JP3716623B2 - Thermal detector - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a warm body detection device capable of accurately detecting a warm body existing in the traveling direction of a vehicle.
[0002]
[Prior art]
As a conventional warm body detection device, an “image type pedestrian detection device” described in Japanese Patent Laid-Open No. 7-200906 shown in FIG. 14 has been reported.
[0003]
In a conventional warm body detection device, a visible image is captured using a CCD camera 141 during the daytime, while a thermal image is captured using a thermal image sensor camera 143 at night, and input using a video signal switch 145. By switching the image and inputting it to the image processing unit 147, and binarizing the input image, by taking the difference between the binarized data of the latest frame image and the binarized data of the previous frame image, Classification is made into four states: the presence of an object in the shooting area, the absence of an object, the retreat of an object, and the entry of an object. Further, the movement vector is calculated by paying attention only to the area within the imaging area where the object intrusion, the object exists, and the object retreats. A movement vector whose direction is parallel to the pedestrian crossing is regarded as a pedestrian, while a movement vector whose direction is vertical is regarded as a vehicle, so that the pedestrian and the vehicle can be distinguished effectively. As a result, when a pedestrian is detected, the output unit 149 alerts the driver vehicle to the presence of the pedestrian.
[0004]
[Problems to be solved by the invention]
In the conventional warm body detection device, the camera is installed in a fixed place, and the pedestrian detection function is exhibited only when the processed image is a still image.
However, when a conventional warm body detection device is mounted on a vehicle, since all objects in the shooting area are detected as moving objects, it is difficult to distinguish and detect only pedestrians. It becomes impossible to classify into four states, that is, an object, no object, object retreat, and object intrusion.
[0005]
Also, when the camera is mounted on the vehicle and the vehicle is traveling on the road, the moving speed of the vehicle is relatively large compared to the moving speed vector of the pedestrian in the transverse direction. It is difficult to distinguish using the vector direction.
Furthermore, when a road ahead of the vehicle is photographed with an infrared camera, the high-luminance input object includes the headlights, radiators, and mufflers of the preceding vehicle in the opposite lane as well as the face of the pedestrian. , Body, taillights, signal lights, road lighting, etc. These disturbances and the pedestrian's face are similar in both shape and size, so it is difficult to distinguish both . For this reason, even if it tried to detect a pedestrian as a warm body, there were many malfunctions, and there existed a problem that reliability would fall remarkably even if a warm body detection apparatus was utilized for an alarm system.
[0006]
The present invention has been made in view of the above, and an object of the present invention is to provide a warm body detection device capable of accurately detecting a pedestrian without being affected by disturbance factors.
[0007]
In order to solve the above-mentioned problem, the invention described in claim 1 is a thermal image capturing unit that captures a thermal image having a heat source in the traveling direction of the vehicle, and a visible image capturing unit that captures a visible image in the traveling direction of the vehicle. Vehicle detection means for detecting the distance and direction to the other vehicle in the traveling direction of the vehicle, vehicle extraction means for extracting the presence area of the other vehicle based on the distance and direction to the other vehicle, and the visible image An object that emits light is extracted from the visible image based on the luminance of the visible image captured by the image capturing unit, and is extracted by the vehicle extracting unit and the light emitting object from the warm body image captured by the warm body image capturing unit. The present invention includes an image removal processing means for removing the existing region of the other vehicle and a warm body detection means for detecting the presence of a pedestrian based on the warm body image obtained as a result of the removal processing. To do.
[0008]
According to a second aspect of the present invention, in order to solve the above-mentioned problem, the vehicle extracting means is configured to detect a point group in which a horizontal distance between a plurality of point groups represented by a distance and a direction to the other vehicle is a predetermined value or less. Group generation means for grouping, group selection means for selecting a group whose horizontal width exceeds a predetermined value as a vehicle group, area expansion means for extending the area of the selected vehicle group in the height direction, and this expansion The gist of the invention is that it includes presence region generation means for generating a vehicle presence region by setting a predetermined luminance value in the vehicle group region after processing.
[0009]
According to a third aspect of the present invention, in order to solve the above-described problem, the warm body detection unit includes an area determination unit that determines a pedestrian when an area of a pixel group of a warm body image exceeds a predetermined reference value, and a warm body Aspect ratio determining means for determining a pedestrian when the aspect ratio of a rectangle circumscribing a pixel group of an image is within a predetermined range, and a high luminance pixel rate per unit area of a pixel group of a thermal image is within a predetermined range it is determined that the pedestrian and the high luminance pixel rate determining means, when the area determining means and the aspect ratio determination means and determination result of the high luminance pixel ratio determining means are all pedestrian determines walking as a pedestrian The gist is provided with a person determination means.
