JP2005291805A - Obstacle detection method and obstacle detection system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a concrete configuration which detects the width and height of an obstacle ahead of an own vehicle with high precision by combining the detection on the basis of image processing on an image of a camera and the detection on the basis of the results of radar sensing. <P>SOLUTION: The image side detection length of the width of an obstacle is repeatedly calculated by image processing on a taken image ahead of the own vehicle. On the basis of the ratio of a variance of radar side detection length to a variance of the image side detection length, the ratio of the radar side detection length to the image side detection length to be adopted for the detection of the width of the obstacle is determined. The width of the obstacle is detected through the addition of the radar side detection length and the image side detection length on the basis of the determined ratio. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an obstacle detection method and an obstacle detection device that detect the width or width and height of an obstacle such as a preceding vehicle ahead of the host vehicle.

  Conventionally, in an advanced safety vehicle called ASV, a vehicle equipped with an automatic brake system, a brake assist system, etc., in order to determine the presence or absence of a rear-end collision, etc. It is necessary to recognize an obstacle such as a preceding vehicle in front of the vehicle and to detect at least its width.

  If not only the horizontal width but also the height is detected, the size, shape, etc. of the obstacle can be recognized from the detection result.

  By the way, conventionally, there are radars such as laser radars and millimeter wave radars and cameras such as CCD monocular cameras and stereo cameras as the obstacle capturing sensors. In the case of the former radar, scanning in the horizontal direction (horizontal direction) is performed. In the case of the latter camera, the horizontal width can be detected from the search result in front of the vehicle based on the above, and the height can be detected from the search result in front of the vehicle based on the scanning in the height direction (vertical direction). The width and height can be detected by image processing of the image.

  And since the detection accuracy from the captured image is in units of pixels and higher than the detection accuracy from the radar search results, this kind of obstacle recognition has conventionally been performed by image processing of the captured image of the camera. The detection of the width of the object or the detection of the width and height is performed to recognize the presence or absence of an obstacle collision, the size and shape of the obstacle, and the like (see, for example, Patent Document 1).

  When a camera and a radar are mounted on the own vehicle, the radar search result is mainly used for detecting the inter-vehicle distance between the own vehicle and the preceding vehicle.

Japanese Patent Laid-Open No. 7-49999 (paragraphs [0002] and [0014], FIG. 1)

  When detecting the width of an obstacle or the detection of the width and height only from image processing of the image taken by the camera as in the prior art, shooting can be performed depending on the shooting conditions such as the driving environment including the weather conditions and shooting lighting. There is a problem that it cannot always be detected with high detection accuracy because it disappears or the photographing accuracy decreases.

  In order to improve the recognition accuracy by detecting the width of the obstacle or the detection of the width and height as much as possible, the detection based on the image processing of the captured image of the camera and the detection based on the radar search result are combined. However, a specific configuration for detecting the width or width and height of an obstacle by combining them is not invented.

  The present invention provides a specific configuration for detecting the width or width and height of an obstacle ahead of the host vehicle as accurately as possible by combining detection based on image processing of a captured image of a camera and detection based on a radar search result. The purpose is to do.

  In order to achieve the above-described object, the obstacle detection method of the present invention repeatedly calculates the radar-side detection length of the width of an obstacle in front of the host vehicle from the radar search result in front of the host vehicle, and takes a captured image in front of the host vehicle. The image side detection length of the width of the obstacle is repeatedly calculated by the image processing, and is adopted for the detection of the width of the obstacle based on the ratio of the variance value of the radar side detection length and the variance value of the image side detection length. Determining the ratio between the radar-side detection length and the image-side detection length, and detecting the width of the obstacle by adding the determined ratio of the radar-side detection length and the image-side detection length. It is characterized (claim 1).

