs the recognition result to an external device such as a PC. That is, when the determination unit 342 receives a determination result that the imaging data has the feature pattern of the target object from the determination unit 342, the recognition processing unit 343 recognizes the object included in the imaging data as the target object and outputs the recognition result. . For example, if the imaging data has a feature pattern of a light vehicle, the recognition processing unit 343 recognizes the object as a light vehicle, and if it has a track feature pattern, the recognition processing unit 343 recognizes the object as a truck and Output the recognition result.

  Next, the operation of the first embodiment of the present invention will be described with reference to FIG.

  The first imaging device 1 and the second imaging device 2 each output imaging data (first imaging data and second imaging data) acquired at predetermined time intervals to the imaging data storage unit 31 of the object recognition device 3. (Step S1).

  As shown in FIG. 2, when the vehicle moves in the vehicle traveling direction and enters the first region, and the vehicle is present in the first imaging data output by the first imaging device 1 (step S2-Yes), The object detection unit 32 detects a vehicle that is an object in the first imaging data acquired from the imaging data storage unit 31. And the object detection part 32 outputs the 1st detection signal which shows that a vehicle exists to the speed detection part 33 (step S3). At this time, it can be detected that an object is present in the imaging data, but it cannot be recognized that the object in the imaging data is a vehicle.

  Next, when the vehicle enters the second area and the vehicle is present in the second imaging data output from the second imaging device 2 (step S <b> 4 -Yes), the object detection unit 32 starts from the imaging data storage unit 31. A vehicle that is an object in the acquired second imaging data is detected. And the object detection part 32 outputs the 2nd detection signal which shows that a vehicle exists to the speed detection part 33 (step S5).

  When both the first detection signal and the second detection signal are input, the speed detection unit 33 detects the passing speed V of the vehicle from the input time difference between the detection signals, and the detected speed is detected by the object recognition unit 34. (Step S6).

  When the detection speed V from the speed detection unit 33 is input, the image size changing unit 341 of the object recognition unit 34 acquires the imaging data from the imaging data storage unit 31 and is included in the imaging data according to the detection speed V. The size of the vehicle image to be changed is changed (step S7). By changing the size in this way, the length in the line direction of the vehicle corresponds to the length in the line direction when the vehicle passes at the reference speed Vb.

  When the image size changing unit 341 performs the size changing process, the image size changing unit 341 outputs the changed imaging data to the determining unit 342. Here, the imaging data including the image to be resized is either the first imaging data captured by the first imaging apparatus 1 or the second imaging data captured by the second imaging apparatus 2. May be.

  When the imaging data including the changed image is input from the image size changing unit 341, the determination unit 342 compares the changed imaging data with a plurality of template data stored in the imaging data storage unit 31 ( In step S8), it is determined whether or not the imaging data has a feature pattern of an object included in the template data, and the determination result is output to the recognition result output unit 343.

  For example, when template data A, B, and C are stored in the imaging data storage unit 31, and each template data has a characteristic pattern of a light vehicle, a passenger car, and a truck that are objects, the determination unit 342 displays Pattern matching processing is performed on the imaging data and the template data A, B, and C, respectively. Here, description will be made assuming that the imaging data includes the feature pattern of the template data C.

  At this time, the length in the line direction of the vehicle included in the imaging data is resized to correspond to the length in the line direction when the vehicle passes at the reference speed Vb, regardless of the speed. Further, the feature patterns of the objects included in the template data A, B, and C are made to correspond to the size when passing at the reference speed Vb. Therefore, it is possible to prevent mismatch due to different passing speeds of the vehicles and to prevent the matching accuracy of the pattern matching process from being lowered. In addition, since it is not necessary to prepare template data corresponding to various passing speeds, it is possible to reduce the calculation cost in the pattern matching process.

  The recognition processing unit 343 recognizes the object based on the determination result in the determination unit 342, and outputs the result (step S9). That is, when the determination result that the imaging data includes the feature pattern of the template data C is input from the determination unit 342, the recognition processing unit 343 recognizes that the vehicle that is the object included in the imaging data is a truck. And output the result.

