Camera array extrinsic parameter initial value precision calibration method for spatial laser beam straight extrapolation

Abstract

The application discloses a camera array external parameter initial value accurate calibration method for spatial laser beam linear extrapolation, which comprises inputting an image sequence to be detected, performing accurate identification on a two-dimensional center line of an image laser beam under multi-modal constraint, and extracting a laser beam binary point set from the image sequence by adopting a target background method and a skeleton thinning method; the laser beam is separated to obtain a plurality of laser beam gray value point sets; a laser beam center line is extracted by adopting multi-modal constraint to obtain a laser beam two-dimensional center point set; a two-dimensional center line linear equation of a single laser beam is obtained by performing linear fitting on the laser center point set; and camera array external parameter calibration is performed based on the two-dimensional center line linear equation and a laser control line, wherein the application adopts a movable high-strength collimated laser beam as a high-precision control line, and realizes accurate calculation of external parameter initial values, camera array external parameter calibration of a large field of view and solution to laser beam diffusion and cross-interference by adopting an external parameter calibration method based on linear extrapolation.

Description

Technical Field

[0001] This invention relates to the field of microwave remote sensing geological disaster monitoring technology, and in particular to a method for accurate calibration of initial values ​​of camera array extrinsic parameters by linear extrapolation of a space laser beam. Background Technology

[0002] With the widespread application of computer vision and 3D measurement technologies in aerospace, industrial manufacturing, autonomous driving, and smart cities, the demand for large field-of-view (FOV) camera calibration technology is increasingly prominent. However, the widespread use of FAV cameras has also brought calibration challenges. First, the inherent distortion problem of camera lenses is significant, and traditional pinhole imaging models cannot accurately describe this highly nonlinear distortion, severely affecting measurement accuracy. Second, due to the large field of view coverage, control points or targets during the calibration process are often difficult to distribute uniformly and effectively across the entire field of view, leading to increased calibration errors. Furthermore, the complexity of the application environment for FAV cameras, such as variations in lighting conditions, interference from obstructions, and differences in target reflection, also greatly increases the challenge of camera calibration.

[0003] Currently, traditional small field-of-view calibration methods are no longer effective in addressing the problems that arise under large field-of-view conditions. By researching imaging models and calibration strategies that better suit the characteristics of large field-of-view cameras, we can not only effectively improve the measurement performance of vision systems but also promote the expansion of computer vision technology applications in complex environments. Furthermore, efficient and accurate calibration methods for large field-of-view cameras are expected to drive the development of intelligent unmanned systems, smart city monitoring, and industrial production line automation, bringing significant economic and social benefits.

[0004] Extrinsic parameter calibration of large field-of-view camera arrays is a prerequisite and fundamental issue for realizing the above-mentioned technologies. The core challenge lies in the accurate calibration of the initial frame extrinsic parameters of the camera array. In large field-of-view camera array extrinsic parameter calibration, factors such as inconsistent lighting conditions (differences in position and angle) and a lack of large-scale, high-precision calibration targets lead to high difficulty and low accuracy in calibration. Currently, large field-of-view camera array calibration methods mainly include planar target methods, three-dimensional control field methods, feature point calibration methods, self-calibration methods, and lidar calibration methods. Planar target methods obtain high-precision and reliable camera extrinsic parameters by geometrically transforming a checkerboard or dot array calibration board through imaging, but are limited by the size of the calibration board and cannot cover a large field of view. Three-dimensional control field methods establish mapping relationships using a control field with known three-dimensional coordinates, suitable for large field-of-view nonlinear distortion correction, but large calibration objects are complex to manufacture and difficult to deploy. Feature point calibration methods calculate the relative position of the camera by matching naturally or artificially labeled feature points. They do not require a calibration board and are suitable for large fields of view; however, matching accuracy is affected by image quality, and accuracy decreases when feature points are sparse. Self-calibration methods rely on camera motion or geometric constraints, do not require a calibration board, and are highly flexible, but have lower accuracy and extremely high requirements for scene structure. Deep learning methods estimate camera parameters from images using models such as CNNs, offering a high degree of automation, but require significant training and have limited generalization ability, thus limiting calibration accuracy. Therefore, a method for accurate initial value calibration of camera array extrinsic parameters using linear extrapolation of spatial laser beams is needed. Summary of the Invention

[0005] The purpose of this invention is to provide a method for accurately calibrating the initial values ​​of camera array extrinsic parameters by linear extrapolation of a spatial laser beam.

