JP2019156276A - Vehicle detection method and vehicle detection device - Google Patents

Abstract

To provide a vehicle detection method and a vehicle detection device capable of narrowing a light spot in a photographed image, and correctly and quickly detecting a vehicle.SOLUTION: A vehicle detection device comprises: imaging means (camera) 2 for imaging a surrounding of a vehicle including other vehicle, and being mounted on a vehicle; and vehicle detection means (lamp ECU 6 (image analysis part 62)) for detecting the vehicle based on a light spot reflected on the photographed image. The vehicle detection means comprises: a light spot identification part 65 for identifying an object light spot which is a vehicle detection object, and a non-object light spot other than the object light spot, out of the plurality of light spots existing in the image; and a vehicle detection part 66 for detecting a vehicle based on a behavior of the object light spot in the image. The light spot identification part recognizes shapes of the plurality of light spots, and based on the recognition, identifies the object light spot and non-object light spots.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a vehicle detection method and a vehicle detection device that are mounted on a vehicle such as an automobile and detect other vehicles such as an oncoming vehicle and a preceding vehicle.

  ADB (Adaptive Driving Beam) that detects the other vehicle in the front area and controls the light distribution of the headlamp so as not to dazzle the other vehicle in order to improve the visibility of the front area of the host vehicle Control and AHB (Automatic High Beam) control are performed. In recent years, other vehicles have been detected for automatic driving control. As a technique for detecting such other vehicles, a technique has been proposed in which a surrounding area of the host vehicle is imaged by an imaging unit, and an image obtained by the imaging is analyzed to detect the other vehicle.

  In Patent Document 1, a light spot in an image obtained by an imaging unit is detected, and the attribute of the detected light spot, for example, the color and behavior (moving state) of the light spot, A technique has been proposed in which the headlamps and taillights of other vehicles are discriminated to detect other vehicles. Patent Document 1 also discloses a technique for shortening the detection time by setting a detection region for detecting another vehicle to a predetermined region based on the attribute of the detected light spot and determining the set region. Proposed.

JP 2012-20660 A

  The technique of Patent Document 1 is effective in shortening the detection time by limiting the detection area, but in the recent situation where the driving environment of automobiles is diversified, it is possible to detect other vehicles over a wide area. Because it is required, it is difficult to meet this requirement. At the same time, illuminants such as road signs and advertisements increase, and the number of light spots in the captured image tends to increase. Therefore, a process for discriminating other vehicles based on all the behaviors of these light spots is required, and the number of detection steps increases and the detection time increases.

  An object of the present invention is to provide a vehicle detection method and a vehicle detection device that can narrow down light spots in a captured image and detect a vehicle accurately and quickly.

  The vehicle detection method of the present invention is a detection method for detecting a vehicle from the behavior of a light spot that is mounted on a vehicle, images the surroundings of the vehicle, and is captured in an image obtained by the imaging. A step of recognizing the shapes of a plurality of existing light spots, a step of identifying a light spot that is a vehicle detection target among the recognized shapes as a target light spot, and a light spot other than that as a non-target light spot, and an image The method includes detecting a behavior of a target light spot in the vehicle and detecting a light spot having a predetermined behavior as a vehicle.

  A vehicle detection apparatus according to the present invention includes an imaging unit that is mounted on a vehicle and images the surroundings of the vehicle, and a detection unit that detects a vehicle from a light spot that is captured in an image captured by the imaging unit. The vehicle is detected from the target light spot that is the vehicle detection target from the plurality of light spots existing in the light source, the light spot identification unit that identifies the other non-target light spots, and the behavior of the identified target light spot in the image A light detection unit that recognizes shapes of the plurality of light spots, and identifies the target light spot and the non-target light spot based on the recognition;

  According to the present invention, the shape of the light spot in the image picked up by the image pickup means is identified, the light spot that is likely to be a vehicle is identified as the target light spot from this shape, and only the identified target light spot is identified. A process for detecting the vehicle is executed. As a result, even when there are many light spots in the image, there is no need to perform detection processing for light spots that are not likely to be light spots of the vehicle, and the number of processing steps for detecting the vehicle is reduced. In addition, rapid vehicle detection is possible.

