JP2014006179A - Stereo camera calibrating apparatus and method, and distance measuring apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance, in a distance measuring apparatus using a stereo camera, the accuracy of correcting parallax detecting errors of the stereo camera and thereby improve the accuracy of distance measurement.SOLUTION: Distances between a stereo camera and an object for correction use are set according to a plurality of available designed parallax values, and the object for correction use is shot with the stereo camera at different distances to detect a real parallax value for each horizontal coordinate value. Then, for each of the plurality of designed parallax values, a real measured parallax value table indicating the designed parallax values and real parallax values corresponding to the horizontal coordinate values is generated, and from these real measured parallax value tables a parallax correction table indicating the real parallax values and parallax correction quantities corresponding to the horizontal coordinate values is generated for each of the multiple real parallax values corresponding to the plurality of designed parallax values.

Description

  The present invention relates to a stereo camera calibration apparatus and method for correcting a parallax detection error of a stereo camera, and a distance measuring apparatus using the stereo camera.

  In recent years, a stereo camera consisting of a set of two cameras is mounted on a moving body such as an automobile, the front of the moving body is photographed by the stereo camera, and the distance from the captured stereo image to the object is measured. In response to the measurement results, a driving support system that warns the driver to prevent a collision and controls brakes, steering, etc. to maintain the inter-vehicle distance has been put into practical use.

  In such a driving support system, corresponding points are searched by block matching using one stereo image obtained by photographing the same object with a stereo camera including two cameras as a reference image and the other as a reference image. The amount of point displacement (parallax) is detected, and the distance to the object is calculated based on the principle of triangulation.

  FIG. 10 is a diagram for explaining the principle of distance measurement using a stereo camera. The stereo camera is configured by arranging two cameras 11 and 12 parallel to the optical axis. Here, the distance Z to the object is determined from the distance (base line length) B between the optical axes of the cameras 11 and 12, the focal length f, and the parallax d.

として計算される。したがって、基線長Ｂと焦点距離ｆが既知ならば、視差ｄから距離Ｚを求めることができる。

Is calculated as Therefore, if the baseline length B and the focal length f are known, the distance Z can be obtained from the parallax d.

  In distance measurement using a stereo camera, the accuracy of the distance Z depends on the parallax d. For this reason, it is ideal not to depend on a positional shift other than the original parallax that occurs according to the distance between two stereo images obtained by photographing with a stereo camera. There is misalignment due to mechanical distortion. If the parallax is not correctly detected due to a positional shift caused by optical distortion of the lens, the ranging accuracy is lowered, and the reliability and safety of the driving support system are impaired.

  Therefore, in Patent Document 1, for example, the ratio of the minute actual field angle for each predetermined region of the comparison image and the designed ideal field angle is measured and stored in advance as a field angle correction ratio table and imaged with a stereo camera. Proposes to improve the distance measurement accuracy by correcting the parallax data detected by matching corresponding points from the reference image and comparison image using the angle of view correction ratio table to reduce the effects of lens distortion and assembly errors. doing.

  FIG. 11 is a diagram showing the positional relationship between the optical system (lens) of the reference image camera and the comparative image camera and the subject. As shown in FIG. 11, subjects a and b at different distances in the same direction with respect to the comparative image optical system are imaged on Xa ′ and Xb ′ on the image sensor (sensor) through the comparative image optical system. Xa ′ = Xb ′. However, in the reference image optical system, since the subjects a and b are incident from different directions, the images are formed at different positions on the sensor as indicated by Xa and Xb in the drawing. Thus, the image formation position on the reference image camera is not determined only from the image formation position of the comparative image camera.

  In the prior art of Patent Document 1, the image formation positions Xa ′ and Xb ′ of the comparative image are corrected using an angle-of-view correction ratio table. In this case, the same shift correction amount can be obtained for the subjects a and b. However, since the parallax is obtained from the difference in the imaging positions of the reference image camera and the comparative image camera (for example, Xa′−Xa for the subject a), the deviation of the reference image Xa and Xb is also Xa ′, The parallax value is affected similarly to Xb ′.

