JP2008309637A - Obstruction measuring method, obstruction measuring apparatus, and obstruction measuring system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an obstruction measuring method, an obstruction measuring apparatus, and an obstruction measuring system which can ensure distance accuracy, reduce processing time, and reduce a circuit scale by setting a window size in accordance with the distance to an obstruction. <P>SOLUTION: The obstruction measuring method measures the distance to the obstruction by parallax calculation through a stereo image. In the method, on the basis of distance information determined by a low resolution image, a window size for corresponding point search for parallax calculation in a high resolution image is set. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an obstacle measurement method, an obstacle measurement device, and an obstacle measurement system for measuring a distance to an obstacle.

  In distance measurement using a stereo image, a reference image is searched for an area where the pattern matches in one area of a standard image of a pair of images, a parallax between corresponding points is calculated, and the principle of triangulation is calculated from the parallax. In this case, conventionally, the corresponding point search is performed by a correlation method with a fixed region size.

In Patent Document 1, after digitizing images captured by two cameras, each partial image to be analyzed is input to a matching unit corresponding to each distance range to which the image is expected to belong. By changing the resolution of the input image, the distances are collated with a necessary and sufficient spatial resolution in each distance range, and each collating unit converts the input image to 2 ⁻ⁿ (in accordance with each corresponding distance range. Disclosed is a distance distribution detection device capable of reducing the amount of calculation by converting to an image having a resolution of n = 0, 1, 2, 3). That is, the amount of calculation can be reduced by converting the number of pixels and resolution to be collated for each distance range where the object is expected to belong, or changing the size of the area so that it is small at a long distance and large at a short distance. Reduce.
Japanese Patent Laid-Open No. 2001-126065

  Conventionally, the window (area) size for searching for corresponding points in parallax calculation is fixed, or in most cases, a different window size is applied at a specific position in the image area, and the distance in the image There is no example of setting the window size from the information or the expected distance range. However, when the object is a short distance, if the window size of the corresponding point search in the parallax calculation is increased as in Patent Document 1, the influence of the occlusion (concealment) with the background increases in the vicinity of the edge where the feature can be easily taken. , There is a possibility that the distance accuracy is lowered without taking corresponding points.

  In view of the above-described problems of the prior art, the present invention provides an obstacle capable of ensuring distance accuracy, reducing processing time, and reducing circuit scale by setting a window size according to the distance to the obstacle. It aims at providing a measuring method, an obstacle measuring device, and an obstacle measuring system.

  In order to achieve the above object, a first obstacle measurement method according to the present invention is an obstacle measurement method for measuring a distance to an obstacle by parallax calculation using a stereo image, and distance information obtained from a low-resolution image. Based on the above, the window size of the corresponding point search for the parallax calculation in the high resolution image is set.

  According to this obstacle measurement method, the window size of the corresponding point search for calculating the distance information to the obstacle in the high resolution image is determined based on the approximate parallax (distance information) obtained from the low resolution image. For example, when the object is at a short distance, the influence of the occlusion with the background becomes large near the edge where the feature is easy to take, but it becomes easier to distinguish from the background by making the window size smaller than at the long distance. Therefore, the distance accuracy is improved and reliability can be ensured, and the processing time and circuit scale can be reduced.

  In the obstacle measuring method, it is preferable to use the SAD method (difference absolute value sum) or the POC method (phase-only correlation method) as a correlation calculation method for calculating parallax in the low-resolution image. In addition, it is preferable to use a POC method (phase-only correlation method) as a correlation calculation method for calculating parallax in the high-resolution image.

  When the distance to the obstacle determined by the low-resolution image is a short distance, by setting the window size smaller than in the case of a long distance, it becomes easier to distinguish from the background, the influence of occlusion with the background Can be suppressed.

  A second obstacle measurement method according to the present invention is an obstacle measurement method for measuring a distance to an obstacle by a parallax calculation using a stereo image, and sets a window size of corresponding point search for the parallax calculation at a lower part of the image. It is characterized by being set smaller than the upper part of the image.

