A four-quadrant hybrid binocular vision three-dimensional measurement system and method

Abstract

The present invention belongs to the field of measurement technology, and provides a four-quadrant hybrid binocular vision three-dimensional measurement system and method. The four-quadrant hybrid binocular vision three-dimensional measurement system is composed of a four-quadrant hybrid binocular vision sensor and a computer. The four-quadrant hybrid binocular vision sensor is composed of a main camera, a lower right auxiliary camera, a lower left auxiliary camera, an upper left auxiliary camera and an upper right auxiliary camera. The measurement method of the present invention uses a four-quadrant hybrid binocular vision sensor to collect images and process them in a computer. The four quadrants of the main camera image respectively form a binocular pair with the four auxiliary cameras. The four binocular pairs constitute a four-quadrant hybrid binocular measurement model, extract the image coordinates of the feature points, and calculate the three-dimensional coordinates of the object to be measured. The present invention constructs a hybrid binocular structure composed of the main camera quadrant image and the auxiliary camera image for the first time, which reduces the volume of the traditional multi-camera vision sensor while ensuring the measurement accuracy and measurement range, and can broaden the application scope of visual precision measurement in confined spaces.

Description

Technical Field

[0001] The invention belongs to the technical field of visual measurement, and in particular relates to a four-quadrant mixed binocular vision three-dimensional measurement system and method. Background Art

[0002] As a sensor carrier and direction constraint device, precision pipelines are widely used as key components of heavy equipment in the pillar industries of the national economy and in the fields of national defense and security, such as drill collars and drill pipes for oil exploration, nuclear reactor steam pipes, weapon barrels, and ship water, oil, and gas transmission power pipelines. This type of key component has a small caliber and a length of more than several meters. The internal space is limited, and the parameters of its inner surface, including the inner diameter, coaxiality, and the transition area morphology of the stepped hole, have strict high-precision design requirements, but there is a lack of effective means to perform precise detection in the early stages of processing. Therefore, it is of great significance to study three-dimensional measurement methods suitable for confined spaces such as precision pipelines.

[0003] Existing visual three-dimensional measurement methods generally use a single camera or multiple cameras to collect images. A single camera usually needs to be combined with a reflector or a rotating device to provide images from different viewpoints. This method has a complex hardware structure and requires high processing accuracy for the reflector and the rotating device. Due to the limited overlapping area of ​​images from different viewpoints, the number of effective matching points of images from different viewpoints is limited, the sensor field of view and measurement range are small, and the three-dimensional reconstruction accuracy is not high. Although the visual three-dimensional measurement method using multiple cameras can ensure a larger field of view and measurement range than a single camera, the volume occupied by multiple cameras is large, and the measurement principle of multiple cameras forming multiple binocular pairs requires a certain distance between the cameras, which further increases the overall volume of the sensor. Therefore, the visual three-dimensional measurement method using multiple cameras is difficult to apply in confined spaces.

[0004] In order to ensure high 3D reconstruction accuracy and a large field of view, and reduce the size of the visual sensor, a four-quadrant hybrid binocular vision 3D measurement method is proposed. Four auxiliary cameras are distributed around the central main camera. The four quadrant images of the main camera and the four auxiliary cameras respectively form four quadrant binoculars. According to the four-quadrant hybrid binocular model, high-precision 3D reconstruction of the image edge in a larger field of view is performed. Since the auxiliary camera only needs to match a part of the main camera image, the auxiliary camera resolution is 1 / 4 to 1 / 2 of the main camera resolution. Therefore, the auxiliary camera can use a small low-resolution camera, which effectively reduces the size of the hardware structure in the traditional multi-camera visual 3D measurement method. The four-quadrant hybrid binocular 3D measurement method using the main camera combined with four auxiliary cameras reduces the size of the hardware structure while using multiple cameras to ensure measurement accuracy, and realizes high-precision 3D measurement in confined space. Summary of the invention

[0005] The technical problem solved by the present invention is: to overcome the shortcomings of the existing single-camera method in the visual three-dimensional measurement method, such as the complex hardware structure and low measurement accuracy, and the large measurement system volume and difficulty in applying to confined spaces, and to provide a four-quadrant hybrid binocular vision three-dimensional measurement method, in which four auxiliary cameras are distributed around the central main camera. The main camera image is divided into four quadrant images of the same size, and the four quadrant images of the main camera and the four auxiliary cameras respectively form four quadrant binoculars, a four-quadrant hybrid binocular measurement model is established, and the three-dimensional coordinates of the measured object are calculated, so as to realize high-precision three-dimensional measurement in a confined space.