  In the obstacle detection method of the present invention, the radar side detection length of the width and height of the obstacle is repeatedly calculated from the radar search result in front of the host vehicle, and the obstacle is detected by image processing of the captured image in front of the host vehicle. The image side detection length of the width and height is repeatedly calculated, and the obstacle is based on the ratio of the dispersion value of the radar side detection length and the dispersion value of the image side detection length for each of the width and height of the obstacle. The ratio of the radar-side detection length and the image-side detection length employed for the detection of each of the horizontal width and height of the obstacle is determined, and the radar-side detection length and the image-side detection length of the horizontal width and height of the obstacle are determined. The lateral width and height of the obstacle are detected by adding the determined ratios.

  Further, according to the obstacle detection method of the present invention, the ratio between the dispersion value of the radar side detection length and the dispersion value of the image side detection length is corrected by the dispersion characteristics of the detection results of the radar exploration and the captured image. (Claim 3), the adoption ratio between the radar-side detection length and the image-side detection length is determined from a data map in which the adoption ratio based on the ratio of the dispersion values of the two detection lengths is set in advance. It is also characterized in that it is determined (Claim 4), and the dispersion value of the radar side detection length and the dispersion value of the image side detection length are corrected according to the distance between the vehicle and the obstacle (Claims). 5) Further, the adoption ratio between the radar side detection length and the image side detection length is corrected by the detection environment conditions such as the weather condition of the driving environment and the photographing illumination condition (Claim 6), Detect horizontal width or horizontal width and height for a certain period of time. It is also characterized in that it is returned and updated (claim 7), and the detection of the width of the obstacle or the detection of the width and height is repeated until the dispersion values of the radar side detection length and the image side detection length are almost converged. (Claim 8).

  Next, the additional obstacle detection device of the present invention includes a radar for exploring the front of the own vehicle, a camera for photographing the front of the own vehicle, and a radar side detection length of the width of the obstacle ahead of the own vehicle based on the result of the radar exploration. Detection length calculation means for repeatedly calculating and calculating the image side detection length of the width of the obstacle repeatedly by image processing of the captured image of the camera, a variance value of the radar side detection length, and a variance of the image side detection length A ratio determining means for calculating a value and determining a ratio between the radar side detection length and the image side detection length to be used for detecting the width of the obstacle based on the ratio of the two variance values; and the radar side detection Detection processing means for detecting the width of the obstacle by adding the determined ratio of the length and the detection length on the image side is provided (claim 9).

  The obstacle detection device according to the present invention includes a radar for exploring the front of the own vehicle, a camera for photographing the front of the own vehicle, and a radar side of the width and height of the obstacle ahead of the own vehicle according to the result of the radar exploration. A detection length calculation unit that repeatedly calculates a detection length, and repeatedly calculates an image side detection length of the width and height of the obstacle by image processing of the captured image of the camera, and the width and height of the obstacle respectively. The both detections that are used to detect the lateral width and height of the obstacle based on the ratio of the dispersion values of the both detection lengths, by calculating the dispersion value of the radar side detection length and the dispersion value of the image side detection length. A ratio determining means for determining a ratio of the length, and by adding the determined ratio of the radar side detection length and the image side detection length of each of the obstacle width and height respectively, the width and height of the obstacle Detection processing means for detecting It is characterized by comprising (claim 10).

  Further, according to the obstacle detection of the present invention, the ratio between the dispersion value of the radar-side detection length and the dispersion value of the image-side detection length is corrected by the dispersion characteristics of the detection results of the radar search and the captured image. The ratio determination means holds a data map in which the adoption ratio based on the ratio between the dispersion value of the radar side detection length and the dispersion value of the image side detection length is set in advance in the ratio determination means, The adoption ratio of the both detection lengths is also determined from the data map (claim 12), and the ratio determination means automatically determines the dispersion value of the radar side detection length and the dispersion value of the image side detection length. A function of correcting according to the distance between the vehicle and the obstacle is provided (Claim 13), and the ratio determining means determines the adoption ratio of the radar-side detection length and the image-side detection length to the weather in the driving environment. Correction according to detection conditions such as shooting conditions and shooting illumination conditions It is also characterized in that it has a function (Claim 14), and the detection processing means repeats and updates the detection of the width of the obstacle or the detection of the width and height for a certain time (Claim 15), The detection processing means repeats and updates the detection of the width of the obstacle or the detection of the width and height until the dispersion values of the radar side detection length and the image side detection length are almost converged (claim 16). .