[0006] To achieve the above objectives, the present invention is implemented according to the following technical solution:

[0007] The first aspect of this invention provides a method for accurately calibrating the initial values ​​of camera array extrinsic parameters by linear extrapolation of a space laser beam, comprising the following steps:

[0008] Input the image sequence to be detected.

[0009] Accurate identification of the two-dimensional centerline of an image laser beam under multimodal constraints, specifically including:

[0010] a1 uses the target background method and skeleton thinning method to extract the laser beam binarized point set from the image sequence;

[0011] a2 establishes a method for determining intersection points with connectivity and neighborhood density constraints, dividing the binarized point set of the laser beam into intersection points and non-intersection points;

[0012] The Jenks slope clustering method under the constructed topology is used to separate the laser beams and obtain multiple sets of gray value points of the laser beams;

[0013] a4 uses the multimodal constrained Steger algorithm to extract the laser beam centerline, obtains a two-dimensional center point set of the laser beam, and performs linear fitting on the laser center point set to obtain the two-dimensional centerline linear equation of a single laser beam;

[0014] The alignment of the laser control line camera array extrinsic parameter calibration based on the aforementioned two-dimensional centerline linear equation specifically includes:

[0015] b1 uses the measured three-dimensional laser line, the identified two-dimensional laser center line, the camera parameter matrix and the collinearity condition equation to construct the external parameter calibration equation for a large field of view multi-camera array based on the extrapolation of the straight line B2, and solves the initial value of the external parameters of the camera array in the first frame.

[0016] b3 uses the image-side line of the camera array and the reprojection spatial distance error to construct a function that minimizes the reprojection error of the camera array line;

[0017] b4 uses the Levenberg-Marquardt nonlinear optimization method to iteratively optimize the extrinsic parameters of the camera array, and obtains the accurate extrinsic parameter values ​​of the first frame of the large field-of-view camera array.

[0018] Further, in step a2, the intersection point determination method uses the connectivity function A(P) and the neighborhood density function B(P) for determination. When both function A(P) and function B(P) satisfy the set conditions, it is determined to be an intersection point.

[0019]

[0020] Where m is the number of emitted laser beams, P(x i ,y i P'(x) is the set of grayscale points of the laser beam. i ,y i A(P) is an optimized binarized point set of the laser beam, optimized using the target background method connected component analysis and skeleton refinement method. i ') is a connectivity function, B(P) i ') is the neighborhood density function, P j 'It's like point P' i 'The surrounding 8 neighboring image points, when B(P i ')>6 and A(P i When ')≥3, it is determined to be an intersection point.

[0021] Furthermore, the slope clustering method calculates the minimum variance and determines the optimal split point through dynamic recursion, as shown in the formula:

[0022]

[0023] Where D(n,m) represents dividing n slopes into m classes, D(t,m-1) is the dynamic recursive process, t is the initial value of the temporary split point, and Var(kt +1,…,k n Let be the slope variance at time t. By dynamically recursively calculating the sum of variances at time t, the split point corresponding to the minimum sum of variances is the optimal split point.

[0024] Furthermore, the method for extracting image points along a single laser centerline based on the Steger algorithm constrained by the deviation of the normal direction angle and the slope direction specifically includes:

[0025]

[0026] Among them, P o (x i ,y i Let be the set of points along the center line of the o-th laser beam. Let θ be the normal direction of the i-th image point. th The threshold for determining the angle of the normal direction, k i Let k be the slope of the i-th image point, k′ be the slope cluster, and Δk be the slope of the i-th image point. max The slope direction deviation threshold;

[0027] Normal direction angle threshold θ th The threshold Δk of the slope direction deviation max The calculation formula is as follows:

[0028] θ th =arcsin(κ) local ·d),Δk max =κ local ·d (5)

[0029] Where d is the pixel spacing, κ local For maximum curvature;

[0030] Based on the RANSAC-improved least squares method, the laser center point set P is... o (x i ,y i By performing a straight line fitting, the two-dimensional centerline equation of a single laser beam is obtained:

[0031] y = a o x+b o (6)

[0032] Furthermore, we construct a function E that minimizes the linear reprojection error of the camera array:

[0033]

[0034] Where m is the number of paired laser lines.