  Here, the determination unit 342 may determine whether a part of the object included in the resized imaging data is the same as the target object. That is, the template data may be data having an axle feature amount, and pattern matching processing may be performed using image data including a vehicle image and template data having an axle feature amount. By doing so, it is possible to determine and recognize whether or not the axle is present in a part of the vehicle that is the object. As for template data, it is only necessary to store template data having axle feature values when passing at a reference speed, and a plurality of template data having axle feature values corresponding to a plurality of speeds are not required. Therefore, the amount of template data patterns to be saved can be greatly reduced.

  As described above, according to the first embodiment of the present invention, the size of the object image included in the imaging data is changed according to the detected speed of the object, and the changed imaging data and the template data are subjected to pattern matching. Since the object is recognized by processing, the accuracy of matching is improved and the object recognition can be performed with high accuracy.

  In addition, since it is not necessary to prepare a plurality of template data corresponding to a plurality of speeds in advance, the amount of template data to be subjected to pattern matching processing with imaging data is reduced, the calculation cost in the pattern matching processing is reduced, and the processing time is increased. It becomes possible to prevent.

  In addition, by changing the size of the image of the object included in the imaging data in accordance with the speed of the vehicle that is the detected object, pattern matching processing is performed on the imaging data after the change and the template data having the axle feature amount. Thus, it is possible to improve the matching accuracy and reduce the calculation cost, thereby recognizing the axle of the passing vehicle.

  The first imaging data output from the first imaging device 1 and the second imaging data output from the second imaging device 2 are temporarily stored in the imaging data storage unit 31 before the object detection. However, the present invention is not limited to this configuration, and is output directly from the first imaging device 1 and the second imaging device 2 to the object detection unit 32 and the object recognition unit 34, for example. It is good. The same applies to the following.

  The imaging data storage unit 31 has been configured to be provided in the object recognition device 3 so far, but is not limited thereto, and may be configured to be provided in a storage device outside the object recognition device 3. The same applies to the following.

  In the above description, a plurality of template data has been described as being stored in the imaging data storage unit 31. However, the present invention is not limited to this, and may be stored in another memory in the object recognition device 3, for example. You may preserve | save in the internal memory of the object recognition part 34. FIG. The same applies to the following.

Embodiment 2. FIG.
The second embodiment of the present invention will be described below with reference to the drawings. FIG. 8 is a configuration diagram of an object recognition system according to the second embodiment. FIG. 9 is a diagram for explaining the arrangement of the first imaging device and the speedometer in the second embodiment. FIG. 10 is a flowchart for explaining the operation of the second embodiment. Portions corresponding to the configuration of the image display system according to the first embodiment are denoted by the same reference numerals as those in FIG.

  As shown in FIG. 8, the object recognition system of Embodiment 2 includes a first imaging device 1, an object recognition device 3, and a speedometer 4. The object recognition device 3 includes an imaging data storage unit 31 and an object. It has a recognition unit 34. Similarly to the first embodiment, the object recognition unit 34 includes an image conversion unit 341, a determination unit 342, and a recognition result output unit 343 illustrated in FIG. The object recognition system of the second embodiment is different from that of the first embodiment in that there is one imaging device and that the object recognition device 3 does not include the object detection unit 32 and the speed detection unit 33.

  As shown in FIG. 9, the speedometer 4 is installed at a predetermined distance from the first imaging device 1 near the traffic route. The speedometer 4 detects the speed of the object and outputs it to the speed comparison section 340 of the object recognition section 34 when an object such as a passing vehicle enters the speed detection region. The speedometer 4 constitutes an object speed detection means.

  The operation of the second embodiment of the present invention will be described with reference to FIG.

  The first imaging device 1 outputs the first imaging data acquired at predetermined time intervals to the imaging data storage unit 31 of the object recognition device 3 (step S01).

  When the vehicle moves in the vehicle traveling direction and enters the region where the speedometer 4 detects the speed, and the speedometer 4 detects the speed of the vehicle (step S02-Yes), the detected speed is displayed on the image of the object recognition unit 34. It outputs to the size change part 341 (step S03). At this time, it cannot be recognized that the object in the imaging data is a vehicle.

  Since the processing (steps S04 to S06) after the speed detected by the speedometer 4 is input to the image size changing unit 341 is the same as steps S7 to S9 of FIG. 7 described in the first embodiment, Description is omitted.