  In the case of the prior art described in Patent Document 1, a subject that does not match the distance of the data measurement screen used when creating the angle-of-view correction ratio table is affected by the parallax of the reference image optical system and the like.

  It is an object of the present invention to increase the accuracy of correcting parallax detection errors of two cameras constituting a stereo camera and improve the distance measurement accuracy.

  The present invention provides a stereo camera calibration device that generates a correction value for correcting a parallax detection error of a stereo camera including a set of two cameras, for each of a plurality of actual parallax values corresponding to a plurality of design parallax values. The correction value of the actual parallax value corresponding to the actual parallax value and the horizontal coordinate value is generated.

  In the stereo camera calibration device according to an embodiment, a unit that sets a plurality of distances between the stereo camera and a correction subject in association with a plurality of design parallax values, and the stereo camera at each set distance. From a stereo image acquired by photographing the correction subject, a means for detecting an actual parallax value for each horizontal coordinate value, and a plurality of design parallax values based on the detected actual parallax value, Means for generating an actual parallax measurement value table indicating actual parallax values corresponding to the design parallax values and horizontal coordinate values; and a plurality of real parallax values corresponding to a plurality of design parallax values from the actual parallax value measurement value table. And a means for generating a parallax correction table indicating correction values of the real parallax values corresponding to the real parallax values and the horizontal coordinate values.

  According to the present invention, since the actual parallax value is detected for each distance and horizontal coordinate value between the subject and the stereo camera and the correction amount is calculated, the correction accuracy of the parallax detection error of the stereo camera is improved and the distance measurement accuracy is improved. be able to.

1 is an overall block diagram of a calibration apparatus according to an embodiment of the present invention. It is explanatory drawing of the calibration of this embodiment. It is a block diagram of a real parallax measurement value table. It is a block diagram of a parallax correction table. It is a processing flowchart of real parallax measurement value table generation. It is explanatory drawing which produces | generates a parallax correction table from a real parallax measurement value table. It is a processing flowchart of parallax correction table generation. It is a block diagram of the distance measuring device concerning the embodiment of the present invention. It is a processing flowchart of distance measurement. It is explanatory drawing which calculates distance from parallax. It is a figure explaining the problem of a prior art.

  Embodiments of the present invention will be described below with reference to the drawings. In the embodiment, the range of possible parallax values is 0 to 64, but of course not limited to this.

<Calibration>
FIG. 1 is an overall functional block diagram of the calibration apparatus according to the present embodiment. The calibration apparatus includes a calibration chart 100 as a correction subject, a stereo camera 10 including a left camera 11 and a right camera 12, a parallax detection unit 20 that detects parallax of a stereo image captured by the stereo camera 10, and a design. An actual parallax measurement value table generation unit 30 that generates an actual parallax measurement value table 42 indicating the correspondence between the upper parallax value (design parallax value) and the actually measured parallax value (real parallax value), and the generated actual difference measurement value table The parallax correction table 62 for correcting the parallax value obtained by the stereo camera camera 10 actually shooting the subject is generated from the memory (1) 40 that temporarily stores 42 and the actual parallax measurement value table 42. A parallax correction table generation unit 50, a memory (2) 60 that finally stores the generated parallax correction table 62, and controls the overall operation That the control unit 90 and,, and the like distance control mechanism 95 for controlling the distance between the stereo camera 10 and the calibration chart 100.

  Here, each part except the stereo camera 10, the distance control mechanism 95, and the calibration chart 100 in FIG. 1 is generally configured by a computer device. In other words, the parallax detection unit 20, the actual parallax measurement value table generation unit 30, the parallax correction table generation unit 50, and the control unit 90 realize each processing function based on the CPU, a program, and the like. The memory (1) 40 is a volatile memory such as a RAM, and in addition to the actual parallax measurement value table 42, data and the like being processed by the CPU are also stored. The memory (2) 60 is a non-volatile memory such as EPROM or EPPROM. In addition to the parallax correction table 62, the memory (2) 60 stores programs executed by the CPU, parameters necessary for processing by the CPU, and the like. Is done.