In order to solve the above-mentioned problem, the gist of the present invention is that the vehicle detection means uses any one of a laser radar, a millimeter wave radar, and a stereo camera.
[0010]
【The invention's effect】
According to the first aspect of the present invention, a warm body image and a visible image in the traveling direction of the vehicle are captured, and the distance and direction to the other vehicle in the traveling direction of the vehicle are detected. Based on the distance and the direction, the existence area of the other vehicle is extracted. Next, the presence area of the visible image and the other vehicle is removed from the warm image, and the presence of the pedestrian is detected based on the warm image obtained as a result of the removal process. It is possible to accurately detect a pedestrian without being influenced by.
[0011]
According to the second aspect of the present invention, a group of point groups in which the horizontal distance between a plurality of point groups represented by the distance and direction to another vehicle is a predetermined value or less is grouped, and the horizontal width exceeds a predetermined value. As a vehicle group. Next, the area area of the selected vehicle group is expanded in the height direction, a predetermined luminance value is set in the area of the vehicle group after the expansion process, and a vehicle existence area is generated. It is possible to accurately extract the existence area of other vehicles in the traveling direction, and as a result, it is possible to specify the location where the disturbance factor occurs.
[0012]
According to the third aspect of the present invention, when the area of the pixel group of the warm body image exceeds a predetermined reference value, it is determined that the person is a pedestrian, and then a rectangular vertical and horizontal circumscribing the pixel group of the warm body image. If the ratio is within a predetermined range, it is determined that the person is a pedestrian. Furthermore, when the high luminance pixel rate per unit area of the pixel group of the thermal image is within a predetermined range, it is determined as a pedestrian, and when all of these determination results are pedestrians, it is determined as a pedestrian. By doing so, it is possible to contribute to improving the accuracy of pedestrians.
[0013]
According to the present invention as set forth in claim 4, the distance and direction to the other vehicle in the traveling direction of the vehicle can be detected by using any one of a laser radar, a millimeter wave radar, and a stereo camera. Then, the location where the other vehicle is present can be extracted to identify the location where the disturbance factor occurs.
[0014]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram illustrating an overall configuration of a warm body detection apparatus according to an embodiment of the present invention.
[0015]
As shown in FIG. 1, the warm body detection device has a camera unit 13 disposed in the back of a radiator grill 11 in front of the vehicle 9. The camera unit 13 detects the infrared camera 1 that captures a warm body image in front of the vehicle, the visible camera 3 that captures a visible image in front of the vehicle, and the distance and direction of the vehicle in the traveling direction of the vehicle as existing regions. A vehicle detection unit 17 composed of a laser radar is disposed, and each output is transmitted via the harness 15 to the image processing unit 5 disposed in the vehicle interior.
[0016]
When the image processing unit 5 detects the presence of a pedestrian, it sends a detection signal to the alarm unit 7 to generate an alarm using sound, LED, video, etc., to alert the driver to the pedestrian. To do.
[0017]
Although the camera unit 13 is described as being disposed in the radiator grille 11 portion, it may be disposed in the vehicle headlight, bumper portion, windshield portion, roof portion, and the like. The area may be the rear of the vehicle in addition to the front of the vehicle.
[0018]
In addition, the horizontal and vertical field angles of the infrared camera 1 and the horizontal and vertical field angles of the visible camera 3 are roughly matched in advance optically or by enlargement / reduction processing. Furthermore, it is assumed that the horizontal angle of view of the infrared camera 1 and the vehicle detection unit 17 are matched in advance by the above processing.
[0019]
Next, FIG. 2 is a block diagram of the warm body detection apparatus.
The warm body image is captured by the infrared camera 1 and then converted to digital image data having a predetermined two-dimensional luminance by the A / D converter 21. Further, after the window comparator 25 selects only pixels having a predetermined luminance, the binarization unit 29 binarizes the selected pixels.
[0020]
On the other hand, the visible image is captured by the visible camera 3 and then converted into digital image data having a two-dimensional predetermined number of pixels by the A / D converter 23. Further, after the pixel having only a predetermined luminance is selected by the window comparator 27, the pixel is binarized by the binarization unit 31.