  First, according to the configurations of claims 1 and 9, since the dispersion values of the radar side detection length and the image side detection length correspond to the respective detection accuracy, the dispersion value of the radar side detection length and the image side detection length Based on the ratio to the variance value, the detection width is detected by adding both detection lengths in proportion to the detection accuracy, and detection based on image processing of the captured image of the camera and detection based on the radar search result In combination, it is possible to provide a specific configuration capable of detecting the width of an obstacle as accurately as possible even when shooting is difficult.

  According to the second and tenth aspects of the present invention, both detection lengths are determined based on the ratio between the dispersion value of the radar side detection length and the dispersion value of the image side detection length for each of the width and height of the obstacle. By adding a ratio according to the detection accuracy of the camera, it is possible to combine detection based on image processing of the captured image of the camera and detection based on the radar search results, and to obtain the width of the obstacle as accurately as possible even when shooting is difficult. In addition, a specific configuration capable of detecting the height can be provided.

  Furthermore, according to the configurations of claims 3 and 11, the inherent dispersion characteristics of the radar exploration and the photographed image are taken into consideration, and the ratio of the dispersion value of the radar side detection length to the dispersion value of the image side detection length is determined by their inherent characteristics. Therefore, the detection accuracy is further improved.

  According to the configuration of claims 4 and 12, the data map in which the adoption ratio between the dispersion value of the radar side detection length and the image side detection length is set in advance based on the ratio of the dispersion values of both detection lengths. It can be easily determined by determining from

  Next, according to the configurations of claims 5 and 13, detection can be performed by performing radar exploration according to the distance between the vehicle and the obstacle, and correcting the change in detection accuracy of the camera-captured image. According to the configuration, detection can be performed by performing correction in consideration of detection conditions such as weather conditions of the driving environment and shooting illumination conditions.

  Furthermore, according to the configurations of claims 7 and 15, the detection of the width of the obstacle or the detection of the width and height is repeated and updated for a certain period of time, thereby reducing the processing burden and the like. According to the configuration of claims 8 and 16, the detection of the width of the obstacle or the detection of the width and height is performed by the radar side detection length and the image side detection length dispersion. By repeatedly updating the values until they almost converge, the processing load can be reduced, and the width of the obstacle can be detected with high accuracy in the width and height under stable conditions.

  Next, in order to describe the present invention in more detail, an embodiment thereof will be described in detail with reference to FIGS.

  1 is a block diagram of obstacle detection / collision avoidance of the vehicle 1, FIG. 2 is a flowchart for explaining the operation of obstacle detection in FIG. 1, and FIG. 3 is a schematic diagram of a data map.

(Constitution)
A vehicle 1 shown in FIG. 1 is, for example, an advanced safety vehicle (ACC). In FIG. 1, 2 is a control switch unit including various control switches operated by a driver, 3 is a wiper, and 4 is a front image of the vehicle. The CCD single-lens camera camera 5 is a radar having a laser radar configuration for radar exploration in front of the host vehicle. In this embodiment, the width of an obstacle such as a preceding vehicle in front of the host vehicle and the vehicle and the obstacle In order to detect the distance, the vehicle front is searched only in the vehicle width direction (horizontal direction).

  The various control signals of the control switch unit 2, the operation display signal of the wiper 3, the photographed image signal of the camera 4, and the search signal of the radar 5 are microcomputers that perform various operations for obstacle detection / collision avoidance calculation and control. This EUU 6 executes a preset obstacle detection / collision avoidance program, performs the obstacle detection process in steps S1 to S9 in FIG. i) to (iii) are formed.

(I) Detection length calculation means This means repeatedly calculates the radar-side detection length Lr of the width of the obstacle ahead of the host vehicle, for example, every scanning scan of the radar 5 with a period of seconds. The image side detection length Li of the width of the obstacle is repeatedly calculated by image processing of the captured image.