  As described above, according to the second embodiment of the present invention, the speed of the object is detected by the speedometer 4, and the size of the image of the object included in the imaging data is changed according to the detected speed. Since the object is recognized by pattern matching processing of the imaging data and the template data, the number of installed imaging devices is reduced as compared with the first embodiment, and the object recognition processing and the speed detection processing are not required in the object recognition device 3. Therefore, in addition to having the same effect as that of the first embodiment, it is possible to further reduce the calculation cost and reduce the scale of the apparatus.

Embodiment 3 FIG.
Embodiment 3 of the present invention will be described below with reference to the drawings. FIG. 11 is a configuration diagram of an object recognition unit according to the third embodiment. FIG. 12 is a diagram for explaining an example of the altitude histogram calculation process according to the third embodiment. FIG. 13 is a diagram for explaining an example of a detection flag calculation process for imaging data according to the third embodiment. FIG. 14 is a flowchart for explaining the operation of the third embodiment. FIG. 15 is a diagram for explaining another example of the altitude histogram calculation process according to the third embodiment. FIG. 16 is a diagram for explaining another example of the detection flag calculation process for the imaging data in the third embodiment. Portions corresponding to the configuration of the image display system according to the first embodiment are denoted by the same reference numerals as those in FIG.

  In the object recognition system of the third embodiment, the configuration of the object recognition device 3 is different from that of the first embodiment. As shown in FIG. 11, a speed comparison unit 340, an averaging processing unit 344, and a histogram calculation region setting are newly added. Unit 345, altitude histogram calculation unit 346, data existence probability calculation unit 347, and detection flag generation unit 348. The speed comparison unit 340, the averaging processing unit 344, the histogram calculation region setting unit 345, the altitude histogram calculation unit 346, the data existence probability calculation unit 347, and the detection flag generation unit 348 are respectively compared with the speed comparison unit 340 and the averaging. The processing unit 344, the histogram calculation region setting unit 345, the altitude histogram calculation unit 346, the data existence probability calculation unit 347, and the detection flag generation unit 348 are configured.

  When the detection speed V is input from the speed detection unit 33, the speed comparison unit 340 compares the detection speed V with a preset speed threshold Vth, and if the detection speed V is higher than the speed threshold Vth, the speed comparison unit 340 detects the detection speed V. The speed V is output to the image size changing unit 341. On the other hand, when the detected speed V is slower than the speed threshold Vth, or when an error signal indicating that the vehicle speed cannot be detected is input from the speed detector 33, the speed comparison unit 340 The detection speed V is not output to 341, and an instruction signal instructing the averaging processing unit 344 to perform processing is output. Here, the case where the speed of the vehicle cannot be detected corresponds to the case where the vehicle stops in front of the imaging device, for example.

  The speed threshold Vth is the lowest speed value at which the vehicle can be recognized within the time limit for performing object recognition. That is, it is the minimum speed necessary for storing an image including the entire shape of the vehicle, and when the vehicle passing speed falls below the Vth, imaging data including an image of the entire shape of the vehicle within the time limit is obtained. It will not be obtained. Details of how to obtain the speed threshold Vth will be described later.

  When an instruction signal is input from the speed comparison unit 340, the averaging processing unit 344 acquires imaging data from the imaging data storage unit 31, and performs a moving average process on the line direction of the vehicle image included in the imaging data. . By performing this moving average processing, assuming that the moving average score is NA, the SNR (Signal to Noise Ratio) increases by the square root multiple of the NA, and the image quality can be improved. The moving average score NA can be arbitrarily determined. As an example, when the time limit for performing object recognition is TL, the moving average score NA is obtained as shown in Expression (4).

  Instead of the averaging processing unit 344, an IIR filter such as a FIR (Finite Impulse Response) filter or a Butterworth filter may be used, and the order may be NA.

  The imaging data here corresponds to three-dimensional data composed of intensity data and distance data, but is not limited to this, and may correspond to only three-dimensional data of distance data, for example. Further, the imaging data here may be either the first imaging data or the second imaging data, and the following will be described simply as imaging data.