  According to this obstacle measurement method, when shooting forward with an in-vehicle camera while traveling, there are mainly long-distance objects at the top of the image and short-distance objects at the bottom of the image. By making the window size at the bottom of the image smaller than the corresponding point search window size, it becomes easier to distinguish the object from the background when the object is a short distance, thus improving the distance accuracy and ensuring reliability. Thus, the processing time can be shortened and the circuit scale can be reduced.

  A third obstacle measurement method according to the present invention is an obstacle measurement method for measuring a distance to an obstacle by parallax calculation using a stereo image, and a window size for searching for corresponding points for parallax calculation in an image peripheral portion. Is set smaller than the center of the image.

  According to this obstacle measurement method, when shooting forward with a vehicle-mounted camera while traveling, the long-distance objects are concentrated in the image center, and the short-distance objects are concentrated in the image periphery. By making the window size at the periphery of the image smaller than the window size of the corresponding point search, it becomes easier to distinguish the object from the background when the object is a short distance, thus improving the distance accuracy and ensuring reliability. In addition, the processing time can be shortened and the circuit scale can be reduced.

  In the second and third obstacle measuring methods, it is preferable that the upper part of the image at a certain position or more is not subject to the parallax calculation. An image that does not need to be detected as an obstacle such as a background (mountain, sky, etc.) appears in the upper part of the image above a certain position, but since the parallax is not calculated in the upper part of the image, unnecessary calculations are effectively suppressed. it can.

  A first obstacle measuring apparatus according to the present invention is an obstacle measuring apparatus that measures a distance to an obstacle by parallax calculation using a stereo image, and a plurality of cameras that acquire images, and at least different resolutions from the images. An image generation unit that generates a high-resolution image and a low-resolution image, a first distance information generation unit that generates first distance information from the low-resolution image by parallax calculation, and a second by parallax calculation from the high-resolution image A second distance information generation unit that generates the distance information of the corresponding point, and sets a window size of the corresponding point search for the parallax calculation by the high-resolution image based on the first distance information. And

  According to this obstacle measuring apparatus, a corresponding point search window for calculating distance information to an obstacle in a high resolution image based on the approximate parallax (distance information) obtained from the low resolution image (first distance information). The size can be determined. For example, when the object is at a short distance, the influence of the occlusion with the background becomes large near the edge where the feature is easy to take, but it becomes easier to distinguish from the background by making the window size smaller than at the long distance. Therefore, the distance accuracy is improved and reliability can be ensured, and the processing time and circuit scale can be reduced.

  In the obstacle measuring apparatus, it is preferable that the first distance information generation unit uses an SAD method (sum of absolute differences) or a POC method (phase-only correlation method) as the parallax calculation correlation calculation method. The second distance information generation unit preferably uses a POC method (phase-only correlation method) as a correlation calculation method for the parallax calculation.

  A second obstacle measurement device according to the present invention is an obstacle measurement device that measures a distance to an obstacle by parallax calculation using a stereo image, and a plurality of cameras that acquire images and a distance by parallax calculation from the image. A distance information generation unit that generates information, and when generating the distance information, a window size of the corresponding point search for the parallax calculation in the lower part of the image is set smaller than that in the upper part of the image.

  According to this obstacle measuring device, when a forward image is taken with a vehicle-mounted camera while traveling, there are mainly long-distance objects at the top of the image and short-distance objects at the bottom of the image. By making the window size at the bottom of the image smaller than the corresponding point search window size, it becomes easier to distinguish the object from the background when the object is a short distance, thus improving the distance accuracy and ensuring reliability. The processing time and circuit scale can be reduced.

  A third obstacle measuring device according to the present invention is an obstacle measuring device that measures a distance to an obstacle by parallax calculation using a stereo image, and a plurality of cameras that acquire images and a distance by parallax calculation from the image. A distance information generation unit that generates information, and when generating the distance information, a window size of the corresponding point search for the parallax calculation in the image peripheral part is set smaller than the image center part .