1 is a schematic horizontal sectional view of a headlamp to which the present invention is applied. The block block diagram of lamp ECU containing a vehicle detection means. Schematic of an example of the front area | region of the own vehicle imaged with a camera. Schematic of the light spot obtained by imaging with a camera. The conceptual diagram for demonstrating the extraction area | region and shape recognition of a light spot. Schematic of target light spot. Schematic detected as a light spot of a vehicle. The conceptual diagram explaining the vehicle detection method when a camera exists in a high position. The conceptual diagram explaining the vehicle detection method when a camera exists in a low position. The conceptual diagram explaining the vehicle detection method when an image inclines. The conceptual diagram explaining the vehicle detection method using the light spot of a white line.

  Next, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a schematic horizontal sectional view of an embodiment in which a vehicle detection device according to the present invention is incorporated in headlamps HL provided on the left and right of a front part of a vehicle body 8 of an automobile. Although FIG. 1 shows the internal configuration of the right headlamp HL, the left headlamp has the same configuration. The lamp housing 3 of the headlamp HL includes a lamp body 31 and a translucent front cover 32, and the lamp unit 1 and the camera 2 are housed in the lamp housing 3. The lamp unit 1 and the camera 2 are supported by a base plate 5, and the respective optical axis directions can be adjusted by an aiming mechanism 4 including an aiming screw 41 and the like.

  The lamp unit 1 includes a light source 11 having a plurality of LEDs arranged in a simplified manner, and a projection lens 12 that projects light emitted from the light source 11 toward the front of the automobile, and controls light emission of the light source 11. Thus, it is possible to illuminate the front area of the automobile with a desired light distribution.

  The camera 2 is an image pickup means in the present invention, and although it does not appear in the drawing, it is constituted by a digital camera provided with an image pickup device. The camera 2 images a front area of the host vehicle, at least an area including an area irradiated with light from the lamp unit 1, and outputs an imaging signal. The objects imaged by the camera 2 include signs, signs, and vehicles such as preceding vehicles and oncoming vehicles that are present in the area where the host vehicle is traveling.

  A lamp ECU 6 is housed in the lamp housing 3 and is connected to the lamp unit 1 and the camera 2, respectively. The lamp ECU 6 can set a suitable light distribution based on an image obtained by imaging with the camera 2, and can perform light distribution of the lamp unit 1, for example, ADB control based on the setting. The lamp ECU 6 is connected to a vehicle ECU or the like not shown in the figure via a CAN (Controller Area Network) 100, and sends and receives required signals to and from them.

  As shown in the block diagram of FIG. 2, the lamp ECU 6 performs signal processing on an image signal output from the image sensor of the camera 2 to generate an image signal as image data, and this image. An image analysis unit 62 that analyzes a signal to form an image and detects a vehicle present in the image, and a lighting control unit that recognizes the detected vehicle and controls the light distribution of the lamp unit 1 63.

  In the lamp ECU 6, the signal processing unit 61 and the image analysis unit 62 perform an operation for detecting the vehicle, and thus these constitute vehicle detection means in a broad sense, but in particular, the vehicle is directly detected. In a sense, the image analysis unit 62 constitutes a vehicle detection means in a narrow sense. Therefore, the lamp ECU 6 detects the vehicle by the image analysis unit 62, that is, the vehicle detection means based on the image signal obtained by imaging with the camera 2, and controls the lighting so as to shield the area where the detected vehicle exists. The ADB control is executed by controlling the light distribution of the lamp unit 1 in the unit 63.

  As shown in FIG. 2, the image analysis unit 62 as a vehicle detection unit identifies a light spot extraction unit 64 that extracts a light spot existing in the image, and identifies the shape of the extracted light spot. A light spot identifying unit 65 that removes a light spot and a vehicle detection unit 66 that detects a vehicle by determining the behavior of the identified light spot, here, the movement state of the light spot in the image.

