CN111242901A - Space point-based global calibration system and method for automobile detection camera without common view field - Google Patents

Abstract

The invention discloses a system and method for global calibration of cameras without common field of view for vehicle detection based on space points, aiming at solving the problem of global calibration of cameras without common field of view for vehicle detection based on space points. The global calibration system for cameras without a common field of view for vehicle detection based on spatial points is mainly composed of a total camera (1), a total camera bracket (2), a sub-camera left (3), a sub-camera bracket left (4), and a sub-camera right (5). . The camera bracket right (6) and the cylindrical target (7) are composed. The global calibration method for cameras without a common field of view for vehicle detection based on spatial points is composed of image acquisition, and calculation of the homography matrix transformed from the cylindrical target (7) to the coordinate system of the general camera (1) according to the image collected by the general camera (1), etc. The invention provides a global calibration system and method for a vehicle detection without a common field of view based on a simple structure and reliable performance.

Description

System and method for global calibration of cameras without common field of view for vehicle detection based on spatial point

technical field

The invention relates to a measuring device and a measuring method in the field of automobile detection, more specifically, it is a global calibration system and method of a camera without a common field of view for automobile detection based on space points.

Background technique

The purpose of research in the field of automobile performance testing is to regularly test the safety, reliability, passability and other important performance of automobiles, so as to ensure the safe operation of vehicles and the safety of life and property of road users. Due to its advantages of non-contact, low cost and high precision, visual inspection is used in important research directions of automobile performance detection such as automobile wheel alignment parameter detection, automobile wheelbase difference detection, automobile shape detection, multi-axis axle angle detection, and vehicle model identification. It has important research significance and broad application prospects. Due to the limited field of view, a single camera cannot meet the shape measurement requirements of large measured objects such as vehicles and the position measurement requirements of long-distance measured objects such as positioning parameters between multiple wheels. A wide range of camera measurement fields or inspection fields are used for measurement and calibration. For dual cameras or multiple cameras, there may be situations where the calibration reference object cannot appear in the common field of view of the two cameras at the same time, but at the same time, the measurement results of the two cameras need to be combined into one coordinate system for measurement. and reconstruction, therefore, determining the pose relationship between two cameras without a common field of view is an indispensable and important step for the entire measurement system. Since the two-camera method can be extended to the study of multiple coordinate systems without a common field of view camera, it is equally important to study the wheel alignment, wheel alignment parameters, wheelbase difference, and topography reconstruction of high-speed rail vehicles. research value. The two-camera calibration method in the prior art requires the participation of large-scale external auxiliary equipment such as a guide rail table, which is complicated in structure, high in cost, and inconvenient in operation, and cannot be used in on-site calibration. The present application adopts an independent camera and a cylindrical target to construct a three-dimensional space point as a bridge between the two cameras, and realizes the coordinate system 1 of cameras without a common field of view for vehicle detection based on space points.

SUMMARY OF THE INVENTION

The invention aims to solve the problems that a single camera cannot meet the measurement requirements of a large-scale system due to the limited field of view in the process of automobile detection, and the calibration reference objects of the dual cameras may not appear in the public field of view of the two cameras at the same time. A flexible, simple, safe and reliable calibration system and method with a simple structure. The images in each field of view are acquired by three cameras respectively, the pose relationship between the three cameras has been determined, and the coordinate system 1 of cameras without a common field of view is realized for vehicle detection of spatial points.

In conjunction with the accompanying drawings of the description, the present invention adopts the following technical solutions to be realized:

The spatial point-based global calibration system for cameras without a common field of view for vehicle detection includes a total camera, a total camera bracket, a sub-camera left, a sub-camera bracket left, a sub-camera right, a sub-camera bracket right, and a cylindrical target;

The main camera bracket, the right sub-camera bracket, the left sub-camera bracket and the cylindrical target are placed on the ground. The sub-camera left and sub-camera right have no common field of view. The main camera bracket, the left of the sub-camera bracket and the top right of the sub-camera bracket are fixedly connected with bolts and threads.

The total camera bracket described in the technical solution is a height-adjustable tripod bracket.

The left side of the sub-camera bracket described in the technical solution is a height-adjustable tripod bracket.

The right side of the sub-camera bracket described in the technical solution is a height-adjustable tripod bracket.