  According to this obstacle measuring device, when shooting forward with a vehicle-mounted camera while traveling, the long-distance objects are concentrated in the image center, and the short-distance objects are concentrated in the image periphery. By making the window size at the periphery of the image smaller than the window size of the corresponding point search, it becomes easier to distinguish the object from the background when the object is a short distance, thus improving the distance accuracy and ensuring reliability. In addition, the processing time can be shortened and the circuit scale can be reduced.

  In the obstacle measuring apparatus, it is preferable that the distance information generation unit does not set the parallax calculation target for the upper part of the image at a certain position or more. An image that does not need to be detected as an obstacle such as a background (mountain, sky, etc.) appears in the upper part of the image above a certain position, but since the parallax is not calculated in the upper part of the image, unnecessary calculations are effectively suppressed. it can.

  The obstacle measurement system according to the present invention includes the above-described first, second, or third obstacle measurement device, a determination unit that determines a danger related to the obstacle based on the measured distance information, and the danger. A warning unit that issues a warning when it is determined.

  According to this obstacle measurement system, attention is given to danger by warning from the warning section, and it is possible to perform danger avoidance operations, improve distance accuracy and ensure reliability in the obstacle measurement system, calculation time And circuit scale can be reduced.

  According to the obstacle measurement method, the obstacle measurement device, and the obstacle measurement system of the present invention, the distance accuracy is ensured and the processing time is set by setting the window size for searching for corresponding points according to the distance to the obstacle. And shortening the circuit scale can be achieved.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

<First Embodiment>
FIG. 1 is a block diagram showing a schematic configuration of an obstacle measuring apparatus according to the first embodiment. FIG. 2 is a schematic diagram (a) showing a left image, a schematic diagram (b) showing a right image, and a search position and a correlation value in order to explain corresponding point search for parallax calculation in the obstacle measuring apparatus of FIG. It is a figure (c) which shows the example of these roughly.

  As shown in FIG. 1, the vehicle obstacle measurement device 10 includes a right camera 11, a left camera 12, an image correction unit 13 that corrects images from the cameras 11 and 12, and images from the cameras 11 and 12. Similarly, a multi-resolution image generation unit 14 that generates images of a plurality of resolutions, a first distance information generation unit 15 that generates distance information / distance images based on the low-resolution image generated by the multi-resolution image generation unit 14, Disparity in the second distance information generation unit 16 based on the distance information obtained by the second distance information generation unit 16 that generates the distance information / distance image based on the high-resolution image and the first distance information generation unit 15 And a window size setting unit 17 that sets a window size of corresponding point search for calculation, and is mounted on the vehicle.

  Each of the right camera 11 and the left camera 12 includes an optical system such as a lens and an image pickup device such as a CCD or a CMOS image sensor. The right camera 11 and the left camera 12 are arranged at a predetermined interval in the front part of the vehicle to form a stereo camera, Image information of existing obstacles such as preceding vehicles and pedestrians is obtained as an electrical signal.

  The image correction unit 13 corrects distortion or the like generated in the periphery of the lenses of the cameras 11 and 12 with respect to the images from the cameras 11 and 12.

  As shown in FIGS. 2A and 2B, a left image (reference image) obtained from the left camera 12 and a right image (reference image) obtained from the right camera 11 for the preceding vehicle P, which is the same obstacle, are obtained. Although the image is displayed on the imaging surface while being shifted in the horizontal direction, the amount of deviation of the object on these two imaging surfaces is parallax. In the first and second distance information generating units 15 and 16, a window W having a predetermined size is set in the left image of the preceding vehicle P, and a window W ′ having the same size is set in the right image, as shown in FIG. As shown in FIG. 2, parallax can be obtained by searching a region having a high correlation in the windows W and W ′ of two images and obtaining a search position having the largest correlation value as shown in FIG. A distance image can be generated by three-dimensionally calculating the distance to the obstacle based on the parallax. That is, since the parallax and the distance are in an inversely proportional relationship, the distance can be measured by measuring the parallax.

  The first distance information generation unit 15 obtains approximate parallax (distance information) from the low-resolution image by using a correlation method based on a sum of absolute differences (SAD) as a parallax calculation method. The following POC method may be used as a parallax calculation method in the first distance information generation unit 15.