  The detection lengths Lr and Li are calculated from the distance between the reflection points obtained by the radar 5 search, for example, and detected from the distance between the vertical peak level points in the horizontal direction of the captured image. It is obtained by calculating the length Li.

(Ii) Ratio determination means This means that each time a new detection length Lr, Li is obtained, the radar side detection is performed based on a certain number of data from all or the latest detection lengths Lr, Li. The ratio Pr of the detection lengths Lr and Li, which is used for detecting the lateral width of the obstacle, is calculated based on the ratio of the calculated dispersion values σR and σI. (%) And Pi (%) are determined.

  In this embodiment, considering that the dispersion characteristics of the detection results of the radar exploration and the captured images are different, the characteristics error is compensated for and the conditions of the dispersion values σR and σI are equalized. The dispersion coefficients σR and σI are multiplied by the correction coefficients K1 and K2 set by the above, and the ratio between the dispersion values σR and σI is corrected to the corrected dispersion values K1 × σR and K2 × σI. Note that K1 + K2 = 100.

  Further, the ratios Pr (%) and Pi (%) of the detection lengths Lr and Li are determined by either the calculation method or the map method described below.

(A) Calculation method In this method, for example, the ratio Pi (%) is obtained from the calculation of the following calculation formula based on the ratio of the dispersion values K1 × σR (or σR), K2 × σI (or σI), The ratio Pr (%) is obtained from the above calculation.

Pi (%) = 100 × σI / (K1 × σR + K2 × σI)
(B) Map method This method is a data map in which the ratio determination means uses a variance value σR (or K1 × σR), σI (or K2 × σI) in advance, and the two-dimensional parameter shown in FIG. This is a method of holding the data map MP of σR and σI, and determining the adoption ratio of both detection lengths Lr and Li from the data map MP. Specifically, the ratio Pi (%) is determined from the data map MP, and 100 This is a method for obtaining the ratio Pr (%) from -Pi (%), and it is possible to determine the ratio Pi (%) and Pr (%) very easily without calculation.

  The characteristics of the data map MP are set based on various experiments.

  The ratios Pr (%) and Pi (%) indicate that the new detection lengths Lr and Li are obtained during the obstacle width detection period, and the variance value σR (or K1 × σR), σI (or K2 × Each time (σI) is updated, it is newly determined by the calculation method or the map method.

(Iii) Detection processing means This means is obtained by adding the determined ratio Pr (%) and Pi (%) of the detection lengths Lr and Li (= Lr × (Pr / 100) + Li × (Pi / 100)). Repeatedly calculate and detect the width of the obstacle.

  In this embodiment, in order to reduce the processing load on the ECU 6, the width of the obstacle is detected for a certain period of time of about one minute after detecting the presence or change of the obstacle, for example, from a radar survey or a captured image. The updating of the detected width is stopped after that.

  Next, the ECU 6 determines the possibility of collision with the obstacle and a method for avoiding the collision based on the detected width of the obstacle, the own vehicle, the predicted travel path of the obstacle, and the like.

  In FIG. 1, 7 is a display alarm unit for displaying the possibility of collision and its avoidance method on a screen, 8 is a throttle control unit, 9 is an automatic transmission mechanism (AT) control unit, 10 is a brake control unit, 11 Is a steering control unit, and the units 8 to 11 are operation-controlled based on the possibility of the collision of the ECU 6 and the determination of the avoidance method, and automatically and variably control the traveling of the host vehicle 1 to avoid the collision.

(Operation)
Next, the obstacle detection operation of FIG. 1 will be described with reference to the flowchart of FIG.

  First, when the ignition key of the host vehicle 1 is turned on and the engine is started, the radar 5 and the camera 4 execute radar search and photographing in front of the host vehicle in steps S1 and S2 of FIG.

  Next, at steps S3 and S4 in FIG. 2, the detection length calculation means of the ECU 6 operates to calculate the detection lengths Lr and Li.