The total camera described in the technical solution is a wide-angle industrial camera.

The left side of the sub-camera described in the technical solution is a wide-angle industrial camera.

The right sub-camera described in the technical solution is a wide-angle industrial camera.

The cylindrical target described in the technical solution is a hollow cylinder with a checkerboard pattern pasted on the outer surface.

The specific steps of the global calibration method for cameras without a common field of view for vehicle detection based on spatial points are as follows:

The first step: vehicle detection based on spatial points Image acquisition without a common field of view camera global calibration:

The main camera, sub-camera left and sub-camera right are respectively fixed on the main camera bracket, sub-camera bracket left and sub-camera bracket upper right, and the main camera bracket, sub-camera bracket left and sub-camera bracket right are placed on the ground. To meet the needs of a large detection range, there is no common field of view for the left and right of the sub-camera. The cylindrical target is placed on the ground and is in the field of view of the main camera and the left of the sub-camera. The main camera and the left of the sub-camera collect an image respectively. , move the cylindrical target so that it is in the field of view of the main camera and the right of the sub-camera, and the main camera and the right of the sub-camera collect an image respectively;

Step 2: When the cylindrical target is in the left common field of view of the total camera and the sub-camera, calculate the homography matrix converted from the cylindrical target to the total camera coordinate system according to the image collected by the total camera:

The conversion relationship from the cylindrical target coordinate system to the image coordinate system obtained by the total camera is:

P _{T1, I0} X _T1 = sx _{I0, T1}

Using the RANSAC point extraction method and the DLT calibration method, the projection matrix P _{T1, I0} = K _{T1, I0} [R _{T1, I0} t _{T1, I0} ], K _{T1, I0} are the internal parameters of the total camera, R _{T1, I0} , t _{T1, I0} are the external parameters of the total camera obtained according to the QR decomposition, X _T1 are the coordinates of the cylindrical target coordinate system of the cylindrical target feature point, x _{I0, T1} are the cylindrical target feature point X _T1 under the image obtained by the total camera image Coordinates, s is the scale factor, the homography matrix converted from the coordinate system of the cylindrical target in the left field of view of the sub-camera to the coordinate system of the total camera can be obtained from the rotation matrix R _{T1, I0} and the translation vector t _{T1, I0}

Step 3: When the cylindrical target is in the right common field of view of the total camera and the sub-camera, calculate the homography matrix converted from the cylindrical target to the total camera coordinate system according to the image collected by the total camera 1:

The conversion relationship between the cylindrical target coordinate system and the image coordinate system obtained by the total camera 1 is:

P _{T2, I0} X _T2 =sx _{I0, T2}

Using the RANSAC point extraction method and the DLT calibration method, the projection matrix P _{T2, I0} =K _{T2, I0} [R _{T2, I0} t _{T2, I0} ], K _{T2, I0} are the internal parameters of the total camera, R _{T2, I0} , t _{T2, I0} are the external parameters of the total camera obtained by QR decomposition, X _T2 are the coordinates of the cylindrical target coordinate system of the cylindrical target feature point, x _{I0, T2} are the image of the cylindrical target feature point X _T2 under the image obtained by the total camera Coordinates, the homography matrix converted from the coordinate system of the cylindrical target in the right field of view of the sub-camera to the coordinate system of the total camera can be obtained from the rotation matrix R _{T2, I0} and the translation vector t _{T2, I0}

According to the obtained H _{T1, CO} and H _{T2, CO} , the homography matrix from the left coordinate system of the sub-camera to the right coordinate system of the sub-camera can be obtained

H _{T2, T1} = H _{T2, C0} (H _{T1, C0} ) ^-1

Expand H _{T2 and T1} further to get

Step 4: When the cylindrical target is in the left common field of view of the total camera and the sub-camera, calculate the homography matrix converted from the cylindrical target to the left coordinate system of the sub-camera according to the image collected by the left sub-camera:

The conversion relationship between the cylindrical target coordinate system and the image coordinate system obtained by the left sub-camera is:

P _{T1, I1} X _T1 =sx _{I1, T1}

Using the RANSAC point extraction method and the DLT calibration method, the projection matrix P _{T1, I1} = K _{T1, I1} [R _{T1, I1} t _{T1, I1} ], K _{T1, I1} are the left internal parameters of the sub-camera, R _{T1, I1} , t _{T1, I1} are the external parameters of the left sub-camera obtained by QR decomposition, x _{I1, T1} are the image coordinates of the cylindrical target feature point X _T1 under the image obtained by the sub-camera left, by the rotation matrix R _{T1, I1} and translation The vectors t _{T1, I1} can obtain the homography matrix from the coordinate system of the cylindrical target in the left field of view of the sub-camera to the left coordinate system of the sub-camera

Step 5: When the cylindrical target is in the right common field of view of the main camera and the sub-camera, calculate the homography matrix converted from the cylindrical target to the right coordinate system of the sub-camera according to the image collected from the right of the sub-camera:

The conversion relationship between the cylindrical target coordinate system and the image coordinate system obtained by the right sub-camera is as follows:

P _{T2, I2} X _T2 = sx _{I2, T2}

Using the RANSAC point extraction method and the DLT calibration method, the projection matrix P _{T2, I2} = K _{T2, I2} [R _{T2, I2} t _{T2, I2} ], K _{T2, I2} are the internal parameters of the right sub-camera, R _{T2, I2} , t _{T2, I2} are the external parameters of the right sub-camera obtained by QR decomposition, x _{I2, T2} are the image coordinates of the cylindrical target feature point X _T2 under the image acquired by the right sub-camera, by the rotation matrix R _{T2, I2} and translation The vectors t _{T2 and I2} can obtain the homography matrix from the coordinate system of the cylindrical target in the right field of view of the sub-camera to the right coordinate system of the sub-camera

Step 6: Calculate the homography matrix between the left coordinate system of the sub-camera and the right coordinate system of the sub-camera:

According to a series of homography matrices solved in the second to fifth steps, the homography matrix converted from the left coordinate system of the sub-camera to the right coordinate system of the sub-camera can be obtained

H _{C1, C2} = (H _{T2, C2} ) ^-1 H _{T2, T1} H _{T1, C1}

Expanding the homography matrix H _{C1, C2} , the relationship between the homography matrix converted from the left coordinate system of the sub-camera to the right coordinate system of the sub-camera and the rotation matrices and translation vectors obtained in the second to fifth steps is:

The beneficial effects of the present invention are:

(1) Aiming at the problem of global calibration of cameras without a common field of view, the method of the present invention introduces a third camera and a cylindrical target to construct a three-dimensional space point, establishes a conversion bridge between cameras without a common field of view, and realizes a point-based A coordinate system for vehicle detection without a common field-of-view camera. This method can achieve high-precision calibration, and the homography matrix between cameras can be determined without knowing the relationship between targets at different positions.

(2) The system of the present invention is convenient and flexible to use, overcomes the disadvantages of large external auxiliary equipment, limited space and poor convenience in the traditional calibration method, and can further realize the calibration of multiple cameras by expanding the number of cameras.

(3) The system of the present invention has a wide measurement range, reliable performance, simple device structure, simple operation and low cost, and solves the problem that the traditional single-camera reconstruction measurement range is small, and the dual-camera reconstruction has no common field of view calibration method requires a large-scale fixed contact type measurement system High price and low measurement efficiency.

Description of drawings

Fig. 1 is an axonometric view of a global calibration system for a vehicle detection without a common field of view based on a space point;

Fig. 2 is the axonometric view of the total camera 1 in the global calibration system of the vehicle detection without a common field of view based on the space point;

Fig. 3 is the axonometric view of the total camera bracket 2 in the global calibration system for cameras without a common field of view for vehicle detection based on space points;

FIG. 4 is an axonometric view of a cylindrical target 8 in a spatial point-based vehicle detection without a common field of view camera global calibration system;

Figure 5 is a schematic diagram of a global calibration method for a camera without a common field of view for vehicle detection based on spatial points;

6 is a flowchart of solving the homography matrix converted from the cylindrical target 7 coordinate system to the total camera 1 and to the left 3 coordinate system of the sub-camera in the spatial point-based vehicle detection without a common field of view camera global calibration method;

7 is a flow chart of solving the homography matrix converted from the cylindrical target 7 coordinate system to the total camera 1 and to the right 5 coordinate system of the sub-camera in the spatial point-based vehicle detection without a common field of view camera global calibration method;

8 is a flowchart of solving the homography matrix converted from the left 3 coordinate system of the sub-camera to the right 5 coordinate system of the sub-camera in the global calibration method of the camera without a common field of view based on the space point;

In the picture: 1. Main camera, 2. Main camera bracket, 3. Sub-camera left, 4. Sub-camera bracket left, 5. Sub-camera right, 6. Sub-camera bracket right, 7. Cylindrical target.