  The second distance information generation unit 16 obtains parallax from a high-resolution image using a phase-only correlation (POC) as a parallax calculation method. In this case, the window size setting unit 17 sets the window size of the corresponding point search for parallax calculation in the high resolution image smaller than the far distance at a short distance. Whether or not the obstacle is a short distance is determined based on the approximate parallax obtained by the first distance information generation unit 15.

  The SAD method is a commonly used parallax measurement method. For example, a paper by Motoki Arai, Kazuhiko Sumi, and Takashi Matsuyama “Optimization of correlation function and sub-pixel estimation method in image block matching” ( Information Processing Society of Japan Research Report CVIM-144-5, pp.33-40, May 2004) and Japanese Patent Application Laid-Open No. 2000-283753 can be referred to.

  The POC method (phase-only correlation method) uses phase information obtained by performing Fourier transform on an image signal converted into a digital signal to obtain a phase signal and an amplitude signal, and performing inverse Fourier transform on only the phase signal. Since this is a dimension measurement method and uses only phase components, highly accurate matching is possible, and the accuracy of distance measurement is higher than that of the SAD method. The POC method is, for example, a paper by Yasuhiro Fukuda, Masahide Abe, and Masayuki Kawamata “Realization of video image misregistration using phase-only correlation by DSP” (Tohoku Section of the Society of Instrument and Control Engineers 215th Meeting (2004.5.27 ) You can refer to document number 215-3).

  Each of the distance information generation units 15 and 16 can process the calculation for the distance calculation by the above-described SAD method and POC method by hardware using an integrated element or the like, but performs the processing by software by a CPU (Central Processing Unit). It may be. In this case, the CPU executes a predetermined calculation according to a predetermined program.

  FIG. 3 is a schematic diagram showing a camera image photographed by the camera 11 (or 12) of the obstacle measuring apparatus 10 of FIG. 1, and a schematic diagram showing a low-resolution distance image obtained from the image of FIG. Schematic diagram (c) showing the window size of the corresponding point search set for each image in FIGS. (B) and (a), and a graph (d) showing changing the window size with distance th as the boundary It is.

  While the vehicle equipped with the obstacle measuring device 10 of FIG. 1 is traveling, the vehicles 2 and 3 ahead on the road 1 are photographed by the camera 12, for example, as shown in the screen 9 of FIG. The window size setting unit 17 sets the window size for the corresponding point search based on the distance information (see the distance image in FIG. 3B) generated by the first distance information generation unit 15 for each of the images.

  That is, for example, as shown in FIG. 3D, if the distance to the vehicle 3 in FIG. If the size of the window W3 with respect to the car 3 is set to N, and the distance to the vehicle 2 in FIG. (<N) is set. In this way, as shown in FIG. 3C, the size M of the window W2 for the vehicle 3 at a short distance is set smaller than the size N of the window W3 at a long distance.

  Next, the overall operation (steps S01 to S06) of the obstacle measuring apparatus 10 of FIG. 1 will be described with reference to the flowchart of FIG.

  First, when images captured by the cameras 11 and 12, for example, images of the vehicles 2 and 3 running on the road 1 in FIG. 3A are input (S01), the image correction unit 13 corrects the images. (S02) The multi-resolution image generation unit 14 generates images of a plurality of resolutions (S03).

  Next, the first distance information generation unit 15 generates distance information (distance image) based on the low-resolution image among the above-described plurality of resolution images (S04), and is calculated from this low-resolution image. Based on the distance information, the window size setting unit 17 sets the window size for the corresponding point search (S05).

  Next, distance information (distance image) is similarly generated by the second distance information generation unit 16 based on the high-resolution image (S06). The window size for corresponding point search at this time is set in step S05. The distance information obtained in step S06 is output from the obstacle measuring device 10 in FIG.

  Next, window size setting steps S11 to S13 in step S05 of FIG. 4 will be described with reference to the flowchart of FIG.

  If the distance calculated based on the low-resolution image by the first distance information generation unit 15 is equal to or greater than the threshold th for the vehicle 3 in FIG. 3A, for example (S11), FIG. ), The window size for the corresponding point search is set to N by the window size setting unit 17 (S12).