[0006] The technical solution of the present invention is: a four-quadrant hybrid binocular vision three-dimensional measurement system, comprising a four-quadrant hybrid binocular vision sensor (1) and a computer (9); the four-quadrant hybrid binocular vision sensor (1) comprises a main camera (2) and four auxiliary cameras, the four auxiliary cameras being respectively a lower right auxiliary camera (3), a lower left auxiliary camera (4), an upper left auxiliary camera (5) and an upper right auxiliary camera (6); the four-quadrant hybrid binocular vision sensor (1) transmits the captured image to the computer (9) for storage and processing;

[0007] The structure of the four-quadrant hybrid binocular vision sensor (1) is arranged such that the main camera (2) is located in the middle, the four auxiliary cameras are distributed around the main camera (2), and the four quadrants of the main camera (2) image and the four auxiliary camera images respectively form a four-quadrant binocular pair;

[0008] The computer (9) establishes a four-quadrant hybrid binocular measurement model based on the four-quadrant binocular pair; simultaneously, a matching point search algorithm is used to perform stereo matching on the four-quadrant binocular pair to obtain four-quadrant matching point sets; and then, based on the four-quadrant hybrid binocular measurement model and the image coordinates of the feature points in the four-quadrant matching point sets, the three-dimensional coordinates of the feature points in the main camera coordinate system are calculated to achieve three-dimensional measurement of the object to be measured.

[0009] The distance between the main camera and each auxiliary camera is 100-200 mm, and the angle between the optical axis of each auxiliary camera and the optical axis of the main camera is 0°-30°, so as to ensure that the main camera and the auxiliary cameras do not overlap and block each other; the resolution of the main camera is 2-4 times the resolution of each auxiliary camera, so as to ensure that there are enough feature points in the main camera image to match the feature points in the four auxiliary camera images.

[0010] The implementation of establishing the four-quadrant hybrid binocular measurement model is as follows:

[0011] (1) Establish the main camera coordinate system O c0 -x c0 y c0 z c0 , lower right auxiliary camera coordinate system O c1 -x c1 y c1 zc1 , lower left auxiliary camera coordinate system O c2 -x c2 y c2 z c2 , upper left auxiliary camera coordinate system O c3 -x c3 y c3 z c3 and the upper right auxiliary camera coordinate system O c4 -x c4 y c4 z c4 , O c0 ,x c0 ,y c0 ,z c0 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the main camera coordinate system, O c1 ,x c1 ,y c1 ,z c1 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the coordinate system of the lower right auxiliary camera, O c2 ,x c2 ,y c2 ,z c2 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the coordinate system of the lower left auxiliary camera, O c3 ,x c3 ,y c3 ,z c3 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the upper left auxiliary camera coordinate system, O c4 ,x c4 ,y c4 ,z c4 They represent the coordinate origin, x-axis, y-axis, and z-axis of the upper right auxiliary camera coordinate system respectively;

[0012] (2) The rule for establishing the quadrant coordinate system in the four-quadrant hybrid binocular model is that the origin of the quadrant coordinate system is the upper left corner of the main camera image after quadrant segmentation;

[0013] (3) Determine the parameters of the four-quadrant hybrid binocular model as the structural parameters of the first quadrant binocular, the second quadrant binocular, the third quadrant binocular and the fourth quadrant binocular, and the structural parameters include the main camera coordinate system O c0 -x c0 y c0 z c0 To the lower right auxiliary camera coordinate system O c1 -x c1 y c1 z c1 , lower left auxiliary camera coordinate system O c2 -x c2 y c2 z c2, upper left auxiliary camera coordinate system O c3 -x c3 y c3 z c3 , upper right auxiliary camera coordinate system O c4 -x c4 y c4 z c4 The rotation matrix and translation vector.