[0035] Furthermore, the Levenberg-Marquardt nonlinear optimization method is used to iteratively optimize the rotation matrix R and the translation vector T:

[0036]

[0037] T (k) =T (k-1) +ΔT (13)

[0038] Where Δθ is the Lie algebraic form of the rotation vector increment of the rotation matrix R, ΔT is the translation vector increment, J is the Jacobian matrix of the error with respect to the optimization variables, and λ is the step size. If the gradient error J in the k-th iteration is... T E is less than 10 -6 When the iteration stops, output the optimal values ​​of the extrinsic parameters of the camera array in the first frame, including the rotation matrix R and the translation vector T.

[0039] Secondly, embodiments of this application also provide an electronic device, including: a processor; and a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform the steps of the method described in the first aspect.

[0040] Thirdly, embodiments of this application also provide a computer-readable storage medium storing one or more programs that, when executed by an electronic device including multiple applications, cause the electronic device to perform the steps of the method described in the first aspect.

[0041] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects:

[0042] (1) This invention effectively suppresses the errors caused by background light interference and laser beam intersection by constructing a laser beam intersection determination method with connectivity and neighborhood density constraints and an improved Steger algorithm with joint constraints of normal angle and slope direction deviation, and significantly improves the stability and robustness of laser beam two-dimensional center line extraction.

[0043] (2) This invention effectively reduces the accumulation of multi-view matching errors, and further uses the Levenberg-Marquardt nonlinear optimization technique to perform iterative optimization of the rotation matrix and translation vector with the two-dimensional straight line reprojection error of the laser beam as the optimization target, and finally achieves high-precision calibration of the large field-of-view camera array.

[0044] (3) By constructing a collimated laser beam as a high-precision spatial constraint, this invention can not only adapt to image laser beam extraction under complex environmental conditions, but also effectively solve the problem of multi-view matching difficulties for large field-of-view camera arrays. It has strong versatility and scalability, and is suitable for precision measurement and calibration tasks of various complex environments and multi-view camera arrays. Attached Figure Description

[0045] Figure 1This is a flowchart illustrating the steps of the method for accurately calibrating the initial values ​​of camera array extrinsic parameters using linear extrapolation of a space laser beam, as described in this invention.

[0046] Figure 2 This is a schematic diagram of the structure of an electronic device in an embodiment of this specification. Detailed Implementation

[0047] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0048] Reference Figure 1 As shown, this invention provides a method for accurately calibrating the initial values ​​of camera array extrinsic parameters by linear extrapolation of a spatial laser beam, comprising:

[0049] Input the image sequence to be detected.

[0050] Accurate identification of the two-dimensional centerline of an image laser beam under multimodal constraints, specifically including:

[0051] a1 uses the target background method and skeleton thinning method to extract the laser beam binarized point set from the image sequence;

[0052] a2 establishes a method for determining intersection points with connectivity and neighborhood density constraints, dividing the binarized point set of the laser beam into intersection points and non-intersection points;

[0053] The Jenks slope clustering method under the constructed topology is used to separate the laser beams and obtain multiple sets of gray value points of the laser beams;

[0054] a4 uses the multimodal constrained Steger algorithm to extract the laser beam centerline, obtains a two-dimensional center point set of the laser beam, and performs linear fitting on the laser center point set to obtain the two-dimensional centerline linear equation of a single laser beam;

[0055] The alignment of the laser control line camera array extrinsic parameter calibration based on the aforementioned two-dimensional centerline linear equation specifically includes:

[0056] b1 uses the measured three-dimensional laser line, the identified two-dimensional laser center line, the camera parameter matrix and the collinearity condition equation to construct the external parameter calibration equation for a large field of view multi-camera array based on the extrapolation of the straight line B2, and solves the initial value of the external parameters of the camera array in the first frame.

[0057] b3 uses the image-side line of the camera array and the reprojection spatial distance error to construct a function that minimizes the reprojection error of the camera array line;

[0058] b4 uses the Levenberg-Marquardt nonlinear optimization method to iteratively optimize the extrinsic parameters of the camera array, and obtains the accurate extrinsic parameter values ​​of the first frame of the large field-of-view camera array.