  A vehicle detection operation in the image analysis unit 62 as vehicle detection means will be described. Here, a description will be given of a case where the image of the front area obtained by imaging with the camera 2 is in the state of FIG. 3 when the automobile in which the ADB control of the headlamp HL is performed travels at night. In this image, the preceding vehicle CAR1 and the oncoming vehicles CAR2 and CAR3, whose lamps are lit, are traveling on a road on which a white line (including a yellow line) LINE is drawn. On the side of the road, lighting facilities such as a road lamp B and a delineator D are provided, and a rectangular road sign S1 and a rhombus road sign S2 are provided. Here, various signs are also included in the sign.

  An image signal (image data) is obtained by capturing an image of the front region with the camera 2 and processing the image signal with the signal processing unit 61. Further, the image shown in FIG. 4 is obtained from this image signal. This image is obtained by arranging bright and dark pixels (light dots) corresponding to the image sensor of the camera 2 in the X direction (row direction) and the Y direction (column direction). And in this image, the imaged vehicle, lighting equipment, a sign, etc. are displayed as a light spot. Here, for the sake of convenience, the dark area of the background is represented by a white background, and the light spot is drawn with dots.

  Normally, the light spots LP1 to LP3 due to the head lamps or tail lamps of the vehicles CAR1 to CAR3 and the light spots LP4 and LP5 due to the lighting facilities B and D are directly captured by the camera 2, so that they are circular or circular. The light spot has a shape close to. Further, the light spots LP6 and LP7 by the signs S1 and S2 are picked up by the camera 2 with the high brightness surface illuminated by the light from the illuminant, so that each shape, that is, a rectangle (square, rectangle) or rhombus The light spot has the shape of The white line LINE on the road surface is a long and narrow light spot LP8.

  The light spot extraction unit 64 extracts light spots that are candidates for inspection from this image. For example, a luminance value of a predetermined level is set as a threshold value, and an area having a luminance higher than this threshold value is detected. As a result, light spots of low brightness due to light reflected simply by the object, for example, road surface areas illuminated by the headlamps of the host vehicle are excluded, and light spots by a vehicle equipped with a sign or lamp equipped with a light emitter are detected. In the case of FIG. 4, all such light spots are detected.

  The light spot extraction unit 64 further sets the light spot area with the xy coordinates in the image for the detected light spot. That is, for the rectangular light spot LP6, a rectangular light spot area indicated by diagonal coordinates (xa, ya) to (xb, yb) is set as shown by a broken line in FIG. The same applies to the diamond-shaped light spot LP7 shown in FIG. 5B, the circular light spots LP1 to LP5 shown in FIG. 5C, and the white light spot LP8 shown in FIG. 5D, respectively. The light spot region is set with the maximum coordinate (xm, ym) to (xn, yn). In these figures, n and m are values set for each light spot.

  Next, the light spot identifying unit 64 identifies the shape of each light spot for each set light spot area. For example, as indicated by an arrow in FIG. 5A, while scanning in the x direction (row direction) from the position on the left side of the set light spot region, that is, from the position (xa, ya) to the right side. The luminance of the pixels in one row is measured. In this measurement, the above-described luminance threshold value can be used. Next, the above-mentioned line is performed, and this is repeated for several lines, for example, 2 to 5 lines in the y direction (column direction). Actually, about 3 lines are sufficient.

  As shown in FIG. 5A, when most of the pixels in each measured row have a luminance higher than the threshold value, the shape of the light spot is identified as a rectangle. When the number of pixels with higher luminance than the threshold decreases toward the upper row, the shape of the light spot is a rhombus or a circle as shown in FIG. 5B or 5C. Identify. In this case, the number of lines to be measured is increased by several lines, and the decrease state of the high luminance pixels is measured. When the rate of decrease is large, it is identified as a diamond in FIG. 5B, and when the rate of decrease is smaller than that, it is identified as a circle in FIG.

  Further, when the number of pixels having higher luminance than the threshold value is substantially the same, but the pixels having higher luminance gradually change to the left or right as the line moves to the upper row, FIG. As shown in d), the shape of the intersection is identified as linear. When the amount of change in position is constant, it is identified as a linear shape, and when the amount of change in position changes, it is identified as a curved shape.