[0014] The matching point search algorithm is implemented as follows:

[0015] (1) Segment the main camera image into four quadrant images and extract feature points from the main camera quadrant image and the auxiliary camera image;

[0016] (2) Using epipolar constraints, search for feature points in the first quadrant image and the lower right auxiliary camera image, perform stereo matching, and construct a first quadrant matching point set;

[0017] (3) Using epipolar constraints, search for feature points in the second quadrant image and the lower left auxiliary camera image, perform stereo matching, and construct a second quadrant matching point set;

[0018] (4) Using epipolar constraints, search for feature points in the third quadrant image and the upper left auxiliary camera image, perform stereo matching, and construct a third quadrant matching point set;

[0019] (5) Using epipolar constraints, search for feature points in the fourth quadrant image and the upper right auxiliary camera image, perform stereo matching, and construct a fourth quadrant matching point set;

[0020] The first, second, third and fourth quadrant matching point sets constitute four quadrant matching point sets.

[0021] like Figure 1 As shown, a four-quadrant hybrid binocular vision three-dimensional measurement method of the present invention comprises the following steps:

[0022] Step 1: adjust the position of the four-quadrant hybrid binocular vision sensor (1) to ensure that the object to be measured is imaged in the main camera (2), the lower right auxiliary camera (3), the lower left auxiliary camera (4), the upper left auxiliary camera (5) and the upper right auxiliary camera (6) at the same time; adjust the focal length and aperture of the main camera (2), the lower right auxiliary camera (3), the lower left auxiliary camera (4), the upper left auxiliary camera (5) and the upper right auxiliary camera (6) to ensure that the image is clear, and tighten the focal length and aperture after adjustment; use the main camera (2), the lower right auxiliary camera (3), the lower left auxiliary camera (4), the upper left auxiliary camera (5) and the upper right auxiliary camera (6) to take an image respectively;

[0023] Step 2: split the image sequence captured by the main camera (2) into four images of the same size, which are respectively called the first quadrant image, the second quadrant image, the third quadrant image and the fourth quadrant image; the first quadrant image and the lower right auxiliary camera (3) constitute a first quadrant binocular pair, the second quadrant image and the lower left auxiliary camera (4) constitute a second quadrant binocular pair, the third quadrant image and the upper left auxiliary camera (5) constitute a third quadrant binocular pair, and the fourth quadrant image and the upper right auxiliary camera (6) constitute a fourth quadrant binocular pair; finally, a four-quadrant binocular pair is obtained;

[0024] Step 3: Based on the four-quadrant binocular pair, a four-quadrant hybrid binocular measurement model is established. The specific steps are as follows:

[0025] (1) Establish the main camera coordinate system O c0 -x c0 y c0 z c0 , lower right auxiliary camera coordinate system O c1 -x c1 y c1 z c1 , lower left auxiliary camera coordinate system O c2 -x c2 y c2 z c2 , upper left auxiliary camera coordinate system O c3 -x c3 y c3 z c3 and the upper right auxiliary camera coordinate system O c4 -x c4 y c4 z c4 , O c0 ,x c0 ,y c0 ,z c0 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the main camera coordinate system, O c1 ,x c1 ,y c1 ,z c1 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the coordinate system of the lower right auxiliary camera, O c2 ,x c2 ,y c2 ,z c2 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the coordinate system of the lower left auxiliary camera, O c3 ,x c3 ,y c3 ,z c3 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the upper left auxiliary camera coordinate system, O c4 ,x c4 ,y c4 ,z c4They represent the coordinate origin, x-axis, y-axis, and z-axis of the upper right auxiliary camera coordinate system respectively;

[0026] (2) The rule for establishing the quadrant coordinate system in the four-quadrant hybrid binocular model is that the origin of the quadrant coordinate system is the upper left corner of the main camera image after quadrant segmentation;