Detailed ways

Below in conjunction with accompanying drawing, the present invention is described in further detail:

Referring to Figures 1 to 4, the spatial point-based vehicle detection global calibration system for cameras without a common field of view includes a total camera 1, a total camera bracket 2, a left sub-camera 3, a sub-camera bracket left 4, a sub-camera right 5, and a sub-camera. The bracket right 6 and the cylindrical target 7;

The main camera bracket 2, the sub-camera bracket left 4 and the sub-camera bracket right 6 are the same height-adjustable triangular brackets. The main camera bracket 2, the sub-camera bracket left 4, the sub-camera bracket right 6 and the cylindrical target 7 are placed on the ground. , the main camera 1, the sub-camera left 3 and the sub-camera right 5 are wide-angle industrial cameras. The main camera 1, the sub-camera left 3 and the sub-camera right 5 are connected to the main camera bracket 2 and the sub-camera bracket left 4 respectively through the threaded holes at the bottom. It is fixedly connected with the bolts on the top of the right 6 of the sub-camera bracket. According to the needs of automobile detection for a large detection range, the left 3 of the sub-camera and the right of the sub-camera 5 have no common field of view. The cylindrical target 7 is a hollow cylinder with a chessboard pasted on the outer surface. grid pattern.

Referring to Figure 5 to Figure 8, the global calibration method for cameras without a common field of view for vehicle detection based on spatial points can be divided into the following six steps:

The first step: vehicle detection based on spatial points Image acquisition without a common field of view camera global calibration:

The main camera 1, the sub-camera left 3 and the sub-camera right 5 are respectively fixed on the main camera bracket 2, the sub-camera bracket left 4 and the sub-camera bracket right 6, and the main camera bracket 2, sub-camera bracket left 4 and sub-camera bracket right 6 is placed on the ground. According to the needs of car detection for a large detection range, the left 3 of the sub-camera and the right of the sub-camera 5 have no common field of view. In the field of view, the main camera 1 and the sub-camera left 3 collect an image respectively, move the cylindrical target 7 to make it in the field of view of the main camera 1 and the sub-camera right 5, and the main camera 1 and the sub-camera right 5 capture an image respectively. an image;

Step 2: When the cylindrical target 7 is in the common field of view of the total camera 1 and the left 3 of the sub-camera, calculate the homography matrix converted from the cylindrical target 7 to the coordinate system of the total camera 1 according to the image collected by the total camera 1:

The conversion relationship from the coordinate system of the cylindrical target 7 to the coordinate system of the image obtained by the total camera 1 is:

P _{T1, I0} X _T1 = sx _{I0, T1}

Using the RANSAC point extraction method and the DLT calibration method, the projection matrix P _{T1, I0} =K _{T1, I0} [R _{T1, I0} t _{T1, I0} ], K _{T1, I0} are the internal parameters of the total camera 1, R _{T1, I0} , t _{T1, I0} are the external parameters of the total camera 1 obtained according to the QR decomposition, X _T1 is the cylindrical target 7 coordinate system coordinates of the cylindrical target 7 feature point, x _{I0, T1} is the cylindrical target 7 feature point X _T1 in the total camera 1 The image coordinates under the acquired image, s is the scale factor, from the rotation matrix R _{T1, I0} and translation vector t _{T1, I0} can be obtained from the coordinate system of the cylindrical target 7 in the left 3 field of view of the sub-camera to the total camera 1 coordinate Homography matrix for system transformation

Step 3: When the cylindrical target 7 is in the common field of view of the total camera 1 and the sub-camera right 5, calculate the homography matrix converted from the cylindrical target 7 to the coordinate system of the total camera 1 according to the image collected by the total camera 1:

The conversion relationship between the coordinate system of the cylindrical target 7 and the image coordinate system obtained by the total camera 1 is:

P _{T2, I0} X _T2 =sx _{I0, T2}

Using the RANSAC point extraction method and the DLT calibration method, the projection matrix P _{T2, I0} =K _{T2, I0} [R _{T2, I0} t _{T2, I0} ], K _{T2, I0} are the internal parameters of the total camera 1, R _{T2, I0} , t _{T2, I0} are the external parameters of the total camera 1 obtained by QR decomposition, X _T2 is the cylindrical target 7 coordinate system coordinate of the cylindrical target 7 feature point, x _{I0, T2} is the cylindrical target 7 feature point X _T2 in the total camera 1 The image coordinates under the acquired image can be obtained from the rotation matrix R _{T2, I0} and the translation vector t _{T2, I0} from the coordinate system of the cylindrical target 7 in the right 5 field of view of the sub-camera to the coordinate system of the total camera 1. matrix

According to the obtained H _{T1, CO} and H _{T2, CO} , the homography matrix from the left 3 coordinate system of the sub-camera to the right 5 coordinate system of the sub-camera can be obtained

H _{T2, T1} = H _{T2, C0} (H _{T1, C0} ) ^-1

Expand H _{T2 and T1} further to get

Step 4: When the cylindrical target 7 is in the common field of view of the main camera 1 and the sub-camera left 3, the homography matrix converted from the cylindrical target 7 to the sub-camera left 3 coordinate system is calculated according to the image collected by the sub-camera left 3:

The conversion relationship between the coordinate system of the cylindrical target 7 and the image coordinate system obtained by the left 3 of the sub-camera is:

P _{T1, I1} X _T1 =sx _{I1, T1}

Using the RANSAC point extraction method and the DLT calibration method, the projection matrix P _{T1, I1} = K _{T1, I1} [R _{T1, I1} t _{T1, I1} ], K _{T1, I1} are the internal parameters of the left 3 of the sub-camera, R _{T1, I1} , t _{T1, I1} are the external parameters of the sub-camera left 3 obtained by QR decomposition, x _{I1, T1} are the image coordinates of the cylindrical target 7 feature point X _T1 under the image obtained by the sub-camera left 3, and the rotation matrix R _{T1 , I1} and translation vector t _{T1, I1} can obtain the homography matrix from the coordinate system of the cylindrical target 7 in the left 3 field of view of the sub-camera to the left 3 coordinate system of the sub-camera

Step 5: When the cylindrical target 7 is in the common field of view of the total camera 1 and the sub-camera right 5, calculate the homography matrix converted from the cylindrical target 7 to the sub-camera right 5 coordinate system according to the image collected by the sub-camera right 5:

The conversion relationship between the coordinate system of the cylindrical target 7 and the image coordinate system obtained by the right 5 of the sub-camera is:

P _{T2, I2} X _T2 = sx _{I2, T2}

Using the RANSAC point extraction method and the DLT calibration method, the projection matrix P _{T2, I2} =K _{T2, I2} [R _{T2, I2} t _{T2, I2} ], K _{T2, I2} are the internal parameters of the right 5 of the sub-camera, R _{T2, I2} , t _{T2, I2} are the external parameters of the right 5 of the sub-camera obtained by QR decomposition, x _{I2, T2} are the image coordinates of the feature point X _T2 of the cylindrical target 7 under the image obtained by the right 5 of the sub-camera, by the rotation matrix R _{T2 , I2} and translation vector t _{T2, I2} can obtain the homography matrix from the coordinate system of the cylindrical target 7 in the right 5 field of view of the sub-camera to the right 5 coordinate system of the sub-camera

Step 6: Calculate the homography matrix between the left 3 coordinate system of the sub-camera and the right 5 coordinate system of the sub-camera:

According to a series of homography matrices solved in the second to fifth steps, the homography matrix converted from the left 3 coordinate system of the sub-camera to the right 5 coordinate system of the sub-camera can be obtained

H _{C1, C2} = (H _{T2, C2} ) ^-1 H _{T2, T1} H _{T1, C1}

Expand the homography matrix H _{C1, C2} , the relationship between the homography matrix converted from the left 3 coordinate system of the sub-camera to the right 5 coordinate system of the sub-camera and the rotation matrices and translation vectors obtained in the second to fifth steps is:

Claims (9)