  In the above-described identification, when identifying a diamond in FIG. 5B or a circle in FIG. 5C, 1 from the center of the left side of the light spot region in the y direction to the center of the bottom side in the x direction. You may make it measure several lines step by step down line by line. Or you may measure from the right side of a light spot area | region, and stepwise upwards. In such a case, if all the pixels have high luminance, the possibility of being a rhombus increases. In the case of a circle, the luminance of the pixel being measured is lowered, and finally the pixel having a higher luminance is measured again. This circle includes an ellipse.

  Then, if the light spot is identified as a rectangle or a diamond shape, a tag “non-target light spot” that is not a detection target is attached because it is highly likely that the light spot is a light spot by a marker. In addition, even when the light spot is identified as a line shape, the light spot is considered to be a light spot by a white line on the road, and therefore, a tag “non-target light spot” is attached. On the other hand, since a light spot identified as a circle is likely to be a light spot by a vehicle, a tag “target light spot” to be detected is attached. Then, the light spot identification unit 65 outputs each light spot with these tags attached to the vehicle detection unit 66. FIG. 6 shows a light spot with a tag “target light spot”. Here, the light spots LP6 and LP7 are tagged as “non-target light spots” and excluded from extraction.

  Of the identified light spots, the vehicle detection unit 66 detects the vehicle by determining the behavior, in this case, the moving state in the image, of the light spot of FIG. 6 with the tag “target light spot”. . In this detection method, for example, as in Patent Document 1, the relative position change of the light spot with respect to the host vehicle, that is, the movement direction and the movement speed are tracked over time, and the road is traveled based on the result. Detects a preceding vehicle or an oncoming vehicle. FIG. 7 shows the light spot of the vehicle detected in this way. Here, the light spots LP1 and LP2 are detected as the preceding vehicle CAR1 and the oncoming vehicle CAR2. Note that the light spot LP3 by the oncoming vehicle CAR3 is small in size, so that it can be detected that the vehicle is far away, and is excluded because there is little possibility of dazzling the oncoming vehicle CAR3 by the headlamp of the own vehicle at this time.

  Information on the oncoming vehicle and the preceding vehicle detected by the vehicle detection unit 66 in this way is output to the lighting control unit 63. Based on this information, the lighting control unit 63 sets the light distribution so as not to dazzle the preceding vehicle CAR1 and the oncoming vehicle CAR2, and executes the light distribution control in the lamp unit 1. Thereby, suitable ADB control is performed.

  As described above, in the embodiment, the light spot identifying unit 65 identifies a light spot that is highly likely to be a vehicle based on the image obtained by imaging with the camera 2, and the vehicle detecting unit 66 identifies this light spot. A process for detecting the vehicle only for the light spot is executed. Therefore, even when there are a large number of light spots in the captured image, detection processing is performed for light spots that are clearly not from the vehicle or light spots that are likely not from the vehicle. This eliminates the need to reduce the number of processing steps when detecting a vehicle, and enables rapid vehicle detection.

  Here, it is preferable to correct the moving direction of the light spot when detecting the vehicle, particularly when detecting the oncoming vehicle. That is, in the embodiment, the camera 2 is provided integrally with the headlamp HL, and is substantially at the same height as the headlamp of the oncoming vehicle. In some cases, it is lower than the headlamps of the oncoming vehicle. Therefore, when the camera is arranged at a position higher than the lamp of the oncoming vehicle, for example, the moving direction of the oncoming vehicle in the captured image is larger than when the camera is arranged on the front window or side mirror of the automobile. Is different.

  FIG. 8A is an image when the camera is located at a position higher than the headlamp HL of the oncoming vehicle CAR2 as in the case where the camera is disposed at the top of the front window of the host vehicle. In this case, the moving direction of the light spot LP2 by the headlamp HL of the oncoming vehicle CAR2 is directed to the lower right with respect to the horizontal line H in the image as indicated by an arrow.

  FIG. 8B is an image in the case where the camera is integrated with the headlamp as in this embodiment and is at a height or a position substantially equal to the headlamp HL of the oncoming vehicle CAR2. In this case, the moving direction of the light spot LP2 by the headlamp HL of the oncoming vehicle CAR2 is directed to the right along the horizontal line H in the image, or to the lower right to some extent with respect to the horizontal line H.