[0027] (3) Determine the parameters of the model as the structural parameters of the first quadrant binocular, the second quadrant binocular, the third quadrant binocular and the fourth quadrant binocular, wherein the structural parameters include the main camera coordinate system O c0 -x c0 y c0 z c0 To the lower right auxiliary camera coordinate system O c1 -x c1 y c1 z c1 , lower left auxiliary camera coordinate system O c2 -x c2 y c2 z c2 , upper left auxiliary camera coordinate system O c3 -x c3 y c3 z c3 , upper right auxiliary camera coordinate system O c4 -x c4 y c4 z c4 The rotation matrix and translation vector of

[0028] Step 4: Use the matching point search algorithm to perform stereo matching on the four-quadrant binocular pair to obtain a four-quadrant matching point set. The specific steps are:

[0029] (1) Segment the main camera image into four quadrant images and extract feature points from the main camera quadrant image and the auxiliary camera image;

[0030] (2) Using epipolar constraints, search for feature points in the first quadrant image and the lower right auxiliary camera image, perform stereo matching, and construct a first quadrant matching point set;

[0031] (3) Using epipolar constraints, search for feature points in the second quadrant image and the lower left auxiliary camera image, perform stereo matching, and construct a second quadrant matching point set;

[0032] (4) Using epipolar constraints, search for feature points in the third quadrant image and the upper left auxiliary camera image, perform stereo matching, and construct a third quadrant matching point set;

[0033] (5) Using epipolar constraints, search for feature points in the fourth quadrant image and the upper right auxiliary camera image, perform stereo matching, and construct a fourth quadrant matching point set;

[0034] Step 5: According to the four-quadrant hybrid binocular measurement model, the image coordinates of the feature points in the first, second, third and fourth quadrant matching points are used to calculate the feature points in the main camera coordinate system O c0 -x c0 y c0 z c0 The three-dimensional coordinates under the test are saved in the data file to realize the three-dimensional measurement of the object under test.

[0035] The advantages of the present invention compared with the prior art are:

[0036] (1) A visual sensor structure is formed by combining a main camera and four small auxiliary cameras. The existing multi-camera visual sensor uses multiple cameras with the same resolution to form a binocular pair, while the present invention uses a part of the main camera image and the auxiliary camera to form a binocular pair, which reduces the size of the multi-camera visual sensor and facilitates high-precision three-dimensional reconstruction in confined spaces.

[0037] (2) The present invention mainly combines a main camera with four auxiliary cameras, splits the main camera image into four quadrants, and matches them with the images of the four auxiliary cameras respectively. The existing single-camera 3D reconstruction technology has low reconstruction accuracy, and the multi-camera 3D reconstruction technology uses multiple high-resolution cameras for matching. Although the reconstruction accuracy is high, it occupies a large space and is difficult to apply to confined spaces. The main-auxiliary camera structure designed by the present invention uses a small auxiliary camera to reduce the sensor volume while using multiple cameras to ensure reconstruction accuracy, thereby achieving high-precision 3D reconstruction in confined spaces.

[0038] (3) The present invention establishes the four-quadrant image coordinate system of the main camera and determines the establishment rules of the quadrant coordinate system in the four-quadrant hybrid binocular model, providing a reference for the subsequent construction of the transformation relationship between the normalized image coordinate system of the main camera and the four-quadrant image coordinate system of the main camera. The transformation relationship between the normalized image coordinate system of the main camera and the four-quadrant image coordinate system of the main camera is constructed, and the four-quadrant hybrid binocular model is described from the perspective of coordinate system transformation. It has the advantages of high three-dimensional reconstruction accuracy and large measurement field of view in confined space.

[0039] (4) The present invention designs a matching point search algorithm for the four quadrants of the main camera image and the four auxiliary camera images, calculates the matching points of the four quadrants according to the four-quadrant hybrid binocular measurement model, obtains the three-dimensional coordinates of the object to be measured, and realizes the four-quadrant hybrid binocular vision three-dimensional measurement method. Under the premise of ensuring high-precision measurement and a large field of view, the volume of the traditional multi-camera vision sensor is reduced, and three-dimensional measurement in confined space is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flow chart of the four-quadrant hybrid binocular vision three-dimensional measurement method in the present invention;

[0041] Figure 2 It is a structural schematic diagram of the four-quadrant hybrid binocular vision three-dimensional measurement system in the present invention;