  FIG. 2 is a diagram for explaining how the distance between the stereo camera 10 and the calibration chart 100 is set during calibration. As shown in FIG. 2, in this embodiment, at the time of calibration, the design parallax value is set to d, and d is changed to 64, 63,..., I,. .., Bf / i,..., And Bf / i,..., Bf / i,..., And a calibration chart photographed by the stereo camera 10 for each distance. The parallax for each horizontal coordinate value (horizontal pixel position) of 100 images is measured. Then, for each horizontal coordinate value, an actual parallax measurement value table 42 indicating the correspondence between the designed parallax value (design parallax value) and the actually measured parallax value (actual parallax value) is generated. Based on the above, the parallax correction table 62 for correcting the parallax value obtained by the stereo camera 10 actually photographing the subject is generated.

  FIG. 3 is a diagram showing a configuration example of the actual parallax measurement value table 42. As shown in FIG. 3, the actual parallax measurement value table 42 is a two-dimensional table of design value parallax (design parallax value) d and x-coordinate (horizontal coordinate value x) of the captured image. The parallax value (actual parallax value) actually measured by the calibration is represented by a two-dimensional array m (x, d). Here, m (1,64), m (2,64),... Is a design parallax value d = 64, and the distance Z between the stereo camera 10 and the calibration chart 100 is Z = Bf / 64. The actual parallax values at positions where the horizontal coordinate value x is 1, 2,. Similarly, when m (1, i), m (2, i),... Are set as the design parallax value d = i and the distance Z between the stereo camera 10 and the calibration chart 100 is Z = Bf / i, The actual parallax values at the positions where the coordinate value x is 1, 2,.

  FIG. 4 is a diagram illustrating a configuration example of the parallax correction table 62. As shown in FIG. 4, the parallax correction table 62 is a two-dimensional table of the actual parallax value d ′ and the x coordinate (horizontal coordinate value x) of the captured image. A correction value (parallax correction amount) required when correcting the parallax value (actual parallax value) actually obtained by photographing the subject with the stereo camera 10 is represented by a two-dimensional array Δd (x, d ′). Here, Δd (1, 64), Δd (2, 64),... Are designed in each position at the actual parallax value d '= 64 and the horizontal coordinate value x is 1, 2,. The difference from the parallax value is shown. Similarly, Δd (1, i), Δd (2, i),... Is a design at each position where the actual parallax value d '= i and the horizontal coordinate value x is 1, 2,. The difference from the parallax value is shown. Therefore, by subtracting Δd from the actual parallax value according to the horizontal coordinate value, the original parallax that occurs according to the distance can be obtained.

  Returning to FIG. 1, the operation of the calibration apparatus will be described in detail. The operation is roughly divided into a generation process of the actual parallax measurement value table 42 in FIG. 3 and a generation process of the parallax correction table 62 in FIG.

  First, generation processing of the actual parallax measurement value table 42 will be described in detail. FIG. 5 is a process flowchart of the actual parallax measurement value table generation.

  First, the control unit 90 sets a design parallax value (set parallax value) d to a maximum value dmax (here, dmax = 64) that can be taken (step 101), and the stereo camera 10 and the calibration chart 100 are set. The distance Z is calculated (step 102). As described above, the distance Z is calculated as Z = Bf / d. B is the baseline length of the cameras 11 and 12 constituting the stereo camera 10, and f is the focal length.

  The control unit 90 sends a design parallax value (design parallax value) d to the actual parallax measurement value table generation unit 30 and sends the calculated distance value Z to the distance control mechanism 95.

  The distance control mechanism 95 moves the calibration chart 100 so that the distance to the stereo camera 10 becomes Z based on the distance value Z from the control unit 90 (step 103). This is performed, for example, as follows. A laser distance meter is attached to the calibration chart 100, and the distance between the calibration chart 100 and the stereo camera 10 is measured with the laser distance meter. The distance control mechanism 95 receives the distance measurement value from the laser distance meter and drives the calibration chart 100 so that the distance measurement value matches the distance value Z sent from the control unit 90. Note that an encoder may be attached to a carriage or the like that moves the calibration chart 100, and the distance between the calibration chart 100 and the stereo camera 10 may be measured by the encoder.