  If the distance calculated based on the low-resolution image is, for example, the threshold value th or less for the vehicle 2 in FIG. 3A and a short distance (S11), the window size setting is performed as shown in FIG. The window size for the corresponding point search is set to M in the unit 17 (S13).

  As described above, for the long-distance vehicle 3 in FIG. 3A, the size of the window W3 is set to N × N as shown in FIG. 3C, and for the short-distance vehicle 2 the window W2 After the size is set to M × M (M <N), in step S06 in FIG. 4, a corresponding point search is performed with each window size to generate distance information.

  As described above, the distance information output from the obstacle measuring apparatus 10 of FIG. 1 can be generated. Conventionally, if the target object is a short distance, the occlusion with the background is near the edge where the feature is easy to be taken. However, according to the present embodiment, the window size for searching for corresponding points for short-distance objects is long distance. By setting it smaller, it becomes easier to distinguish from the background. Therefore, the distance accuracy can be improved and the reliability can be ensured, the processing time in the second distance information generation unit 16 can be shortened, and the second distance can be secured. The circuit scale of the integrated element of the information generation unit 16 can be reduced.

<Second Embodiment>
FIG. 6 is a block diagram showing a schematic configuration of an obstacle measuring apparatus according to the second embodiment.

  As shown in FIG. 6, the obstacle measurement device 20 for a vehicle includes a right camera 11, a left camera 12, an image correction unit 13 that corrects images from the cameras 11 and 12, and images by the cameras 11 and 12. A distance information generation unit 18 that generates distance information and a distance image based on the distance information and is mounted on a vehicle. The right camera 11, the left camera 12, and the image correction unit 13 have the same configuration as that in FIG.

  The distance information generation unit 18 calculates the distance to the obstacle three-dimensionally based on the horizontal position shift (parallax) between the right image and the left image from the cameras 11 and 12 as in FIG. Is generated. Note that the distance information generation unit 18 can be processed in hardware or software in the same manner as the distance information generation units 15 and 16 using the POC method as a parallax calculation method.

  FIG. 7 is a schematic diagram showing a camera image taken in front by the camera of the obstacle measuring device 20 of FIG. 6 and shows a window size of corresponding point search set for each image of FIG. 7A. It is a schematic diagram (b).

  The obstacle measuring device 20 shown in FIG. 6 is mounted on a vehicle, and while traveling on the road 1 as shown in FIG. A few camera images are obtained. When the distance information generating unit 18 generates distance information from the images of the vehicles 2 and 3 in FIG. 7A, it is assumed that the vehicle 2 in the image lower part 9a of the screen 9 is at a short distance, and the image upper part of the screen 9 It is assumed that the vehicle 3 in 9b is at a long distance. Then, as shown in FIG. 7B, the size of the window W2 for searching for corresponding points for the vehicle 2 in the lower image portion 9a is set smaller than the window W3 for searching for corresponding points for the vehicle 3 in the upper image portion 9b. That is, similarly to FIGS. 3C and 3D, the window W2 is set to size M in the lower image portion 9a, and the window W3 is set to size N (> M) in the upper image portion 9b.

  As described above, in the case of shooting forward with a vehicle-mounted camera while traveling, since there are mainly long-distance objects in the upper image portion 9b and short-distance objects in the lower image portion 9a, search for corresponding points in the upper image portion 9b. By making the size of the window W2 in the lower part 9a of the image lower than the size of the window W3, it becomes easier to distinguish the object from the background when the object is a short distance, thereby improving the distance accuracy and ensuring reliability. In addition, the processing time in the distance information generator 18 and the circuit scale can be reduced.

  Next, a modification of the obstacle measuring apparatus 20 in FIG. 6 will be described with reference to FIG. FIG. 8 is a schematic view similar to FIG. 7A showing a camera image taken in front by the camera of the obstacle measuring device 20 of FIG.