  The histogram calculation region setting unit 345 narrows down the first region (or the second region) to a predetermined region with respect to the imaging data that has been subjected to the moving average processing by the averaging processing unit 344, and performs imaging for the narrowed down predetermined region. Output data. This predetermined area is defined as a histogram calculation area and will be described below.

  As shown in FIG. 12, the histogram calculation area is an area surrounded by an upper limit value H1a in the altitude direction, a lower limit value H2a, an upper limit value S1a in the depth distance direction, and a lower limit value S2a. For example, the upper limit value H1a in the altitude direction is set to the maximum height that can be imaged by the imaging device, the lower limit value H2a in the altitude direction is set to the lowest ground altitude of the domestic vehicle, and the upper limit value S1a in the depth distance direction is set by the vehicle. The lower limit value S2a in the depth distance direction is set to the minimum distance between the imaging device and the vehicle, but is not limited to this.

  The altitude histogram calculation unit 346 divides the histogram calculation area of the imaging data into N (altitude) directions to create N total sections (height bins) Bha and calculates the altitude histogram Ca. The width of the height bin Bha can be set arbitrarily, but may be a value corresponding to the spatial resolution in the height direction of the imaging data, for example.

  In FIG. 12, the division number of the height bin is set to 2, and a plurality of imaging data corresponding to the roof portion Ba of the vehicle are stored in the upper bin, and the vehicle door portion Ba, vehicle floor portion, and tire portion A ′. A plurality of image data corresponding to the above are stored in the lower bin. Since the road surface is outside the histogram calculation area, imaging data corresponding to the road surface is not stored in the height bin. Then, the altitude histogram calculation unit 346 outputs the calculated histogram to the data existence probability calculation unit 347.

  Based on the altitude histogram Ca calculated by the altitude histogram calculation unit 346, the data existence probability calculation unit 347 includes the bin number n in which a predetermined number or more of intensity data or distance data is stored, and the total number N of bins divided in the height direction. The ratio n / N is calculated as the data existence probability Pa. Then, the data existence probability calculating unit 347 acquires the data existence probability Pa along the cross section of the vehicle included in the imaging data shown in FIG. 13A, that is, along the line direction (corresponding to FIG. 13B). Then, the data existence probability calculation unit 347 outputs data about the data existence probability Pa obtained along the line direction of the vehicle to the detection flag generation unit. Note that the image of the vehicle included in the imaging data is an image in which a part of the entire shape of the vehicle is missing as shown in FIG. 13A because the detected speed V of the vehicle is below the speed threshold Vth. It becomes.

  The detection flag generation unit 348 generates a detection flag by comparing the data presence probability Pa calculated by the data presence probability calculation unit 347 with a preset threshold value.

  The detection flag generation unit 348 has a plurality of threshold sets (upper limit threshold, lower limit threshold). When the data existence probability Pa is higher than the upper limit threshold, the detection flag is H, lower than the upper limit threshold, and lower than the lower limit threshold. If it is higher, M, and if it is lower than the lower threshold, L. A plurality of threshold value sets are different for each object, that is, for each vehicle type and each vehicle size. For example, a threshold value set D for trucks, a threshold value set E for light vehicles, and a threshold value set F for passenger cars Are different values. As described above, the detection flag generation unit 348 compares a plurality of threshold sets having different values depending on the object with the data existence probability Pa.

  For example, when the data existence probability Pa shown in FIG. 13B is input, the detection flag generation unit 348 compares the data existence probability Pa with the upper limit threshold and the lower limit threshold. Here, the vehicle that is the object included in the imaging data is a passenger car, and the threshold set for comparison is also the threshold set F for the passenger car. As a result of the comparison, among the imaging data shown in FIG. 13A, the data existence probability Pa in the tire portion of the vehicle is H because it exceeds the upper threshold. The data existence probability Pa in the roof portion of the vehicle is M because it is greater than the lower limit threshold and less than the upper limit threshold. The data existence probability Pa in the other part is L because it is smaller than the lower limit threshold. As a result, the detection flag generation unit 348 generates a detection flag having the pattern shown in FIG.