  When the distance Z between the stereo camera 10 and the calibration chart 100 is set, the control unit 90 instructs the stereo camera 10 to start photographing. The left camera 11 and the right camera 12 of the stereo camera 10 are arranged with their optical axes parallel to each other with a predetermined interval (baseline length). The left camera 11 and the right camera 12 operate in synchronization with each other to photograph the calibration chart 100 of the correction subject from different viewpoints, convert the optical image into an analog electric signal, and further convert it into a digital signal. The image data is output after conversion (step 104). The image data is composed of, for example, 8 bits / pixel.

  The parallax detection unit 20 receives image data (stereo image data) output from the left camera 11 and the right camera 12 respectively, and performs parallax (actual) for each horizontal coordinate value (for each horizontal pixel) of the image as follows. Parallax value) is detected. Here, the image data of the left camera 11 is used as a reference image, and the image data of the right camera 12 is used as a comparison image.

  The parallax detection unit 20 first sets the horizontal coordinate value x of the reference image (image data of the left camera 11) to the minimum position xmin (step 105). Then, the parallax detection unit 20 searches for a corresponding point (corresponding point horizontal coordinate value) x ′ on the comparison image (image data of the right camera 12) corresponding to the horizontal coordinate value x (here, x = xmin) of the reference image. If a corresponding point is found (step 106), the difference (xx ′) between them, that is, the parallax (actual parallax value) d ′ is calculated (step 107). A well-known block matching method is used for the corresponding point search.

  The parallax detection unit 20 sends the horizontal coordinate value x and the calculated parallax (real parallax value) d ′ to the real parallax measurement value table generation unit 30. The design parallax value d (here, d = 64) is sent from the control unit 90 to the actual parallax measurement value table generation unit 30. The actual parallax measurement value table generation unit 30 stores the actual parallax measurement value table in the storage area where the column in the actual parallax measurement value table 42 of the memory (1) 40 is d (here, 64) and the row is x (here, x = xmin). The parallax value d ′ is written (step 108). That is, m (x, d) = d ′. The actual parallax measurement value table 42 in the memory (1) 40 is initially empty.

  Next, the parallax detection unit 20 sets the horizontal coordinate value x of the reference image (image data of the left camera 11) to (x + 1) (step 109). Then, it is determined whether the horizontal coordinate value x exceeds the maximum value xmax (step 110). If not, the process returns to step 106.

  Thereafter, by repeating the processing of steps 106 to 109 until the horizontal coordinate value x reaches the maximum value xmax, the design parallax value d (here, d =) is stored in the actual parallax measurement value table 42 of the memory (1) 40. 64), the actual parallax value d ′ for each horizontal coordinate value xmin,..., Xmax is written.

  If the horizontal coordinate value x exceeds the maximum value xmax, the control unit 90 sets the design parallax value d to (d-1) (step 111). That is, d = 63. Then, it is determined whether or not the design parallax value d is less than the minimum value dmin (step 112).

  Hereinafter, the design parallax value d in the actual parallax measurement value table 42 of the memory (1) 40 is changed to 64, 63,... By repeating the processing of steps 102 to 111 until the design parallax value d reaches the minimum value dmin. ..., The actual parallax value d ′ (d ′ = m (x, d)) for each horizontal coordinate value xmin, xmin,. That is, the actual parallax measurement value table 42 shown in FIG. 3 is generated.

  Next, the generation process of the parallax correction table 62 will be described in detail. The parallax correction table 62 is generated from the actual parallax measurement value table 52 by interpolation processing.

  FIG. 6 assumes the coordinate system of the design parallax d and the real parallax d ′, and the real parallax value d ′ (d ′ = m () at a certain horizontal coordinate value x for the real parallax measurement value table 52 obtained by calibration. It shows how x, d)) are plotted. According to FIG. 6, the correction value (parallax correction amount) Δd (x, i) for the actual parallax value i at a certain horizontal coordinate value x is

として算出される。ここで、ｉ＝０〜６４である。

Is calculated as Here, i = 0 to 64.