  Therefore, the vehicle detection unit 66 of the embodiment can detect the light spot moving in the horizontal direction in the image or in the lower right direction from the horizontal direction as the light spot of the oncoming vehicle. However, when the camera 2 provided on the host vehicle is tilted in the roll direction (rotation direction about the longitudinal axis of the automobile), for example, the camera 2 is mounted tilted in the left-down direction. Alternatively, when the vehicle body is traveling in the same direction due to a deviation in the load weight of the host vehicle, the image captured by the camera 2 is also inclined to the lower left as shown in FIG. 8C.

  In such an image, the light spot LP2 of the headlamp HL of the oncoming vehicle CAR2 is moved to the upper right with respect to the horizontal line H in the image. Therefore, when the light spot is detected only in the moving direction in the image, the light spot LP2 of the oncoming vehicle may be erroneously detected as a light spot other than the vehicle, for example, a sign or a road illumination light. Therefore, when detecting the vehicle, it is necessary to detect not only the moving direction of the light spot in the image but also the moving speed.

  Therefore, the vehicle detection unit 66 refers to the light spot that has been identified as the “non-target light spot” by the light spot identification unit 65. For example, in the example of FIG. 8C, the light spot LP6 of the sign S1 identified as having a rectangular shape is referred to, and the detected horizontal side, that is, the direction of the bottom side or the top side is set as the corrected horizontal direction DH. Then, the moving direction of the light spot LP2 of the oncoming vehicle CAR2 in the image indicated by the arrow is corrected as the direction with respect to the corrected horizontal direction DH.

  Usually, since the horizontal side of the rectangular sign S1 is oriented in the horizontal direction, this correction corrects the horizontal direction of the image even if the camera 2 or the vehicle body is tilted in the roll direction. The direction DH can be corrected. If the moving direction of the light spot LP6 of the oncoming car CAR2 is detected with reference to the corrected horizontal direction DH, it can be detected that the light spot LP6 is moving in the horizontal direction or the lower right direction, and the oncoming car CAR2 is detected. It becomes possible to detect accurately. That is, even if the moving speed of the light spot is not detected at the time of detection, the vehicle can be accurately detected only by detecting the moving direction, thereby simplifying the processing and speeding up the detection.

  In this case, when a plurality of rectangular light spots are identified in the image, an average of the bottom and top sides of the plurality of rectangles may be taken as the reference horizontal direction. Alternatively, the bottom side or the top side of the rectangular light spot having the largest dimension, which can easily identify the shape with high accuracy, may be employed. If a rectangular light spot is not identified, the reference horizontal direction may be set using a light spot of another shape. For example, in the case of a diamond-shaped light spot, a horizontal diagonal line may be used as the reference horizontal direction. If the vehicle has a vehicle height sensor or the like and the roll angle of the host vehicle can be detected, the moving direction of the light spot in the image may be corrected based on the detected roll angle. Good.

  In the present invention, the light spot LP8 of the white line LINE identified by the light spot identifying unit 65 may be used to further improve the accuracy of vehicle detection, particularly the accuracy of oncoming vehicle detection. In this embodiment, the camera 2 is provided integrally with the headlamp HL, and is provided at a lower position closer to the road surface than when provided on the front window or side mirror. The white line LINE can be imaged more clearly. In particular, since the white line LINE is a light spot obtained by imaging light reflected from the headlamps of the oncoming vehicle and the host vehicle, it is difficult to normally capture images with high brightness. By providing in, it can image as a high-intensity light spot.

  Therefore, it becomes possible to clearly image each light spot LP8 of the slightly distant white line LINE illuminated by the headlamp of the host vehicle and the adjacent white line illuminated by the headlamp HL of the oncoming vehicle CAR2 with high brightness, Recognition of the white line LINE becomes easy and accurate. Along with this, it becomes possible to accurately recognize whether the road on which the vehicle is traveling is a straight road or a curved road.

  If the road condition can be accurately recognized, the vehicle detection unit 66 is in the vicinity of the recognized white line light spot LP8 among the light spots of the plurality of “target light spots” identified by the light spot identifying unit 65, and Only the light spot that exists at the position along the extension direction of the light spot LP8 and moves in the extension direction can be detected as the light spot of the vehicle. Therefore, by performing vehicle detection based only on the detected light spot based on the moving direction and moving speed of the light spot, the accuracy of vehicle detection is improved, and the processing steps for vehicle detection are simplified, and the detection speed is increased. Can be realized.