  The detection flag generation unit 348 compares the data existence probability Pa with respect to threshold values other than the threshold value set F for passenger cars, but the threshold values other than F are as shown in FIG. No detection flag pattern is generated. This is because the value of the data existence probability Pa in the line direction differs between a passenger car and a truck, and between a passenger car and a light vehicle. Therefore, the detection flag pattern as shown in FIG. 13C in the threshold set F is generated only for the passenger car threshold set F for a passenger car.

  When the detection flag generation unit 348 generates a plurality of detection flags by comparing the data existence probability Pa and a set of threshold values corresponding to the target object, the detection flag generation unit 348 associates the generated detection flag pattern with the corresponding target object information. To the recognition processing unit 343. For example, when the above threshold value sets D, E, and F are compared with the data existence probability P shown in FIG. 13B, the detection flag pattern generated by comparing with D is compared with the information “track” and E. The generated detection flag pattern and “light vehicle”, and the detection flag pattern (corresponding to FIG. 13C) generated by comparing with F and the information “passenger car” are output to the recognition processing unit 343, respectively.

  The recognition processing unit 343 recognizes an object based on the detection flag pattern output from the detection flag generation unit 348 and information on the target corresponding to the pattern. That is, the recognition processing unit 343 reads a plurality of input detection flag patterns, and recognizes an object from information about a set of threshold values corresponding to a desired detection flag pattern. For example, when the detection flag pattern generated by comparison with the threshold sets D, E, and F and the corresponding target object information are input, the recognition processing unit 343 selects the object shown in FIG. It is recognized as a passenger car corresponding to the detection flag pattern.

  The operation of the third embodiment of the present invention will be described with reference to FIG.

  Since the processing from steps S001 to S006 is the same as steps S1 to S6 in FIG. 7 described in the first embodiment, the description thereof is omitted.

  The speed comparison unit 340 compares the detection speed V detected by the speed detection unit 33 with the speed threshold Vth (step S007), and when the detection speed V is faster than the set speed Vb (step S008-Yes), the image size The detection speed V is output to the changing unit 341, and then the processing from steps S009 to S011 is performed. Since the processing from these steps S009 to S011 is the same as the steps S7 to S9 in FIG. 7 described in the first embodiment, the description thereof is omitted.

  When the detected speed V detected by the speed detecting unit 33 is slower than the set speed Vb, or when an error signal is input from the speed detecting unit 33 (step S008-No), the speed comparing unit 340 performs steps S009 to S011. The process is switched so that the processes from steps S012 to S017 are performed. That is, the speed comparison unit 340 does not output the detection speed V to the image size changing unit 341 but switches the processing by outputting an instruction signal to the averaging processing unit 344.

  When the instruction signal is output from the speed comparison unit 340, the averaging processing unit 344 performs a moving average process on the imaging data acquired from the imaging data storage unit 31 to remove noise (step S012).

  The histogram calculation area setting unit 345 narrows down the first area (or the second area) in the imaging data on which the moving average processing has been performed, and obtains a histogram calculation area (step S013).

  The altitude histogram calculation unit 346 divides the histogram calculation area into N in the height direction to create a total of N height bins Bha and calculates the altitude histogram Ca (step S014). The altitude histogram calculation unit 346 calculates this altitude histogram along the line direction of the vehicle.

  Based on the altitude histogram calculated along the line direction of the vehicle, the data existence probability calculation unit 347 includes a bin number n in which a predetermined number or more of intensity data or distance data is stored, and a bin total number N divided in the height direction. From the ratio n / N, the data existence probability Pa along the line direction of the vehicle is calculated (step S015).

  The detection flag generation unit 348 generates and generates a plurality of detection flags by comparing the data existence probability Pa calculated by the data existence probability calculation unit 347 and a set of thresholds having different values depending on the object. The plurality of detected flag patterns and the information on the object corresponding to each pattern are output to the recognition processing unit 343 (step S016).

  The recognition processing unit 343 is a target corresponding to a desired detection flag pattern from the detection flag pattern output from the detection flag generation unit 348 and information on the target corresponding to the pattern. (Step S017).