  FIG. 7 is a processing flowchart for generating a parallax correction table. This is performed by the parallax correction table generation unit 50 under the control of the control unit 90.

  The parallax correction table generation unit 50 first sets the horizontal coordinate value x to the minimum horizontal coordinate value xmin (step 201), and sets the minimum actual parallax value d'min that can take the actual parallax value d '(step 202). . Then, i is initialized to 0 (step 203).

The parallax correction table generation unit 50 uses the actual parallax measurement value table 42 in the memory (1) 40 to
m (x, i) <d ′ and d ′ ≦ m (x, i + 1)
Whether there is a set of real parallax values m (x, i) and m (x, x + 1) satisfying the condition (step 204). If not, i is set to i + 1 (step 205), and the process returns to step 204.

  If the combination of the actual parallax values m (x, i) and m (x, i + 1) satisfying the above conditions is found, the parallax correction table generation unit 50 calculates the correction value Δd (x, i) according to the above equation (2). ) Is calculated (step 206). Then, the parallax correction table generation unit 50 indicates that the column in the parallax correction table 62 of the memory (2) 60 is d ′ (here, d ′ = d′ min) and the row is x (here, x = xmin). In the storage area, the calculated correction value Δd (x, i) is written. Note that the parallax correction table 62 in the memory (2) 60 is initially empty.

  Next, the parallax correction table generation unit 50 sets the actual parallax value d ′ to (d ′ + 1) (step 207). Then, it is determined whether the actual parallax value d ′ exceeds the maximum possible parallax value d′ max (here, d′ max = 64) (step 208).

  Hereinafter, the parallax correction table generation unit 50 repeats the processing of steps 203 to 207 until the actual parallax value d ′ reaches the maximum parallax value d′ max, whereby the parallax correction table 62 in the memory (2) 60 has The actual parallax value d ′ corresponding to the horizontal coordinate value x (here, x = xmin) is a correction value (parallax correction amount) for each of d′ min,..., I,. Δd is written.

  If the actual parallax value d ′ exceeds the maximum parallax value d′ max (here, d′ max = 64), the parallax correction table generation unit 50 sets the horizontal coordinate value x to (x + 1) (step 209). . Then, it is determined whether the horizontal coordinate value x exceeds the maximum horizontal coordinate value xmax (step 210). If not, the process returns to step 202.

  Hereinafter, the parallax correction table generation unit 50 repeats the processing of steps 202 to 209 until the horizontal coordinate value x reaches the maximum horizontal coordinate value xmax, whereby the horizontal coordinate is stored in the parallax correction table 62 of the memory (2) 60. The actual parallax value d ′ corresponds to the correction value (parallax correction amount) Δd (x for each of d′ min,..., I,. , I) is written. That is, the parallax correction table 62 shown in FIG. 4 is completed.

  Here, the parallax correction table 62 is generated directly in the memory (2) 60, which is a non-volatile memory. However, the parallax using the memory (1) 40, which is a volatile memory, or other work memory is used. It is also possible to generate a correction table, transfer the generated parallax correction table to the memory (2) 60, and finally hold the parallax correction table 62 in the memory (2) 60.

<Distance measurement>
FIG. 8 is a functional block diagram of the distance measuring apparatus using the stereo camera according to the present embodiment. The distance measuring apparatus includes a stereo camera 10 including a left camera 11 and a right camera 12, a parallax detection unit 20 that detects parallax of a stereo image of a subject photographed by the stereo camera 10, and parallax correction generated by the previous calibration. A memory (2) 60 that holds a table 62, a parallax correction unit 70 that corrects parallax using the parallax correction table 62 held in the memory (2) 60, and calculates a distance to the subject using the corrected parallax And a distance calculation unit 80. In addition, although the control part 90 which controls the whole operation | movement shown in FIG. 1 is also provided, it abbreviate | omits in FIG.