[0042] Figure 3 Schematic diagram of the conversion relationship between the normalized image coordinate system of the main camera and the quadrant image coordinate system in the present invention;

[0043] Figure 4 Schematic diagram of the four-quadrant hybrid binocular model in the present invention;

[0044] Figure 5 This is a flow chart of stereo matching in the present invention. DETAILED DESCRIPTION

[0045] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0046] like Figure 2 As shown, the four-quadrant hybrid binocular vision measurement system includes a four-quadrant hybrid binocular vision sensor 1, a data transmission line 8, a computer 9 and a power supply 11; the four-quadrant hybrid binocular vision sensor 1 is composed of a main camera 2, a lower right auxiliary camera 3, a lower left auxiliary camera 4, an upper left auxiliary camera 5, an upper right auxiliary camera 6 and a mechanical bracket 7; the data transmission line 10 is connected to the computer 9; the four-quadrant hybrid binocular vision sensor 1 transmits the captured image to the computer 9 for storage and processing.

[0047] The geometric parameters of the four-quadrant hybrid binocular vision sensor 1 include: the main camera is located in the middle of the vision sensor, the four auxiliary cameras are distributed around the main camera, the distance between the main camera and the auxiliary camera is 100-200 mm, and the angle between the optical axis of the auxiliary camera and the optical axis of the main camera is 0°-30°; the resolution of the main camera is 2-4 times that of the auxiliary camera. Since the auxiliary camera only needs to match a part of the main camera image, the auxiliary camera can use a low-resolution camera that is smaller than the main camera. For example, the volumes of the main camera and the auxiliary camera are 43 cm and 13 cm respectively. 2 and 24cm 2 , the resolutions are 5120×5120 and 2048×1536 respectively, which effectively reduces the volume of the hardware structure.

[0048] The present invention constructs a four-quadrant hybrid binocular measurement model and defines the coordinate system as follows: w -x w y w z w is the world coordinate system, O w ,x w ,y w ,z w Respectively represent the origin, x-axis, y-axis and z-axis of the world coordinate system; ci -x ci y ci zci , i=0,1,2,3,4 are the camera coordinate systems, where i=0 is the main camera coordinate system, and i=1,2,3,4 are the four auxiliary camera coordinate systems corresponding to the first, second, third, and fourth quadrants respectively. ci ,x ci ,y ci ,z ci Respectively represent the origin, x-axis, y-axis and z-axis of the camera coordinate system; ui -x ui y ui , i=0,1,2,3,4 is the image coordinate system, O ci ,x ci ,y ci ,z ci Respectively represent the origin, x-axis, y-axis and z-axis of the camera coordinate system; ni -x ni y ni , i=0,1,2,3,4 is the normalized image coordinate system, O ni ,x ni ,y ni Respectively represent the origin, x-axis and y-axis of the camera coordinate system; li -x li y li , i=1,2,3,4 is the quadrant image coordinate system O li ,x li ,y li They represent the origin, x-axis, and y-axis of the camera coordinate system respectively.

[0049] like Figure 3 As shown, the normalized image coordinate system O of the main camera n0 -x n0 y n0 Divided into four quadrants, the four quadrants are defined as the four quadrant image coordinate system O li -x li y li , i = 1, 2, 3, 4, the origin of the quadrant coordinate system is the upper left corner of the main camera image after quadrant division. Let Q i , i = 1, 2, 3, 4 is the translation matrix from the normalized image coordinate system of the main camera to the quadrant image coordinate system. The four quadrant coordinate systems are O li -x li y li , i=1,2,3,4 are respectively related to the normalized image coordinate system O of the four auxiliary cameras ni -x ni y ni , i=1,2,3,4 correspondingly.

[0050] Suppose a point P in three-dimensional space wThe homogeneous coordinates in the world coordinate system are The homogeneous coordinates in the camera coordinate system are The homogeneous coordinates in the image coordinate system are The homogeneous coordinates in the normalized image coordinate system are The homogeneous coordinates in the quadrant image coordinate system are

[0051] like Figure 4 As shown in FIG. 1 , a four-quadrant hybrid binocular model is constructed, and the transformation relationship between the normalized image coordinate system of the main camera and the four quadrant image coordinate systems of the main camera is constructed to obtain the rotation matrix and translation matrix between the main camera and the four auxiliary cameras. w -x w y w z w To the camera coordinate system O ci -x ci y ci z ci , i=0,1,2,3,4 rigid body transformation is expressed as:

[0052]

[0053] Among them, R i and T i are the extrinsic parameters rotation matrix and translation vector of the i-th camera respectively.