[0059] (1) Accurate identification of the two-dimensional centerline of the image laser beam under multimodal constraints

[0060] Accurate identification of the 2D centerline of laser beams in images under multimodal constraints comprises two stages: laser beam separation using Jenks' slope clustering method and extraction of the 2D centerline of laser beams under multimodal constraints. These stages are crucial steps for accurately identifying the 2D centerline of laser beams in complex environments, significantly impacting the accuracy and efficiency of subsequent camera array extrinsic calibration and 3D coordinate calculation. During image capture, the intensity of ambient light and interference from adjacent laser beams can severely affect the acquired image data. When ambient light is strong, the extraction of the laser centerline may be affected by the background light intensity, leading to centerline extraction failure. Simultaneously, interference between adjacent laser beams can also cause laser centerline extraction failure. To address this issue, a multimodal globally optimized method for accurate identification of the 2D centerline of laser beams is proposed. The specific steps are as follows:

[0061] 1) Define m (≥6) as the number of emitted laser beams, P(x i ,y i P'(x) is the set of grayscale points of the laser beam. i ,y i The optimized laser beam binarized point set is obtained by using the target background method for connected component analysis and the skeleton refinement method.

[0062] 2) Establish a laser beam intersection point determination method with connectivity and neighborhood density constraints, which can divide the optimized laser beam binarized point set into two categories: intersection points and non-intersection points.

[0063]

[0064]

[0065] Among them, A(P) i ') is a connectivity function, B(P) i ') is the neighborhood density function, P j 'It's like point P' i 'The surrounding 8 neighboring image points, when B(P i ')>6 and A(P i When ')≥3, it is determined to be an intersection point, thus constructing an intersection point set P consisting of m laser beams. c (x i ,y i ).

[0066] 3) Establish P cGiven the topology of (xi,yi), a skeleton path search method is used to filter out simply connected path nodes, and the slope clusters k = {k1,k2,…k} between the simply connected path nodes are calculated. n A laser beam slope cluster separation method based on Jenks' natural breakpoint clustering of simply connected paths is constructed.

[0067]

[0068] Where D(n,m) represents dividing n slopes into m classes, D(t,m-1) is the dynamic recursive process, t is the initial value of the temporary split point, and Var(k t+1 ,…,k n Let be the slope variance at time t. By dynamically recursively calculating the sum of variances at time t, the segmentation point corresponding to the minimum sum of variances is the optimal segmentation point. Thus, a slope cluster k' = {k'1, k'2, ..., k''} is obtained, consisting of the slopes of m laser beams. m Finally, based on k', the laser beam grayscale value set P(x) is... i ,y i Separate into m sets of laser beam grayscale points

[0069] 4) To address the issue of sudden grayscale changes in certain areas of the image caused by environmental interference, which leads to errors in the extraction of the two-dimensional centerline of the laser beam, the set of grayscale points of the m laser beams separated in step 3) is used. Based on this, a method for extracting single laser centerline image points using the Steger algorithm, which is jointly constrained by the deviation of the normal direction angle and the slope direction, is constructed:

[0070]

[0071] Among them, P o (x i ,y i Let be the set of points along the center line of the o-th laser beam. Let θ be the normal direction of the i-th image point. th The threshold for determining the angle of the normal direction, k i Let Δk be the slope of the i-th image point, and k′ be the slope cluster calculated in step 3). max This is the slope direction deviation threshold.

[0072] Normal direction angle threshold θ th The threshold Δk of the slope direction deviation max The calculation formula is as follows:

[0073] θ th =arcsin(κ) local ·d),Δk max =κ local ·d (5)

[0074] Where d is the pixel spacing (usually 1 pixel), κ local This represents the maximum curvature.

[0075] Finally, the least squares method based on RANSAC improvement is used to analyze the laser center point set P. o (x i ,y i By performing a straight line fitting, the two-dimensional centerline equation of a single laser beam is obtained:

[0076] y = a o x+b o (6)

[0077] It provides accurate two-dimensional linear information for camera array extrinsic parameter calibration.