  Actually, the distance measuring apparatus of FIG. 8 is configured integrally with the calibration apparatus of FIG. 1 and is used by being mounted on a moving body such as an automobile except for the distance control mechanism 95. That is, in FIG. 8, other than the stereo camera 10 is actually a computer device, and the parallax detection unit 20, the parallax correction unit 70, and the distance calculation unit 80 are the real parallax measurement value table generation unit 30 and the parallax correction table generation. Each processing function is realized together with the unit 50 and the like based on the CPU and the program. The control unit 90 operates each processing module of the parallax detection unit 20, the actual parallax measurement value table generation unit 30, and the parallax correction table generation unit 50 at the time of calibration, but measures the distance to the actual subject (front vehicle or the like). At this time, the processing modules of the parallax detection unit 20, the parallax correction unit 70, and the distance calculation unit 80 are operated.

FIG. 9 is a process flowchart of the distance measuring apparatus according to the present embodiment.
The left camera 11 and the right camera 12 of the stereo camera 10 operate in synchronization with each other to photograph a subject (such as a vehicle ahead) (step 301). The left camera 11 and the right camera 12 each convert the optical image of the subject into an analog electrical signal, further convert it into a digital signal, and output it as image data. The image data is composed of, for example, 8 bits / pixel.

  The parallax detection unit 20 receives image data output from the left camera 11 and the right camera 12 of the stereo camera 10 and detects parallax generated according to the distance to the subject. The parallax detection method is basically the same as that when the previous subject is the calibration chart 100. That is, the parallax detection unit 20 uses, for example, the image data of the left camera 11 as the reference image and the image data of the left camera 11 as the comparison image, and the corresponding point x ′ on the comparison image with respect to the horizontal position (horizontal coordinate value) x of the reference image. (Step 302), and the difference (x′−x) between the found corresponding points is calculated as a parallax (parallax value) d ′ (step 303). A well-known block matching method is used for the corresponding point search.

  The parallax detection unit 20 sends the horizontal coordinate value x and the calculated parallax value d ′ to the parallax correction unit 70.

  The parallax correction unit 70 acquires the correction value Δd from the parallax correction table 62 (FIG. 6) of the memory (2) 60 based on the horizontal coordinate value x and the parallax value d ′ sent from the parallax detection unit 20, The parallax value d ′ is corrected to obtain the original ideal parallax d generated according to the distance (step 304). That is, d = d′−Δd is calculated.

  Here, the parallax value d ′ obtained from a stereo image obtained by photographing an actual subject is not necessarily an integer 64, 63,. Rather, it is usually not an integer. In this case, the correction value Δd may be obtained by performing linear interpolation or the like. That is, a set of correction values Δd (x, i) and Δd (x, i + 1) of i <d ′ and d ′ <i + 1 is obtained from the parallax correction table 62 (FIG. 4), and these correction values are linearly complemented. The correction value Δd is obtained by calculation.

  The parallax correction unit 70 sends the corrected parallax value d to the distance calculation unit 80. The distance calculation unit 80 calculates the distance Z to the subject using the parallax value d sent from the parallax correction unit 70 (step 305). The distance Z is calculated as Z = Bf / d.

  The distance Z obtained by the distance measuring device is sent to a vehicle travel control unit (not shown), and control of the steering wheel and brake, warning to the driver, and the like are performed based on the distance.

<Development examples>
In the embodiment, the parallax correction table is a two-dimensional table of the horizontal coordinate value x and the actual parallax value d ′. However, the parallax offset may change depending on the vertical position. For example, in the case of a lens having barrel distortion, the distortion changes depending on the horizontal coordinate value x and the vertical coordinate value y. Therefore, the parallax correction table is a three-dimensional table of horizontal coordinate values x, vertical coordinate values y, and actual parallax values d ′. For this purpose, at the time of generating the actual parallax measurement value table, for each distance Z corresponding to possible dmin ≦ d <dmax, m (x, y, d) for each xmin ≦ x <xmax and ymin ≦ y <ymax. Should be calculated. By making the parallax correction table a three-dimensional table, parallax can be corrected for parallax offsets in the y direction (vertical position), and ranging accuracy is improved.