[0054] From the camera coordinate system O ci -x ci y ci z ci ,i=0,1,2,3,4 to image coordinate system O ui -x ui y ui , i=0,1,2,3,4 The perspective transformation is expressed as:

[0055]

[0056] Among them, s i is the scale factor of the ith camera, A i is the intrinsic parameter matrix of the i-th camera, (f xi ,f yi ) is the focal length of the i-th camera, (u 0i ,v 0i ) is the intersection of the optical axis of the ith camera and the image plane.

[0057] From the image coordinate system O ui -x ui y ui , i=0,1,2,3,4 to the normalized image coordinate system:

[0058]

[0059] Normalize the image coordinate system O from the main camera n0 -x n0 y n0 The transformation relationship to the quadrant image coordinate system is expressed as:

[0060]

[0061] Among them, Q i is the translation matrix from the normalized image coordinate system of the main camera to the image coordinate system of the i-th quadrant.

[0062] The rotation matrix and translation matrix between the main camera and the four auxiliary cameras are:

[0063]

[0064] Among them, R i0 and T i0 are the rotation matrix and translation matrix between the main camera and the i-th auxiliary camera, R i and T i are the external parameter rotation matrix and translation vector of the i-th auxiliary camera, and R0 and T0 are the external parameter rotation matrix and translation vector of the main camera, respectively.

[0065] like Figure 5 As shown, the matching point search algorithm proposed in the present invention performs stereo matching on the four quadrants of the main camera image and the four auxiliary camera images, and the specific steps are:

[0066] The first step is to segment the main camera image into four quadrant images and extract feature points from the main camera quadrant image and the auxiliary camera image;

[0067] In the second step, the epipolar constraint is used to search for feature points in the first quadrant image and the lower right auxiliary camera image, and stereo matching is performed to construct the first quadrant matching point set;

[0068] The third step is to use the epipolar constraint to search for feature points in the second quadrant image and the lower left auxiliary camera image, perform stereo matching, and construct the second quadrant matching point set;

[0069] The fourth step is to use the epipolar constraint to search for feature points in the third quadrant image and the upper left auxiliary camera image, perform stereo matching, and construct a third quadrant matching point set;

[0070] The fifth step is to use the epipolar constraint to search for feature points in the fourth quadrant image and the upper right auxiliary camera image, perform stereo matching, and construct the fourth quadrant matching point set.

[0071] According to the four-quadrant hybrid binocular measurement model, the image coordinates of the feature points of stereo matching are used to calculate the feature points at O c0 -x c0 y c0 z c0 The three-dimensional coordinates in the coordinate system are saved in the data file to realize the three-dimensional measurement of the object being measured.

Claims

1. A four-quadrant hybrid binocular vision three-dimensional measurement system, characterized in that: The measuring system comprises a four-quadrant hybrid binocular vision sensor (1) and a computer (9); the four-quadrant hybrid binocular vision sensor (1) comprises a main camera (2) and four auxiliary cameras, the four auxiliary cameras being respectively a lower right auxiliary camera (3), a lower left auxiliary camera (4), an upper left auxiliary camera (5) and an upper right auxiliary camera (6); the four-quadrant hybrid binocular vision sensor (1) transmits the captured image to the computer (9) for storage and processing; The structure of the four-quadrant hybrid binocular vision sensor (1) is arranged such that the main camera (2) is located in the middle, the four auxiliary cameras are distributed around the main camera (2), and the four quadrants of the main camera (2) image and the four auxiliary camera images respectively form a four-quadrant binocular pair; The computer (9) establishes a four-quadrant hybrid binocular measurement model based on the four-quadrant binocular pair; simultaneously, a matching point search algorithm is used to perform stereo matching on the four-quadrant binocular pair to obtain four-quadrant matching point sets; and then, based on the four-quadrant hybrid binocular measurement model and the image coordinates of the feature points in the four-quadrant matching point sets, the three-dimensional coordinates of the feature points in the main camera coordinate system are calculated to achieve three-dimensional measurement of the object to be measured.