  Here, as described above, the histogram calculation area setting unit 345 may set an area including the entire vehicle as the histogram calculation area. However, as shown in FIGS. You may set to the area | region of the extent to which the tire of this is included. In this case, for example, the histogram calculation region setting unit 345 sells the upper limit value H1 in the altitude direction as the minimum ground altitude of the domestic vehicle, the lower limit value H2 as the standard deviation value of the altitude relative to the road surface, and the upper limit value S1 in the depth distance direction. The maximum value and the lower limit value S2 of the tire width are set as the minimum distance between the imaging device 1 and the object.

  The altitude histogram calculation unit 346 divides the histogram calculation area into N in the height direction to create a total of N height bins Bh, and calculates an altitude histogram. As shown in FIG. 15 (a), when there is a tire in the histogram calculation region, the histogram C in which intensity data or distance data exists is present at any position of the set height bin Bh. The probability Pa is 100%.

  On the other hand, as shown in FIG. 15 (b), when there is no tire in the histogram calculation region, since intensity data or distance data exists only for a part of the set height bin Bh, a biased histogram C is obtained. The data existence probability Pa is about 25%. As described above, the altitude histogram calculation unit 346 calculates the altitude histogram along the line direction of the vehicle.

  When there is a tire in the histogram calculation area, the data existence probability calculation unit 347 acquires the data existence probability P along the vehicle line direction shown in FIG. 16A (corresponding to FIG. 16B). The data existence probability P obtained here has a higher value when a vehicle tire is detected.

  The detection flag generation unit 348 generates a detection flag by comparing the data existence probability P with a preset threshold value, and outputs the generated detection flag pattern to the recognition processing unit 343.

  In the data existence probability P shown in FIG. 16 (b), the detection flag does not exceed the threshold value in the portion where the tire does not exist and becomes 0, and exceeds 1 in the portion where the tire exists. Therefore, the detection flag generation unit 348 generates one detection flag pattern indicating the presence of the tire as shown in FIG. 16C when the tire is present, and detects zero when the tire is not present. A pattern will be generated.

  The recognition processing unit 343 determines from the detection flag pattern output from the detection flag generation unit 348 whether a tire exists on the object, in other words, whether an axle exists on the object. When the recognition processing unit 343 determines that the axle is present on the object, the recognition processing unit 343 recognizes that the object is a vehicle.

  When the rear wheel tire is also included in the image of the vehicle included in the acquired imaging data, the detection flag pattern is a pattern in which two 1 peaks occur, and only the front wheel tire is detected. Similarly, it is determined that an axle is present in the object, and the recognition processing unit 343 recognizes that the object is a vehicle.

  As described above, according to the third embodiment of the present invention, in addition to the effect of the first embodiment, when the detection speed of the object is lower than the speed threshold or when the speed of the object passing cannot be detected. Then, an altitude histogram is calculated for a histogram calculation area obtained by narrowing down the area in the imaging data, and a detection flag is generated by comparing the data existence probability calculated from the altitude histogram with a threshold, and from the detection flag pattern Since the object is recognized, the object can be recognized even when the image including all the shapes of the object cannot be saved.

  The object recognition unit 347 may include a data deletion unit (not shown in FIG. 11). The data deletion unit inspects the data amount stored in the imaging data storage unit 31. Assuming that the data amount threshold value is M, for example, if the installed memory amount is Me (bit), the number of pixels in one line N and the number of bits in data F are used to calculate the equation (5). It is required as follows.

  In addition, the data deletion unit may obtain the following equation (6) using the time limit TL that requires object recognition. However, it is not limited to Formula (5) or Formula (6). When the stored data amount is M or more, the data deletion unit 347 discards the oldest data and outputs the result to the imaging data storage unit 31 as an electrical signal.

  Here, details of how to obtain the speed threshold value Vth will be described. The speed threshold Vth is expressed as in Expression (7) using M in Expression (5) or Expression (6). The user may arbitrarily define M.

Embodiment 4 FIG.
Embodiment 4 of the present invention will be described below with reference to the drawings. FIG. 17 is a configuration diagram of an object recognition unit according to the fourth embodiment. Portions corresponding to the configuration of the image display system according to the first embodiment are denoted by the same reference numerals as those in FIG.