[0021]
Here, in the difference unit 33, the visible image output from the binarization unit 31 and the presence vehicle image of the preceding vehicle output from the vehicle detection unit 17 are subtracted from the warm body image output from the binarization unit 29. Is done. The pedestrian identification unit 35 identifies whether or not there is a pedestrian in the image obtained from the subtraction result of the difference unit 33. If it is determined that there is a pedestrian, the pedestrian identification unit 35 sends an alarm signal to the alarm unit 7. Output. As a result, the driver is alerted from the alarm unit 7 that a pedestrian is present.
The image processing unit 5 is composed of electronic components such as a CPU, ROM, and RAM, and has the functional blocks as described above.
[0022]
Next, the operation of the image processing unit 5 shown in FIG. 3 will be described with reference to FIGS. The image processing unit 5 is controlled according to a control program stored in the ROM.
As shown in FIG. 4, a traffic light 51 is installed in front of the host lane 41 in front of the host vehicle 41, and a preceding vehicle 43 is traveling on the host lane 41. An oncoming vehicle 47 is traveling on the oncoming lane 45 with the headlights on. A pedestrian 49 is walking on the sidewalk 60.
[0023]
First, in step SlO, the warm body image from the infrared camera 1 is A / D converted by the A / D converter 21 and converted into two-dimensional luminance data having a predetermined number of pixels.
In step S20, the window comparator 25 selects only pixels having a luminance value with a predetermined width. For example, if the luminance value of a pedestrian's face is usually about 200 out of luminance data with a dynamic range of 256 gradations, only the luminance data within the luminance range 190 to 210 is selected, and the luminance values of the other pixels are all 0. Rewrite to As another method, if an appearance histogram of luminance values of all pixels is measured and generated, and the luminance of a predetermined width is adjusted so that, for example, only 100 pixels are extracted as the predetermined number of pixels, automatic It is also possible to extract only the luminance data of the pedestrian's face.
[0024]
Next, in step S30, all of the pixels obtained in step S20 are rewritten to the luminance value 255 for pixels whose luminance data is not 0. Here, the warm body image which passed through the process of step S30 is shown in FIG. Since it is a warm body image picked up by the infrared camera 1, only the warm body image having the heat source has a high luminance value 255, and the portion painted in black shown in FIG. 5 corresponds to the pixel of the high luminance value 255. As shown in FIG. 5, an infrared image 511 of a blue signal, an infrared image 521 of a headlight of an oncoming vehicle, a pedestrian's head 491, a pedestrian's right hand 495, a pedestrian's left hand 493, and a taillight of a preceding vehicle An infrared image 531 and an infrared image 54 of the muffler of the preceding vehicle are extracted.
[0025]
In step S40, the visible image captured by the visible camera 3 is A / D converted by the A / D converter 23 and converted into two-dimensional luminance data having a predetermined number of pixels.
In step S50, the window comparator 27 selects only pixels having a luminance value with a predetermined width.
In step S60, all of the pixels obtained in step S50 are rewritten to the luminance value 255 for pixels whose luminance data is not 0. Here, the visible image which passed through the process of step S60 is shown in FIG. Since it is a visible image captured by the visible camera 3, only the light emitting object has a high luminance value 255, and the portion painted black in FIG. 6 corresponds to the pixel of the high luminance value 255. As shown in FIG. 6, a visible image 513 of a green signal, a visible image 523 of a headlight of an oncoming vehicle, and a visible image 533 of a taillight of a preceding vehicle are extracted. At this time, in the visible image, lights such as the visible image 513 of the green light, the visible image 523 of the headlight of the oncoming vehicle, and the visible image 533 of the taillight of the preceding vehicle are scattered by dust in the air, so-called glare. Since the phenomenon occurs, the area selected from the lamp imaged by the infrared camera 1 is enlarged.
[0026]
In step S70, the vehicle detection unit 17 detects the distance and direction to the object in front of the vehicle and captures the values. As shown in FIG. 7, the distance point groups detected by the vehicle detection unit 17 are the license plate of the traffic light 51, the taillight of the preceding vehicle and the trunk portion 53 therebetween, the headlight of the oncoming vehicle and the portion 52 therebetween. Part. The distance D between these point groups is D3 for the column of the traffic light 51, D1 for the taillight of the preceding vehicle and the trunk portion 53 between them, and D2 for the headlight of the oncoming vehicle and the portion 52 between them. It is measured.