  Further, the ratio determining means of the ECU 6 operates, and the dispersion values σR and σI of the detection lengths Lr and Li are calculated in steps S5 and S6 in FIG. 2. At this time, the conditions of the dispersion values σR and σI are equalized. The ratio of the values σR and σI is corrected to the corrected dispersion values K1 × σR and K2 × σI.

  Further, based on the ratio of the dispersion values σR, σI calculated in step S7 by the operation of the ratio determining means of the ECU 6, for example, the detection lengths Lr, Li used for detecting the width of the obstacle from the data map MP of FIG. The ratios Pr (%) and Pi (%) are determined.

  Then, in step S8, the detection processing means operates, and the lateral width of the obstacle is calculated and detected by adding the determined ratios Pr (%) and Pi (%) of the detection lengths Lr and Li.

  Further, the process returns from step S8 to steps S1 and S2 through step S9 until the set fixed period elapses, and the operations from these steps S1 and S2 are repeated to determine the determined ratio Pr (%) of the detection lengths Lr and Li. ), Pi (%) is updated, the width of the obstacle is calculated and detected repeatedly, and when a certain period of time has elapsed, step S9 is passed in affirmative (YES), and the process is terminated.

  Since the dispersion values σR and σI correspond to the detection accuracy of each of the detection lengths Lr and Li, based on the ratio of the dispersion values σR and σI, both detection lengths Lr and Li are added in proportions according to the detection accuracy. Thus, the lateral width of the obstacle can be detected with high accuracy, and the detection based on the image processing of the captured image of the camera 4 and the detection based on the search result of the radar 5 are combined to be as accurate as possible even when shooting is difficult. It is possible to provide a specific configuration that can detect the width of the obstacle well, and as a result, it is possible to improve the possibility of collision and the control accuracy of avoidance thereof.

  Further, by taking into account the inherent dispersion characteristics of the radar exploration and the captured image, the ratio of the dispersion values σR and σI is obtained by correcting the inherent dispersion characteristics, thereby further improving the width detection accuracy of the obstacle. improves.

  Furthermore, by limiting the update of the detection of the width of the obstacle within a certain time, the width of the obstacle can be detected while reducing the processing load of the ECU 6 and the like.

The present invention is not limited to the above-described embodiment, and various modifications other than those described above can be made without departing from the spirit thereof. For example, in the above-described embodiment, the radar 5 In this way, only the vehicle width direction (horizontal direction) in front of the own vehicle is searched, but the radar 5 is formed by a horizontal scanning radar and a vertical scanning radar, and the height direction (vertical direction) in front of the own vehicle is also set. Further, the detection length calculation means of the ECU 6 repeatedly calculates the radar side detection length of the width and height of the obstacle ahead of the host vehicle from the search result of the radar 5, and performs image processing of the image captured by the camera 4 The image side detection length of the obstacle width and height is repeatedly calculated by using the ratio determining means, and the radar side detection length Lr dispersion value σR and the image side detection length of the obstacle width and height are respectively calculated by the ratio determining means. The variance value σI of i is calculated, and based on the ratio of both variance values σR, σI, the ratio of both detection lengths Lr, Li used for detecting the width and height of the obstacle is determined, and the detection processing means The width and height of the obstacle may be detected by adding the determined ratio of the radar-side detection length Lr and the image-side detection length Li for each of the width and height of the obstacle. The width and height of the obstacle can be detected with high accuracy. As a result, the shape of the obstacle can be detected with high accuracy, and the obstacle can be specified from the detection.

  In the detection of the horizontal width and height, the ratio Pr (%) and Pi (%) of the detection lengths Lr and Li are determined from the map data of the horizontal width and height similar to the data map MP by the map method. The ratio Pr (%) and Pi (%) of the detection lengths Lr and Li can also be determined by calculation of the calculation method. In addition, taking into account the inherent dispersion characteristics of radar exploration and captured images, the ratio of dispersion values σR and σI can be determined by correcting these inherent dispersion characteristics to further improve the accuracy of obstacle width detection. Furthermore, by limiting the update of the detection of the width of the obstacle within a certain time, it is possible to detect the width and height of the obstacle by reducing the processing load of the ECU 6 and the like. Of course.