  Compared with the object recognition system of the third embodiment, the object recognition system of the fourth embodiment newly includes an averaging processing unit 344a, a histogram calculation region setting unit 345a, and an altitude histogram calculation as shown in FIG. The configuration may include a unit 346a, a data existence probability calculation unit 347a, and a detection flag generation unit 348a. Each of these components has the same functions as the averaging processing unit 344, histogram calculation region setting unit 345, altitude histogram calculation unit 346, data existence probability calculation unit 347, and detection flag generation unit 348 of FIG. To do. Further, since the operation until the detection speed is input to the speed comparison unit 340 is the same as that of the third embodiment, the description thereof is omitted.

  The speed comparison unit 340 outputs an instruction signal to the averaging processing unit 344 when the detected speed V detected by the speed detection unit 33 is slower than the speed threshold Vth or when an error signal is input from the speed detection unit 33. To do. Thereafter, the processing of steps S012 to S017 in FIG. 14 is performed to recognize the object.

  When the detection speed V detected by the speed detection unit 33 is faster than the speed threshold Vth, the speed comparison unit 340 outputs the detection speed V to the image size changing unit 341 and the averaging processing unit 344a. Thereafter, the image size changing unit 341 and the determining unit 342 perform the processing of steps S009 and S010 in FIG. 14, and the averaging processing unit 344a, the histogram calculation region setting unit 345a, the altitude histogram calculation unit 346a, Processing corresponding to steps S012 to S016 in FIG. 14 is also performed by the data existence probability calculation unit 347a and the detection flag generation unit 348a.

  Then, the recognition processing unit 343 determines the object recognition result based on the determination result output from the determination unit 342 and the detection flag pattern and target object information output from the detection flag generation unit 348a. Output to. In other words, the recognition processing unit 343 detects the pattern of the detection flag generated from the result of comparison between the captured image data whose image size has been changed and the template data corresponding to the reference speed, and the histogram calculated for the altitude direction of the captured image data. And the object or part of the object is recognized from the information of the object. When the result from the determination unit 342 is different from the result from the detection flag generation unit 348a, the result from the determination unit 342 is preferably prioritized.

  By doing so, it is possible to improve the accuracy of object recognition in the recognition processing unit 343 as compared with the first to third embodiments.

  Note that the averaging processing unit 344a may obtain the moving average score NA using the offset C as shown in Expression (8). It is possible to perform an optimal averaging process by dynamically controlling the moving average score.

  A plurality of speed threshold values Vth may be set. For example, assuming that Vth and Vth_s (Vth <Vth_s), if the detected speed V is larger than Vth_s in the speed comparison unit 340, processing corresponding to steps S009 to S011 in FIG. If the detected speed V is lower than Vth, processing corresponding to steps S012 to 017 in FIG. 14 is performed. If the detection speed V is larger than Vth and smaller than Vth_s, as described in the present embodiment, the image size changing unit 341 and the determining unit 342 perform the processes of steps S009 and S010 in FIG. Processing corresponding to steps S012 to S016 of FIG. 14 is performed by the averaging processing unit 344a, the histogram calculation region setting unit 345a, the altitude histogram calculation unit 346a, the data existence probability calculation unit 347a, and the detection flag generation unit 348a. The recognition processing unit 343 recognizes the object based on the determination result output from the determination unit 342 and the detection flag pattern and target object information output from the detection flag generation unit 348a. Good. Further, Vth_s may be set arbitrarily.

DESCRIPTION OF SYMBOLS 1 1st imaging device 2 2nd imaging device 3 Object recognition apparatus 31 Imaging data storage part 32 Object detection part 33 Speed detection part 34 Object recognition part 340 Speed comparison part 341 Image size change part 342 Determination part 343 Recognition processing part 344 344a Averaging processing unit 345, 345a Histogram calculation area setting unit 346, 346a Altitude histogram calculation unit 347, 347a Data existence probability calculation unit 348, 348a Detection flag generation unit 4 Speedometer

Claims (8)