[0027]
In step S80, the horizontal distance is less than or equal to a predetermined value between the point groups calculated in step S70 using the trigonometric function calculation based on the difference in reach distance from the point group to the target point group and the distance and direction. Group point cloud comrades. For example, if the horizontal distance is set within 0.1 m and the distance to the point group is set within 0.1 m as the predetermined values, the resolution sufficient for performing the vehicle extraction process in the next step S90 is obtained. can get.
[0028]
In step S90, only the group whose group horizontal width is larger than a predetermined value is selected from the groups obtained in step S80. For example, if only a group having a horizontal width of 1.5 m or more is selected as a vehicle group among the groups, at least the vehicle can be reliably extracted.
[0029]
In step S100, the area of the vehicle group selected in step S90 is expanded to about 6 times only in the height direction using the area center of gravity of the vehicle group calculated in advance as an origin, and the areas A and B shown in FIG. Thus, the brightness value is set to 255, and the brightness values of the other areas are set to 0. Here, the reason why the area is expanded six times in the height direction is that a muffler may be arranged in a portion protruding from the vehicle group of a general vehicle depending on the vehicle type. This is because it is included in the vehicle group.
[0030]
In step SlO, the visible image obtained in step S60 is subtracted for each pixel from the warm body image obtained in step S30. When the subtraction result is negative, the subtraction result is rewritten to 0. As a result, as shown in FIG. 8, the pedestrian's head 491, the pedestrian's right hand 495, the pedestrian's left hand 493, and the muffler 32 of the preceding vehicle are extracted, and an infrared image 511 of a green signal, the head of the oncoming vehicle The infrared image 521 of the light and the infrared image 531 of the tail light of the preceding vehicle are removed. Further, the existence area of the preceding vehicle obtained in step S100 is subtracted for each pixel from the removal result, and is rewritten to 0 when the subtraction result is negative. As a result, as shown in FIG. 9, the pedestrian's head 491, the pedestrian's right hand 495, and the pedestrian's left hand 493 are extracted, and the infrared image of the muffler 54 of the preceding vehicle is removed.
[0031]
In step S120, continuous regions of the pixel group extracted in step S110 are sequentially labeled as a cluster.
In step S130, the area of each labeled pixel group is calculated.
In step S140, the center of gravity of each labeled pixel group is calculated, and the position of the center of gravity is further calculated.
In step S150, as shown in FIG. 10, five areas from area A to E are set in advance in the vertical direction of the paper, and each area is labeled from the barycentric position of each pixel group calculated in step S140. It is determined which of the areas A to E the pixel group belongs to.
[0032]
In step S160, the size of the area of each labeled pixel group calculated in step S130 is compared with the size of the area of the pedestrian's face for which a reference value is set in advance for each region. If it is determined that the face is a face, the size flag indicating the size match is set to 1, while if it is determined not, the size flag is set to 0. As a result, the hand of the pedestrian is excluded because it is small relative to the pedestrian's face. Therefore, the pixel group in which the size flag is set to 1 is only the pedestrian's head 491. By setting a standard value for the size of the pedestrian's face for each area, even if the apparent size changes depending on the distance from the vehicle, the size of the pedestrian's face can be compared. Can be done accurately.
[0033]
In step S170, the coordinates of the circumscribed rectangle are calculated for the head 491 of the pedestrian whose size flag is set to 1 in step S160.
In step S180, as shown in FIG. 11, the aspect ratio of the circumscribed rectangle 497 of the head of the pedestrian whose size flag is set to 1, that is, the ratio of the landscape A and the portrait B of the circumscribed rectangle 497 is set. calculate.
[0034]
In step S190, it is determined whether or not the ratio of the horizontally long A and the vertically long B of the circumscribed rectangle 497 of the pedestrian's head is within a predetermined range set in advance. If it is within the predetermined range, the aspect ratio flag indicating the coincidence of the aspect ratio is set to 1, and if not, 0 is set. In general, since the face of a pedestrian is close to a circle, a value with an aspect ratio close to 1 is a threshold value within a predetermined range.
[0035]
In step S200, the area is calculated according to the coordinates of the circumscribed rectangle obtained in step S170. Furthermore, by calculating the ratio with the area of the pedestrian's head 491 obtained in step S130, the ratio of high luminance pixels per unit area relating to the target pixel group, that is, the sufficiency rate is calculated.