  Next, in both the detection of the width of the obstacle and the detection of the width and height, the ratio determination means uses the adoption ratio of the radar side detection length Lr and the image side detection length Li, the daytime and nighttime of the driving environment, bad weather, etc. For example, when the quality of the captured image is reduced due to changes in lighting, etc. at night or in bad weather. From the standpoint of improving detection accuracy, it is preferable to correct the ratio Pr (%) and Pi (%) by applying a correction coefficient so that the Lr adoption ratio increases.

  At night, it can be detected from a change in image brightness, a headlight lighting signal, a car clock, etc., and bad weather can be detected from, for example, the operation of the wiper 3 in FIG. The illumination condition can be detected from a change in image brightness.

  Next, if the ratio determining means has a function of correcting the dispersion value of the radar side detection length and the dispersion value of the image side detection length according to the distance between the vehicle 1 and the obstacle, the vehicle 1 and the obstacle If the distance to the object is within a predetermined distance, the horizontal resolution of the camera 4 is better than the search performance of the radar 5, and if the distance between the vehicle 1 and the obstacle is a long distance, the search performance of the radar 5 is the camera. If the distance between the host vehicle 1 and the obstacle is a short distance, a larger correction coefficient is applied to the ratio Pi (%) of the detection length Li than the ratio Pr (%) of the detection length Lr. When the ratio Pi (%) is corrected so as to increase, and the distance becomes longer, the ratio Pr (%) is obtained by multiplying the ratio Pr (%) of the detection length Lr by a correction coefficient larger than the ratio Pi (%) of the detection length Li. Correct so that it becomes larger, and use more accurate detection lengths Li and Lr. Therefore, the width of the obstacle, a more accurate detection of the width and height allows.

  The camera 4 may be a stereo camera or the like, the radar 5 may be a millimeter wave radar or the like, and the data map may have a different map shape from the data map MP of FIG. The obstacle detection program of the ECU 6 may be a program having a processing procedure different from that shown in FIG.

  The present invention can be applied not only to obstacle recognition of advanced safety vehicles, but also to obstacle recognition of vehicles equipped with an automatic brake system, a brake assist system, and the like.

  By the way, the camera 4 and the radar 5 of FIG.

It is a block diagram of one embodiment of the present invention. It is a flowchart for operation | movement description of FIG. It is a schematic diagram of the data map for operation | movement description of FIG.

Explanation of symbols

1 Vehicle 4 Camera 5 Radar 6 ECU

Claims (16)