1. A space point-based global calibration system for a camera without a common view field for automobile detection is characterized by comprising a total camera (1), a total camera support (2), a branch camera left (3), a branch camera support left (4), a branch camera right (5), a branch camera support right (6) and a cylindrical target (7);

the main camera support (2), the branch camera support left (4), the branch camera support right (6) and the cylindrical target (7) are placed on the ground, the branch camera left (3) and the branch camera right (5) have no public view field, and the main camera (1), the branch camera left (3) and the branch camera right (5) are respectively fixedly connected with the main camera support (2), the branch camera support left (4) and the bolt thread at the top of the branch camera support right (6) through the threaded holes at the bottom.

2. The system for global calibration of cameras without common visual field for automobile inspection based on space points as claimed in claim 1 is characterized in that the total camera (1) is a wide-angle industrial camera.

3. The system for global calibration of cameras without common field of view for vehicle inspection based on space points as claimed in claim 1, wherein said total camera support (2) is a tripod with adjustable height.

4. The system for global calibration of cameras without common visual field for automobile inspection based on space points as claimed in claim 1, wherein the left side (3) of the sub-camera is a wide-angle industrial camera.

5. The system for global calibration of cameras without common field of view for automobile inspection based on space points as claimed in claim 1, wherein the left part (4) of the sub-camera support is a tripod with adjustable height.

6. The system for global calibration of common-view-field-free camera for automobile inspection based on spatial points as claimed in claim 1, wherein the right sub-camera (5) is a wide-angle industrial camera.

7. The system for global calibration of cameras without common field of view for vehicle inspection based on space points as claimed in claim 1, wherein said right branch camera support (6) is a tripod with adjustable height.

8. The global calibration system for the non-common-view-field camera for the automobile inspection based on the space points as claimed in claim 1, wherein the cylindrical target (7) is a hollow cylinder, and a checkerboard pattern is pasted on the outer surface.

9. The calibration method of the spatial point-based global calibration system for the automobile inspection camera without the common view field according to the claims 1 to 8 is characterized by comprising the following specific steps:

the first step is as follows: the method comprises the following steps of (1) acquiring an image of a common-view-field-free camera global calibration in the automobile detection based on space points:

the main camera (1), the branch camera left (3) and the branch camera right (5) are respectively fixed on the main camera support (2), the branch camera support left (4) and the branch camera support right (6), the main camera support (2), the branch camera support left (4) and the branch camera support right (6) are placed on the ground, according to the requirement of the automobile detection on a large detection range, the left (3) of the sub-camera and the right (5) of the sub-camera have no common view field, the cylindrical target (7) is placed on the ground, the cylindrical target (7) is moved to be positioned in the fields of view of the main camera (1) and the branch camera right (5), and the main camera (1) and the branch camera right (5) respectively acquire an image;

the second step is that: when the cylindrical target (7) is in the common field of view of the main camera (1) and the branch cameras (3), a homography matrix for converting the coordinate system of the cylindrical target (7) to the main camera (1) is calculated according to the image acquired by the main camera (1):

the conversion relation from the coordinate system of the cylindrical target (7) to the coordinate system of the image acquired by the total camera (1) is

P_T1，I0X_T1＝sx_I0，T1

Projection matrix P can be obtained by using RANSAC point extraction method and DLT calibration method_T1，I0＝K_T1，I0[R_T1，I0t_T1，I0]，K_T1，I0Is an internal parameter of the overall camera (1), R_T1，I0，t_T1，I0Is an external parameter, X, of the total camera (1) obtained from QR decomposition_T1Coordinates of the coordinate system of the cylindrical target (7) which are the characteristic points of the cylindrical target (7), x_I0，T1Is a cylinder target (7) characteristic point X_T1The coordinates of the image under the image acquired by the overall camera (1), s being a scaling factor, are given by a rotation matrix R_T1，I0And a translation vector t_T1，I0The homography matrix for the transformation from the coordinate system of the cylindrical target (7) under the view field of the branch camera left (3) to the coordinate system of the total camera (1) can be obtained

The third step: when the cylindrical target (7) is in the common field of view of the main camera (1) and the branch cameras (5), a homography matrix for converting the coordinate system of the cylindrical target (7) to the main camera (1) is solved according to the image acquired by the main camera (1):

the conversion relation between the coordinate system of the cylindrical target (7) and the coordinate system of the image acquired by the general camera (1) is