Imaging means for outputting imaging data obtained by irradiating laser light to a predetermined region through which an object passes;
An object speed detecting means for detecting a speed at which the object passes;
The size of the image of the object included in the imaging data output from the imaging means is changed according to the detection speed, and the imaging data after the change is compared with template data corresponding to a reference speed. Or an object recognition means for recognizing a part of the object ,
The object recognition means includes
A speed comparison unit that compares the detected speed with a preset speed threshold;
When the detection speed is faster than the speed threshold, the image size of the object included in the imaging data is changed according to the detection speed, and the changed imaging data and template data corresponding to the reference speed are obtained. Recognize the object or part of the object in comparison,
When the detection speed is slower than the speed threshold or when the speed of the object is not detected, a detection flag is generated from the histogram calculated for the altitude direction of the imaging data, and the object or the An object recognition system characterized by recognizing a part of an object. The object recognition means includes
A size changing unit that changes the size by changing the length of the image of the object or a part of the object based on the ratio of the detection speed to the reference speed; and
A pattern matching process is performed between the imaging data having the image changed by the size changing unit and the template data having the feature amount of the object, and it is determined whether the imaging data has the feature amount of the object. A determination unit to perform,
And a recognition processing unit that recognizes the object or a part of the object as the target when the determination unit determines that the imaging data has a feature amount of the target. Item 4. The object recognition system according to Item 1. The imaging means includes
A first imaging unit that outputs first imaging data acquired by irradiating a first region through which the object passes, and a predetermined interval in the passing direction of the object with respect to the first imaging unit; And a second imaging unit that outputs second imaging data acquired by irradiating a laser beam to a second region through which the object passes, and
The object speed detecting means is
A first detection signal indicating that the object in the first area is detected from the first imaging data is output, and the object in the second area is detected from the second imaging data. An object detection unit that outputs a second detection signal indicating that;
A speed detection unit that detects a speed at which the object passes from a time difference between inputs of the first detection signal and the second detection signal;
The object recognizing means recognizes the object or a part of the object by changing a size of an image of the object included in either the first imaging data or the second imaging data. The object recognition system according to claim 1 or 2. The object recognition means includes
An altitude histogram calculation unit that calculates a histogram for the altitude direction of the imaging data when the detection speed is slower than the speed threshold or when the speed of the object is not detected;
A data existence probability calculating unit that calculates a ratio of the imaging data of a predetermined number or more in the altitude histogram calculated by the altitude histogram calculating unit as a data existence probability;
A detection flag generation unit that generates the detection flag by comparing the data presence probability calculated by the data presence probability calculation unit with a preset threshold;
The said recognition process part recognizes the said object or a part of said object from the pattern of the said detection flag produced | generated by the said detection flag production | generation part, The any one of Claim 1 to 3 characterized by the above-mentioned. Object recognition system. The altitude histogram calculation unit creates N bins by dividing the imaging data in an altitude direction,
The data existence probability calculating unit calculates, as the data existence probability, a ratio n / N of the number of bins n in which more than a predetermined number of the imaging data is stored with respect to the number N of bins. 5. The object recognition system according to 4 . The object recognition means includes
When the detection speed is faster than the speed threshold, a comparison is made between the imaging data whose image size is changed according to the detection speed and the template data corresponding to the reference speed, and the altitude direction of the imaging data is calculated. the object or the object object recognition system of the placing serial any to one of claims 1 to 5, characterized in that to recognize a portion of the based on the the pattern of the detecting flag generated from the histogram . The object recognition system according to any one of claims 1 to 6, wherein the object is a vehicle, and a part of the object is an axle. An object recognition device for recognizing an object or a part of an object by inputting imaging data obtained by irradiating a predetermined area through which the object passes with laser light
An object speed detecting means for detecting a speed at which the object passes;
The size of the image of the object included in the imaging data output from the imaging means is changed according to the detection speed, and the imaging data after the change is compared with template data corresponding to a reference speed. Or an object recognition means for recognizing a part of the object ,
The object recognition means includes
A speed comparison unit that compares the detected speed with a preset speed threshold;
When the detection speed is faster than the speed threshold, the image size of the object included in the imaging data is changed according to the detection speed, and the changed imaging data and template data corresponding to the reference speed are obtained. Recognize the object or part of the object in comparison,
When the detection speed is slower than the speed threshold or when the speed of the object is not detected, a detection flag is generated from the histogram calculated for the altitude direction of the imaging data, and the object or the An object recognition apparatus characterized by recognizing a part of an object.

Images