[0036]
In step S210, it is determined whether or not the sufficiency rate calculated in step S200 is within a predetermined range set in advance. If it is within the predetermined range, a fullness flag indicating that the fullness is the same is set to 1, and if not, 0 is set. As a result, only a pixel group with a high pixel density, such as a pedestrian's face, is selected, while a pixel group with a low pixel density is excluded, so that the accuracy of pedestrian detection can be improved.
[0037]
As a result of the above processing, even if there is a heat source of disturbance, it is possible to eliminate them depending on the size, shape, and fullness. However, the infrared image of the muffler 54 of the preceding vehicle is substantially the same in size, shape, and fullness as the pedestrian's head 491, and thus cannot be excluded in steps S160, S190, and S210, but in advance in step S110. Therefore, the infrared image of the muffler 54 of the preceding vehicle is not mistaken for the head 491 of the pedestrian.
[0038]
In step S220, if there is a pixel group in which the size flag set in step S160, the aspect ratio flag set in step S190, and the fullness flag set in step S210 are all 1, a pedestrian The alarm signal is sent to the alarm unit 7. As a result, the driver is alerted from the alarm unit 7 that a pedestrian is present. The alarm unit 7 includes a buzzer, an indicator, a monitor, and the like. An alarm sound is generated from the buzzer, and red alarm light is generated from the indicator. Further, a pedestrian warm body image extracted by the image processing unit 5 is displayed from the monitor. As a result, when the driver notices the warning sound and looks at the monitor, the pedestrian's warm body image is displayed, so that the pedestrian's presence area can also be visually recognized.
[0039]
In the present embodiment, the warm body image and the visible image captured by the infrared camera 1 and the visible camera 3 are simply subtracted by the difference unit 33 to perform difference processing. It is considered that an afterimage phenomenon called a smear phenomenon occurs in the images from 1 and the visible camera 3. Since this smear phenomenon occurs when a strong luminance signal is input to the camera, it is effective to reduce the aperture of the camera until this phenomenon is resolved. The smear detection may be performed by detecting a change in the luminance value of a predetermined pixel until the strong luminance signal is eliminated. It is also possible to adjust the time delay of the video input to the difference unit 33 according to the time when the smear phenomenon occurs. It is also possible to automatically adjust the aperture of the camera according to the time when the smear phenomenon occurs.
[0040]
Further, in the present embodiment, a laser radar is used for the vehicle detection unit 17, but it can be replaced with a millimeter wave radar instead. When a millimeter wave radar is used for the vehicle detection unit 17, as shown in FIG. 12, the detected distance point group is a metal plate that favorably reflects the millimeter wave. This metal plate is, for example, the column of the traffic light 51, the entire preceding vehicle 43, and the entire oncoming vehicle 47. The distance D of these point groups is measured as D3 for the column of the traffic light 51, Dl for the preceding vehicle 43, and D2 for the oncoming vehicle 47.
[0041]
Furthermore, the vehicle detection unit 17 can be replaced with a stereo camera method. This is a method in which two cameras arranged at a predetermined distance in the horizontal direction are used as the vehicle detection unit 17 and the distance to the object is measured based on the parallax between the two cameras. As shown in FIG. 13, the detected distance point group is the entire preceding vehicle 43 and the entire oncoming vehicle 47. Thus, in the stereo camera method, in order to reduce measurement noise, a small object or a thin object can be excluded in advance and a distance calculation can be performed, so that only a vehicle can be extracted. Further, an object arranged at a high position such as the signal device 51 can be easily excluded by setting the measurement angle of view. The distance D of these point groups is measured as D1 for the preceding vehicle 43 and D2 for the oncoming vehicle 47.
[0042]
Furthermore, the case where an ultrasonic method is used for the vehicle detection unit 17 or the case where optical triangulation is used for an image signal obtained by a one-line sensor composed of a CCD is conceivable.
[0043]
Further, the operating condition of the warm body detection device may be limited by judging the vehicle speed, the distance from the preceding vehicle detected by the vehicle detection unit 17, the relative vehicle speed with respect to the preceding vehicle, and the like.
Furthermore, the effect of the present invention is satisfied even if the thermal image captured by the infrared camera 1 and the visible image captured by the visible camera 3 do not completely match due to a shift in the optical axis or the like. Therefore, high cost for precise optical axis adjustment can be prevented.
[0044]
Furthermore, the lights are limited to traffic lights, headlights of oncoming vehicles, and taillights of preceding vehicles, but the same effect can be obtained for all illuminants such as road lights and house lights. Needless to say.