P_T2，I0X_T2＝sx_I0，T2

Projection matrix P can be obtained by using RANSAC point extraction method and DLT calibration method_T2，I0＝K_T2，I0[R_T2，I0t_T2，I0]，K_T2，I0Is an internal parameter of the overall camera (1), R_T2，I0，t_T2，I0Is an external parameter, X, of the total camera (1) obtained from QR decomposition_T2Coordinates of the coordinate system of the cylindrical target (7) which are the characteristic points of the cylindrical target (7), x_I0，T2Is a cylinder target (7) characteristic point X_T2The coordinates of the image under the image acquired by the overall camera (1) are determined by a rotation matrix R_T2，I0And a translation vector t_T2，I0The homography matrix for the transformation from the coordinate system of the cylindrical target (7) under the view field of the partial camera right (5) to the coordinate system of the total camera (1) can be obtained

According to the obtained H_T1，COAnd H_T2，COHomography matrix from left (3) coordinate system of the partial camera to right (5) coordinate system of the partial camera can be obtained

H_T2，T1＝H_T2，C0(H_T1，C0)^-1

H is to be_T2，T1Further spread out to obtain

The fourth step: when the cylindrical target (7) is in the common field of view of the total camera (1) and the branch cameras (3), calculating a homography matrix converted from the cylindrical target (7) to the left coordinate system of the branch cameras (3) according to the images acquired by the branch cameras (3):

the conversion relation between the coordinate system of the cylindrical target (7) and the image coordinate system obtained by the left camera (3) is

P_T1，I1X_T1＝sx_I1，T1

Projection matrix P can be obtained by using RANSAC point extraction method and DLT calibration method_T1，I1＝K_T1，I1[R_T1，I1t_T1，I1]，K_T1，I1Is an internal parameter of the left (3) of the partial camera, R_T1，I1，t_T1，I1Is an external parameter, x, of the left (3) of the partial camera obtained from QR decomposition_I1，T1Is a cylinder target (7) characteristic point X_T1Image coordinates in the image acquired by the left (3) of the partial camera are determined by the rotation matrix R_T1，I1And a translation vector t_T1，I1The homography matrix from the coordinate system of the cylindrical target (7) under the view field of the left camera (3) to the coordinate system of the left camera (3) can be obtained

The fifth step: when the cylindrical target (7) is in the common visual field of the total camera (1) and the branch cameras (5), calculating a homography matrix converted from the coordinate system of the cylindrical target (7) to the right (5) of the branch cameras according to the image acquired by the right (5) of the branch cameras:

the conversion relation between the coordinate system of the cylindrical target (7) and the image coordinate system obtained by the right side (5) of the sub-camera is

P_T2，I2X_T2＝sx_I2，T2

Projection matrix P can be obtained by using RANSAC point extraction method and DLT calibration method_T2，I2＝K_T2，I2[R_T2，I2t_T2，I2]，K_T2，I2Is an internal parameter of the right (5) of the partial camera, R_T2，I2，t_T2，I2Is an external parameter, x, of the right (5) of the partial camera obtained from QR decomposition_I2，T2Is a cylinder target (7) characteristic point X_T2The coordinates of the image obtained from the right (5) of the partial camera are determined by the rotation matrix R_T2，I2And a translation vector t_T2，I2The homography matrix from the coordinate system of the cylindrical target (7) under the view field of the right (5) of the sub-camera to the coordinate system of the right (5) of the sub-camera can be obtained

And a sixth step: and (3) solving a homography matrix between the left (3) coordinate system of the sub-camera and the right (5) coordinate system of the sub-camera:

according to a series of homography matrixes obtained by the solution of the second step to the fifth step, the homography matrix for converting the left (3) coordinate system of the sub-cameras into the right (5) coordinate system of the sub-cameras can be obtained

H_C1，C2＝(H_T2，C2)^-1H_T2，T1H_T1，C1

Homography matrix H_C1，C2The relationship between the homography matrix for expanding the conversion from the left (3) coordinate system of the scorable camera to the right (5) coordinate system of the scorable camera and each rotation matrix and translation vector obtained in the second step to the fifth step is

Images