2. The four-quadrant hybrid binocular vision three-dimensional measurement system according to claim 1, characterized in that: The distance between the main camera and each auxiliary camera is 100-200 mm, the angle between the optical axis of each auxiliary camera and the optical axis of the main camera is 0°-30°; the resolution of the main camera is 2-4 times the resolution of each auxiliary camera.

3. The four-quadrant hybrid binocular vision three-dimensional measurement system according to claim 1, characterized in that: The implementation of establishing the four-quadrant hybrid binocular measurement model is as follows: (1) Establish the main camera coordinate system O c0 -x c0 y c0 z c0 , lower right auxiliary camera coordinate system O c1 -x c1 y c1 z c1 , lower left auxiliary camera coordinate system O c2 -x c2 y c2 z c2 , upper left auxiliary camera coordinate system O c3 -x c3 y c3 z c3 and the upper right auxiliary camera coordinate system O c4 -x c4 y c4 z c4 , O c0 ,x c0 ,y c0 ,z c0 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the main camera coordinate system, O c1 ,x c1 ,y c1 ,z c1 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the coordinate system of the lower right auxiliary camera, O c2 ,x c2 ,y c2 ,z c2 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the coordinate system of the lower left auxiliary camera, O c3 ,x c3 ,y c3 ,z c3 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the upper left auxiliary camera coordinate system, O c4 ,x c4 ,y c4 ,z c4 They represent the coordinate origin, x-axis, y-axis, and z-axis of the upper right auxiliary camera coordinate system respectively; (2) The rule for establishing the quadrant coordinate system in the four-quadrant hybrid binocular model is that the origin of the quadrant coordinate system is the upper left corner of the main camera image after quadrant segmentation; (3) Determine the parameters of the four-quadrant hybrid binocular model as the structural parameters of the first quadrant binocular, the second quadrant binocular, the third quadrant binocular and the fourth quadrant binocular, and the structural parameters include the main camera coordinate system O c0 -x c0 y c0 z c0 To the lower right auxiliary camera coordinate system O c1 -x c1 y c1 z c1 , lower left auxiliary camera coordinate system O c2 -x c2 y c2 z c2 , upper left auxiliary camera coordinate system O c3 -x c3 y c3 z c3 , upper right auxiliary camera coordinate system O c4 -x c4 y c4 z c4 The rotation matrix and translation vector.

4. The four-quadrant hybrid binocular vision three-dimensional measurement system according to claim 1, characterized in that: The matching point search algorithm is implemented as follows: (1) Segment the main camera image into four quadrant images and extract feature points from the main camera quadrant image and the auxiliary camera image; (2) Using epipolar constraints, search for feature points in the first quadrant image and the lower right auxiliary camera image, perform stereo matching, and construct a first quadrant matching point set; (3) Using epipolar constraints, search for feature points in the second quadrant image and the lower left auxiliary camera image, perform stereo matching, and construct a second quadrant matching point set; (4) Using epipolar constraints, search for feature points in the third quadrant image and the upper left auxiliary camera image, perform stereo matching, and construct a third quadrant matching point set; (5) Using epipolar constraints, search for feature points in the fourth quadrant image and the upper right auxiliary camera image, perform stereo matching, and construct a fourth quadrant matching point set; The first, second, third and fourth quadrant matching point sets constitute four quadrant matching point sets.