[Brief description of the drawings]
FIG. 1 is a diagram showing an overall configuration of a warm body detection apparatus according to an embodiment of the present invention.
FIG. 2 is a block diagram of a warm body detection apparatus.
FIG. 3 is a flowchart for explaining the operation of the image processing unit 5;
FIG. 4 is a diagram for explaining a use state of a warm body detection device.
FIG. 5 is a diagram illustrating a result of image processing performed on an image of a warm body image captured by an infrared camera by an image processing unit 5;
FIG. 6 is a diagram showing a result of image processing performed on a visible image captured by a visible camera by an image processing unit 5;
FIG. 7 is a diagram showing distances and directions detected by a laser radar.
FIG. 8 is a diagram showing a result of subtracting a visible image from a warm image in the image processing unit 5;
FIG. 9 is a diagram illustrating an image processing result in which only a pedestrian's warm body image is extracted by the image processing unit 5;
10 is a diagram showing processing areas A to E used for image processing in the image processing unit 5. FIG.
11 is a diagram showing a pedestrian's head used for image processing in the image processing unit 5. FIG.
FIG. 12 is a diagram showing a range of distance points detected by a millimeter wave radar.
FIG. 13 is a diagram showing a range of distance points detected by a stereo camera.
FIG. 14 is a block diagram of an image type pedestrian detection device reported as a conventional warm body detection device.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Infrared camera 3 Visible camera 5 Image processing unit 7 Alarm part 9 Vehicle 11 Radar grill 13 Camera unit 15 Harness 17 Vehicle detection part 41 Own lane 43 Leading car 45 Oncoming lane 47 Oncoming lane 49 Pedestrian 21, 23 A / D converter 25, 27 Window comparators 29, 31 Binarization unit 33 Difference unit 35 Pedestrian identification unit 51 Traffic light 52 Headlight of oncoming vehicle 53 Taillight of preceding vehicle 54 Infrared image of muffler of preceding vehicle 60 Sidewalk 511 Infrared of blue signal Image 513 Blue signal visible image 521 Oncoming vehicle headlight infrared image 523 Oncoming vehicle headlight visible image 491 Pedestrian head 493 Pedestrian left hand 495 Pedestrian right hand 497 Pedestrian head Circumscribed square 531 infrared image 533 of the taillight of the preceding vehicle visible image of the taillight of the preceding vehicle

Claims (4)

A warm body image capturing means for capturing a warm body image having a heat source in the traveling direction of the vehicle;
Visible image capturing means for capturing a visible image of the traveling direction of the vehicle;
Vehicle detection means for detecting the distance and direction to the other vehicle in the traveling direction of the vehicle;
Vehicle extraction means for extracting the presence area of the other vehicle based on the distance and direction to the other vehicle;
Based on the luminance of the visible image captured by the visible image capturing unit, an object that emits light is extracted from the visible image, and the object that emits light and the vehicle are extracted from the warm body image captured by the warm image capturing unit. Image removal processing means for removing the presence area of the other vehicle extracted by the means;
A warm body detection apparatus comprising warm body detection means for detecting the presence of a pedestrian based on a warm body image obtained as a result of the removal process. The vehicle extraction means includes
A group generating means for grouping point groups having a horizontal distance between a plurality of point groups represented by a distance and a direction to the other vehicle below a predetermined value;
Group selection means for selecting a group having a horizontal width exceeding a predetermined value as a vehicle group;
Area expanding means for expanding the area of the selected vehicle group in the height direction;
2. The warm body detection device according to claim 1, further comprising presence region generation means for generating a vehicle presence region by setting a predetermined luminance value in the vehicle group region after the expansion processing. The warm body detecting means includes
Area determining means for determining a pedestrian when the area of the pixel group of the warm body image exceeds a predetermined reference value;
Aspect ratio determining means for determining a pedestrian when the aspect ratio of a rectangle circumscribing a pixel group of a thermal image is within a predetermined range;
A high-luminance pixel rate determination means for determining a pedestrian when the high-luminance pixel rate per unit area of the pixel group of the warm body image is within a predetermined range;
A pedestrian determination unit for determining as a pedestrian when all of the determination results by the area determination unit, the aspect ratio determination unit, and the high luminance pixel rate determination unit are pedestrians. The warm body detection device according to 1. The vehicle detection means includes
The warm body detection apparatus according to claim 1, wherein any one of a laser radar, a millimeter wave radar, and a stereo camera is used.

Images