Repeatedly calculate the radar-side detection length of the width of obstacles ahead of the vehicle based on the radar search results in front of the vehicle,
The image side detection length of the width of the obstacle is repeatedly calculated by image processing of the captured image in front of the vehicle,
Based on the ratio between the dispersion value of the radar-side detection length and the dispersion value of the image-side detection length, determine the ratio of the radar-side detection length and the image-side detection length to be used for detecting the width of the obstacle,
An obstacle detection method comprising: detecting a lateral width of the obstacle by adding a determined ratio of the radar side detection length and the image side detection length. Repeatedly calculate the radar side detection length of the width and height of the obstacle from the radar survey results in front of the vehicle,
The image side detection length of the width and height of the obstacle is repeatedly calculated by image processing of the captured image in front of the host vehicle,
Based on the ratio between the dispersion value of the radar side detection length and the dispersion value of the image side detection length for each of the width and height of the obstacle, the radar side employed for detection of each of the width and height of the obstacle Determine the ratio between the detection length and the image side detection length,
Obstacle detection characterized by detecting the width and height of the obstacle by adding the determined ratio of the radar side detection length and the image side detection length of each of the width and height of the obstacle. Method.   The ratio between the dispersion value of the radar-side detection length and the dispersion value of the image-side detection length is a ratio of the two dispersion values corrected by the inherent detection dispersion characteristics of each of the radar search and the photographed image. 3. The obstacle detection method according to 1 or 2.   4. The adoption ratio between the radar side detection length and the image side detection length is determined from a data map in which an adoption ratio based on a ratio of dispersion values of the two detection lengths is set in advance. The obstacle detection method according to claim 1.   The obstacle detection according to claim 1, wherein the dispersion value of the radar side detection length and the dispersion value of the image side detection length are corrected according to the distance between the vehicle and the obstacle. Method.   The obstacle detection according to any one of claims 1 to 5, wherein the adoption ratio between the radar side detection length and the image side detection length is corrected by a detection condition such as a weather condition of a traveling environment and a photographing illumination condition. Method.   The obstacle detection method according to any one of claims 1 to 6, wherein the detection of the width of the obstacle or the detection of the width and height is repeated for a predetermined time.   The detection of the width of the obstacle or the detection of the width and height is repeated and updated until the dispersion values of the radar side detection length and the image side detection length are almost converged. The obstacle detection method described. Radar that searches in front of the vehicle,
A camera that captures the front of the vehicle,
Detection length calculation that repeatedly calculates the radar side detection length of the width of the obstacle ahead of the vehicle based on the radar search result, and repeatedly calculates the image side detection length of the width of the obstacle by image processing of the captured image of the camera Means,
The radar-side detection length and the image-side detection are calculated for calculating the dispersion value of the radar-side detection length and the dispersion value of the image-side detection length, and are employed for detecting the width of the obstacle based on the ratio of the two dispersion values. A ratio determining means for determining a ratio with the length;
An obstacle detection apparatus comprising: a detection processing unit configured to detect a width of the obstacle by adding a determined ratio between the radar side detection length and the image side detection length. Radar that searches in front of the vehicle,
A camera that captures the front of the vehicle,
The radar side detection length of the obstacle in front of the vehicle is repeatedly calculated from the radar search result, and the image side detection length of the obstacle width and height is calculated by image processing of the image captured by the camera. Detection length calculation means for repeatedly calculating;
A variance value of the radar side detection length and a variance value of the image side detection length for each of the lateral width and height of the obstacle are calculated, and a variance value of the both detection lengths for each of the lateral width and height of the obstacle is calculated. Based on the ratio, a ratio determining means for determining a ratio between the radar side detection length and the image side detection length adopted for detection of the width and height of the obstacle,
Detection processing means for detecting the width and height of the obstacle by adding the determined ratio of the radar-side detection length and the image-side detection length for each of the width and height of the obstacle. Obstacle detection device characterized by the above.   The ratio between the dispersion value of the radar-side detection length and the dispersion value of the image-side detection length is a ratio of the two dispersion values corrected by the inherent detection dispersion characteristics of each of the radar search and the photographed image. The obstacle detection apparatus according to 9 or 10.   The ratio determining means holds a data map in which the adoption ratio based on the ratio of the dispersion value of the radar-side detection length and the dispersion value of the image-side detection length is preset, and the adoption ratio of both detection lengths is determined from the data map The obstacle detection device according to any one of claims 9 to 11, wherein   The ratio determining means has a function of correcting the dispersion value of the radar side detection length and the dispersion value of the image side detection length in accordance with the distance between the vehicle and the obstacle. The obstacle detection device according to any one of the above.   14. The ratio determining means is provided with a function of correcting the adoption ratio between the radar side detection length and the image side detection length according to detection conditions such as a weather condition of a traveling environment and a photographing illumination condition. The obstacle detection device according to any one of the above.   The obstacle detection device according to claim 9, wherein the detection processing means repeatedly updates the detection of the width of the obstacle or the detection of the width and height for a predetermined time.   15. The detection processing means repeats and updates the detection of the width of the obstacle or the detection of the width and height until the dispersion value of the radar side detection length and the image side detection length almost converges. The obstacle detection device according to any one of the above.

Images