  As shown in FIG. 8, when the distance information generating unit 18 generates the distance information from the images of the vehicles 2 and 3 in the screen 9, it is assumed that the vehicle 2 in the image peripheral part 9c of the screen 9 is at a short distance, It is assumed that the vehicle 3 in the image center part 9d is at a long distance, and further, the image in the image upper part 9e is not subject to parallax calculation. Similarly to FIG. 7B, the size of the window W2 for searching for the corresponding point for the vehicle 2 in the image peripheral portion 9c is set smaller than the window W3 for searching for the corresponding point for the vehicle 3 in the image center portion 9d. To do. For example, as in FIGS. 3C and 3D, the window W2 is set to size M in the image peripheral portion 9c, and the window W3 is set to size N (> M) in the image center portion 9d.

  As described above, in the case of shooting forward with the in-vehicle camera while traveling, the long-distance objects are concentrated on the image center portion 9d, and the short-distance objects are concentrated on the image periphery portion 9c. By making the size of the window W2 of the image peripheral portion 9c smaller than the size of the window W3 for point search, it becomes easy to distinguish the object from the background when the object is a short distance, thereby improving distance accuracy and reliability. In addition, the processing time in the distance information generation unit 18 and the circuit scale can be reduced.

  Also, unlike the upper image portion 9e in FIG. 8, the upper image portion above a certain position is not subject to parallax calculation, but the upper image portion above a certain position is detected as an obstacle such as a background (mountain, sky, etc.). Although an unnecessary video is shown, parallax calculation is not performed on the upper part of the image, so that unnecessary calculation in the distance information generation unit 18 can be effectively suppressed.

<Third Embodiment>
FIG. 9 is a block diagram schematically showing an obstacle measurement system for a vehicle according to a third embodiment.

  The obstacle measurement system 30 of FIG. 9 is based on the distance information from the obstacle information measuring unit 10 or 20 of FIG. 1 or FIG. When the distance between the determination unit 31 and the preceding vehicle 2 is equal to or less than the set distance, a warning sound, a warning sound, a warning lamp, etc. And a warning unit 32 for informing the danger.

  According to the obstacle measurement system 30 in FIG. 9, when the distance to the preceding vehicle 2 is less than the set distance and the vehicle approaches, the operation is performed by a warning from the warning unit 32 based on the distance information from the obstacle measurement devices 10 and 20. The person can pay attention to the danger and can perform the operation of avoiding the danger, and the distance accuracy can be improved and the reliability can be secured, so that the calculation time can be shortened and the circuit scale can be reduced.

  As described above, the best mode for carrying out the present invention has been described. However, the present invention is not limited to these, and various modifications are possible within the scope of the technical idea of the present invention. For example, the obstacle measurement devices 10 and 20 in FIGS. 1 and 6 or the obstacle measurement system 30 in FIG. 9 can be used by being mounted on a self-propelled vehicle such as an automobile or a train, but is not limited thereto. For example, it may of course be mounted on a robot or the like.

  In the obstacle measurement system 30 of FIG. 9, a warning is issued when the distance to the preceding vehicle 2 is equal to or less than the set distance. For example, the vehicle is controlled so as to automatically decelerate. May be.

It is a block diagram showing a schematic structure of an obstacle measuring device by a 1st embodiment. Schematic diagram (a) showing a left image, schematic diagram (b) showing a right image, and an example of a relationship between a search position and a correlation value in order to explain corresponding point search for parallax calculation in the obstacle measurement apparatus of FIG. (C) which shows schematically. . FIG. 1A is a schematic diagram showing a camera image taken by the camera 11 (or 12) of the obstacle measuring apparatus 10 in FIG. 1, and FIG. 3B is a schematic diagram showing a low-resolution distance image obtained from the image in FIG. 4A is a schematic diagram showing a corresponding point search window size set for each image in FIG. 3A, and FIG. 4D is a graph showing that the window size is changed with the distance th as a boundary. It is a flowchart for demonstrating the whole operation | movement (step S01-S06) of the obstacle measuring device 10 of FIG. It is a flowchart for demonstrating step S11-S13 of the window size setting in step S05 of FIG. It is a block diagram which shows schematic structure of the obstacle measuring device by 2nd Embodiment. Schematic diagram (a) showing a camera image taken in front by the camera of the obstacle measuring apparatus 20 in FIG. 6 and schematic diagram (b) showing the window size of corresponding point search set for each image in FIG. 7 (a). ). It is a figure for demonstrating the modification of 2nd Embodiment, and is a schematic diagram similar to Fig.7 (a) which shows the camera image image | photographed the front with the camera of the obstacle measuring device 20 of FIG. It is a block diagram which shows roughly the obstacle measuring system for vehicles of 3rd Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10,20 Obstacle measuring device 11 Right camera 12 Left camera 14 Multi-resolution image generation part 15 1st distance information generation part 16 2nd distance information generation part 17 Window size setting part 9 Screen 9a Image lower part 9b Image upper part 9c Image Peripheral part 9d Image center part 9e Upper part of image 30 Obstacle measurement system W2, W3 Window