5. A four-quadrant hybrid binocular vision three-dimensional measurement method, characterized in that: The following steps are involved: Step 1: adjust the position of the four-quadrant hybrid binocular vision sensor (1) to ensure that the object to be measured is imaged in the main camera (2), the lower right auxiliary camera (3), the lower left auxiliary camera (4), the upper left auxiliary camera (5) and the upper right auxiliary camera (6) at the same time; adjust the focal length and aperture of the main camera (2), the lower right auxiliary camera (3), the lower left auxiliary camera (4), the upper left auxiliary camera (5) and the upper right auxiliary camera (6) to ensure that the image is clear, and tighten the focal length and aperture after adjustment; use the main camera (2), the lower right auxiliary camera (3), the lower left auxiliary camera (4), the upper left auxiliary camera (5) and the upper right auxiliary camera (6) to take an image respectively; Step 2: split the image sequence captured by the main camera (2) into four images of the same size, which are respectively called the first quadrant image, the second quadrant image, the third quadrant image and the fourth quadrant image; the first quadrant image and the lower right auxiliary camera (3) constitute a first quadrant binocular pair, the second quadrant image and the lower left auxiliary camera (4) constitute a second quadrant binocular pair, the third quadrant image and the upper left auxiliary camera (5) constitute a third quadrant binocular pair, and the fourth quadrant image and the upper right auxiliary camera (6) constitute a fourth quadrant binocular pair; finally, a four-quadrant binocular pair is obtained; Step 3: Based on the four-quadrant binocular pair, a four-quadrant hybrid binocular measurement model is established. The specific steps are as follows: (1) Establish the main camera coordinate system O c0 -x c0 y c0 z c0 , lower right auxiliary camera coordinate system O c1 -x c1 y c1 z c1 , lower left auxiliary camera coordinate system O c2 -x c2 y c2 z c2 , upper left auxiliary camera coordinate system O c3 -x c3 y c3 z c3 and the upper right auxiliary camera coordinate system O c4 -x c4 y c4 z c4 , O c0 ,x c0 ,y c0 ,z c0 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the main camera coordinate system, O c1 ,x c1 ,y c1 ,z c1 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the coordinate system of the lower right auxiliary camera, O c2 ,x c2 ,y c2 ,z c2 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the coordinate system of the lower left auxiliary camera, O c3 ,x c3 ,y c3 ,z c3 Respectively represent the coordinate origin, x-axis, y-axis and z-axis of the upper left auxiliary camera coordinate system, O c4 ,x c4 ,y c4 ,z c4 They represent the coordinate origin, x-axis, y-axis, and z-axis of the upper right auxiliary camera coordinate system respectively; (2) The rule for establishing the quadrant coordinate system in the four-quadrant hybrid binocular model is that the origin of the quadrant coordinate system is the upper left corner of the main camera image after quadrant segmentation; (3) Determine the parameters of the model as the structural parameters of the first quadrant binocular, the second quadrant binocular, the third quadrant binocular and the fourth quadrant binocular, wherein the structural parameters include the main camera coordinate system O c0 -x c0 y c0 z c0 To the lower right auxiliary camera coordinate system O c1 -x c1 y c1 z c1 , lower left auxiliary camera coordinate system O c2 -x c2 y c2 z c2 , upper left auxiliary camera coordinate system O c3 -x c3 y c3 z c3 , upper right auxiliary camera coordinate system O c4 -x c4 y c4 z c4 The rotation matrix and translation vector of Step 4: Use the matching point search algorithm to perform stereo matching on the four-quadrant binocular pair to obtain a four-quadrant matching point set. The specific steps are: (1) Segment the main camera image into four quadrant images and extract feature points from the main camera quadrant image and the auxiliary camera image; (3) Using epipolar constraints, search for feature points in the first quadrant image and the lower right auxiliary camera image, perform stereo matching, and construct a first quadrant matching point set; (4) Using epipolar constraints, search for feature points in the second quadrant image and the lower left auxiliary camera image, perform stereo matching, and construct a second quadrant matching point set; (5) Using epipolar constraints, search for feature points in the third quadrant image and the upper left auxiliary camera image, perform stereo matching, and construct a third quadrant matching point set; (6) Using epipolar constraints, search for feature points in the fourth quadrant image and the upper right auxiliary camera image, perform stereo matching, and construct a fourth quadrant matching point set; Step 5: According to the four-quadrant hybrid binocular measurement model, the image coordinates of the feature points in the first, second, third and fourth quadrant matching points are used to calculate the feature points in the main camera coordinate system O c0 -x c0 y c0 z c0 The three-dimensional coordinates under the test are saved in the data file to realize the three-dimensional measurement of the object under test.

Images