  Further, by recognizing the road state using the light spot LP8 by the white line in this way, the moving direction of the light spot in the image is a direction close to the light spot LP8 of the white line and along the extending direction thereof. It is possible to accurately detect the vehicle simply by detecting. That is, even when the host vehicle rolls and the camera and the vehicle body are tilted, accurate vehicle detection is possible without considering the corrected horizontal direction as described with reference to FIG. 8C. The complexity of processing can be avoided.

  When the white line is a curve, that is, when the road state is a curved road, the vehicle detection unit 66 calculates the curvature from the shape of the light spot LP8 when it is identified by the light spot identifying unit 65. When the target light spot in the image is moving in the horizontal direction, correction such as subtraction or addition of the movement amount may be performed based on the direction of the curved path and the calculated curvature. This makes it possible to accurately detect the vehicle from the light spot.

  In the above embodiment, the example of the light distribution control by the ADB control of the headlamp is shown in the vehicle detection device of the present invention. However, the vehicle detection device may be used to perform AHB control and other light distribution control. Further, since the vehicle can be detected quickly and accurately, it can also be used for automatic driving control of an automobile.

  Further, the vehicle detection device of the present invention can be configured to take an image of a side region and a rear region of an automobile and detect other vehicles existing in a side region and a rear region of the host vehicle. In this case, the camera may be integrated into a side mirror or tail lamp.

HL Headlamp 1 Lamp unit 2 Camera (imaging means)
3 Lamp housing 6 Lamp ECU
61 Signal Processing Unit 62 Image Analysis Unit (Vehicle Detection Unit)
63 lighting control part 64 light spot extraction part 65 light spot identification part 66 vehicle detection part LP1-LP8 light spot

Claims (9)

  A vehicle detection method for detecting a vehicle from the behavior of a light spot that is mounted on a vehicle and that captures the surroundings of the vehicle and that is captured in an image obtained by the imaging, wherein a plurality of light spots present in the image are detected. A step of recognizing a shape, a step of identifying a light spot that is a vehicle detection target in the recognized shape as a target light point, and a light spot other than that as a non-target light point, and a behavior of the target light spot in the image And detecting a light spot having a predetermined behavior as a vehicle.   The vehicle detection method according to claim 1, wherein the shape of the non-target light spot is referred to when detecting the behavior of the target light spot.   A vehicle detection apparatus comprising: an imaging unit that is mounted on a vehicle and images the surroundings of the vehicle; and a detection unit that detects a vehicle from a light spot captured in an image captured by the imaging unit, and is present in the image A light spot identifying unit for identifying a target light spot to be detected from a plurality of light spots, a non-target light spot, and a vehicle detection for detecting a vehicle from the behavior of the identified target light spot in the image A vehicle detection device that recognizes shapes of the plurality of light spots and identifies the target light spot and the non-target light spot based on the recognition.   The vehicle detection device according to claim 3, wherein the light spot identifying unit identifies a light spot having a rectangular shape, a diamond shape, or a linear shape as a non-target light spot, and identifies a light spot having a circular shape or a shape close thereto as a target light spot. .   The vehicle detection device according to claim 3, wherein the vehicle detection unit detects a vehicle based on a moving direction of the target light spot in the image.   The vehicle detection device according to claim 5, wherein the vehicle detection unit corrects a moving direction of the target light spot based on a shape recognized by the non-target light spot.   When the shape recognized from the non-target light spot is a rectangle, the vehicle detection unit sets the direction of the bottom or top of the rectangle as a reference horizontal direction, and moves the target light spot based on the reference horizontal direction. The vehicle detection device according to claim 6, wherein the direction is corrected.   The vehicle detection device according to claim 5, wherein the vehicle detection unit recognizes a white line on a road from a linear non-target light spot and detects a light spot in a moving direction along the white line as a vehicle. The vehicle detection device according to claim 3, wherein the imaging unit is provided integrally with a lamp of the vehicle.

Images