Claims (14)

An obstacle measurement method for measuring the distance to an obstacle by calculating parallax using a stereo image,
An obstacle measurement method characterized by setting a window size for searching for corresponding points for parallax calculation in a high-resolution image based on distance information obtained from a low-resolution image.   The obstacle measuring method according to claim 1, wherein an SAD method (sum of absolute differences) or a POC method (phase-only correlation method) is used as a correlation calculation method for calculating parallax in the low-resolution image.   The obstacle measurement method according to claim 1, wherein a POC method (phase-only correlation method) is used as a correlation calculation method for calculating parallax in the high-resolution image.   The obstacle measurement method according to claim 1, 2, or 3, wherein when the distance to the obstacle obtained from the low-resolution image is a short distance, the window size is set smaller than that for a long distance. An obstacle measurement method for measuring the distance to an obstacle by calculating parallax using a stereo image,
An obstacle measurement method, wherein a window size for searching for a corresponding point for calculating the parallax in the lower part of the image is set smaller than that in the upper part of the image. An obstacle measurement method for measuring the distance to an obstacle by calculating parallax using a stereo image,
An obstacle measurement method, wherein a window size for searching for corresponding points for calculating parallax in an image peripheral portion is set smaller than that in an image central portion.   The obstacle measurement method according to claim 5 or 6, wherein an upper part of an image above a certain position is not subject to the parallax calculation. An obstacle measurement device that measures the distance to an obstacle by calculating parallax using a stereo image,
Multiple cameras that capture images,
An image generator that generates at least a high-resolution image and a low-resolution image having different resolutions from the image;
A first distance information generation unit that generates first distance information from the low-resolution image by parallax calculation;
A second distance information generation unit that generates second distance information by calculating parallax from the high-resolution image,
An obstacle measuring apparatus, wherein a window size for searching for a corresponding point for calculating the parallax using the high-resolution image is set based on the first distance information.   The obstacle measurement apparatus according to claim 8, wherein the first distance information generation unit uses a SAD method (sum of absolute differences) or a POC method (phase-only correlation method) as the correlation calculation method of the parallax calculation.   The obstacle measurement device according to claim 8 or 9, wherein the second distance information generation unit uses a POC method (phase-only correlation method) as a correlation calculation method of the parallax calculation. An obstacle measurement device that measures the distance to an obstacle by calculating parallax using a stereo image,
Multiple cameras that capture images,
A distance information generation unit that generates distance information by calculating parallax from the image,
An obstacle measuring device characterized in that, when generating the distance information, a window size for searching for corresponding points for calculating the parallax at the lower part of the image is set smaller than that at the upper part of the image. An obstacle measurement device that measures the distance to an obstacle by calculating parallax using a stereo image,
Multiple cameras that capture images,
A distance information generation unit that generates distance information by calculating parallax from the image,
An obstacle measuring device characterized in that, when generating the distance information, a window size for searching for corresponding points for parallax calculation in an image peripheral portion is set smaller than that in an image central portion.   The obstacle measurement device according to claim 11 or 12, wherein the distance information generation unit does not subject the upper part of the image above a certain position to the parallax calculation target.   An obstacle measurement device according to any one of claims 8 to 13, a determination unit that determines a risk related to the obstacle based on the measured distance information, and a warning when the risk is determined. An obstacle measuring system comprising: a warning unit that emits.

Images