DESCRIPTION OF SYMBOLS 10 Stereo camera 20 Parallax detection part 30 Real parallax measurement value table production | generation part 40 Memory (1)
42 actual parallax measurement value table 50 parallax correction table generation unit 60 memory (2)
62 parallax correction table 70 parallax correction unit 80 distance calculation unit 90 control unit 95 distance control mechanism 100 calibration chart (correction subject)

Japanese Patent No. 3261115

Claims (7)

A stereo camera calibration device for generating a correction value for correcting a parallax detection error of a stereo camera composed of a pair of cameras,
A stereo camera calibration device that generates, for each of a plurality of real parallax values corresponding to a plurality of design parallax values, a correction value of the real parallax value corresponding to the real parallax value and a horizontal coordinate value. Means for setting a plurality of distances between the stereo camera and the correction subject in association with a plurality of design parallax values;
Means for detecting an actual parallax value for each horizontal coordinate value from a stereo image acquired by photographing the correction subject by the stereo camera at each set distance;
Means for generating an actual parallax measurement value table indicating an actual parallax value corresponding to the designed parallax value and a horizontal coordinate value for each of the plurality of designed parallax values based on the detected actual parallax value;
A parallax correction table indicating correction values of the actual parallax values corresponding to the real parallax values and the horizontal coordinate values is generated for each of the plurality of real parallax values corresponding to the plurality of design parallax values from the actual parallax value measurement value table. Means to
The stereo camera calibration apparatus according to claim 1, further comprising:   The parallax correction table is generated for each real parallax value and for each horizontal coordinate value by calculating a correction value by interpolating two real parallax values corresponding to two design parallax values near the real parallax value. The stereo camera calibration apparatus according to claim 2, wherein: The means for detecting the actual parallax value detects the actual parallax value for each horizontal coordinate value and each vertical coordinate value for each distance,
The means for generating the actual parallax measurement value table generates an actual parallax measurement value table indicating the actual parallax value corresponding to the design parallax value and the vertical coordinate value and the vertical coordinate value,
The stereo camera calibration according to claim 2 or 3, wherein the means for generating the parallax correction table generates a parallax correction table indicating an actual parallax value, a horizontal coordinate value, and a correction value corresponding to the horizontal coordinate value. Equipment. A stereo camera calibration method for generating a parallax correction table for correcting parallax detection errors of a stereo camera composed of a pair of cameras,
In association with multiple design parallax values, set multiple distances between the stereo camera and the correction subject,
From the stereo image acquired by photographing the correction subject with the stereo camera at each set distance, an actual parallax value for each horizontal coordinate value is detected for each distance,
Based on the detected actual parallax value, for each of a plurality of design parallax values, generate a real parallax measurement value table indicating the actual parallax value corresponding to the design value parallax value and the horizontal coordinate value,
A parallax correction table indicating correction values of the actual parallax values corresponding to the real parallax values and the horizontal coordinate values is generated for each of the plurality of real parallax values corresponding to the plurality of design parallax values from the actual parallax value measurement value table. To
Stereo camera calibration method characterized by the above. A stereo camera consisting of a pair of cameras,
Means for holding a parallax correction table indicating a correction value of the actual parallax value corresponding to the real parallax value and the horizontal coordinate value for each of the plurality of real parallax values corresponding to the plurality of design parallax values;
Means for detecting a parallax value for each horizontal coordinate value from a stereo image acquired by photographing a subject with the stereo camera;
Means for acquiring a correction value corresponding to the parallax value from the parallax correction table and correcting the parallax value for the detected parallax value;
Means for calculating a distance between the stereo camera and the subject using the corrected parallax value;
A distance measuring device comprising: The parallax correction table indicates, for each of a plurality of real parallax values corresponding to a plurality of design parallax values, a correction value corresponding to the real parallax value, a horizontal coordinate value, and a vertical coordinate value,
The means for detecting the parallax value detects a parallax value for each horizontal coordinate value and vertical coordinate value from the stereo image.
The distance measuring device according to claim 6.

Images