[0078] (2) Precise global calibration of extrinsic parameters of quasi-laser control line camera array

[0079] To address the issue of accumulated straight-line matching errors caused by multi-view shooting with camera arrays, a method for extrinsic parameter calibration of large field-of-view multi-camera arrays based on straight-line extrapolation is constructed. Through synchronous calculation and iterative optimization, accurate calibration of the extrinsic parameters of the first frame of a large field-of-view camera array is achieved. The specific steps are as follows:

[0080] 1) Using measured three-dimensional space laser straight lines The identified laser two-dimensional center line y = a o x+b o Camera parameter matrix M(m0~m 11 Based on the collinearity condition equation, a method for extrinsic parameter checking of a large field-of-view multi-camera array is constructed using linear extrapolation, as follows:

[0081]

[0082] Construct a set of parametric equations for the camera array projection matrix, as shown in formula (7). Based on the mathematical relationship between the rotation matrix R, the translation vector T, the intrinsic parameter matrix K, and the projection matrix M, solve for the initial values ​​of the extrinsic parameters of the camera array in the first frame.

[0083]

[0084] Where, matrix A l and A r These are the coefficient matrices for the left and right cameras, respectively, M. l and M r These are the projection parameters for the left and right cameras, respectively.

[0085] 2) To address the issue of discrepancies in laser 2D centerline extraction due to varying shooting angles of multiple camera arrays, the method uses the image-side straight line of the camera array and the reprojection spatial distance error d(l) as the basis for the method. o ,Lo ) l and d(l o ,L o ) r Construct the function E that minimizes the linear reprojection error of the camera array:

[0086]

[0087] Where m is the number of paired laser lines.

[0088] By calculating the two endpoints p of the two-dimensional straight line of the laser beam s ,p e The perpendicular distance to the reprojection line l′ is used to obtain the distance d(l) between the image-side line and the line in the reprojection space. o ,L o ):

[0089]

[0090] Where, p s =[x1,y1,1] T p e = [x2, y2, 1] T l′=[l1,l2,l3] T This is the representation of the reprojected line on the imaging plane. The rotation matrix R and translation vector T are iteratively optimized using the Levenberg-Marquardt nonlinear optimization method:

[0091]

[0092] T (k) =T (k-1) +ΔT (13)

[0093] Where Δθ is the Lie algebraic form of the rotation vector increment of the rotation matrix R, ΔT is the translation vector increment, J is the Jacobian matrix of the error with respect to the optimization variables, and λ is the step size. If the gradient error J in the k-th iteration... T E is less than 10 -6 When the iteration stops, output the optimal values ​​of the extrinsic parameters of the camera array in the first frame, including the rotation matrix R and the translation vector T.

[0094] Figure 2 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Please refer to it. Figure 2At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and memory. The memory may include main memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for other business operations.

[0095] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. This bus can be categorized into address bus, data bus, control bus, etc.

[0096] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0097] The processor reads the corresponding computer program from non-volatile memory into memory and then runs it, forming a device for constructing a virtual reality-based intelligent product simulation test scenario at the logical level. The processor executes the program stored in memory and is specifically used to execute any of the aforementioned methods for accurately calibrating the initial values ​​of camera array extrinsic parameters through linear extrapolation of spatial laser beams.

[0098] This invention can be applied to or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0099] This application also proposes a computer-readable storage medium that stores one or more programs, the programs including instructions that, when executed by an electronic device including multiple applications, perform any of the aforementioned methods for accurate initial value calibration of camera array extrinsic values ​​for linear extrapolation of spatial laser beams.

[0100] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

Claims

1. A method for accurately calibrating the initial values ​​of camera array extrinsic parameters by linear extrapolation of a spatial laser beam, characterized in that, Includes the following steps: Input the image sequence to be detected. Accurate identification of the two-dimensional centerline of an image laser beam under multimodal constraints, specifically including: a1 uses the target background method and skeleton thinning method to extract the laser beam binarized point set from the image sequence; a2 establishes a method for determining intersection points with connectivity and neighborhood density constraints, dividing the binarized point set of the laser beam into intersection points and non-intersection points; The Jenks slope clustering method under the topological structure constructed by a3 is used to separate the laser beams and obtain multiple sets of gray value points of the laser beams; a4 uses the multimodal constrained Steger algorithm to extract the laser beam centerline, obtains a two-dimensional center point set of the laser beam, and performs straight line fitting on the laser center point set to obtain the two-dimensional centerline straight line equation of a single laser beam; The alignment of the laser control line camera array extrinsic parameter calibration based on the aforementioned two-dimensional centerline linear equation specifically includes: b1 uses the measured three-dimensional laser line, the identified two-dimensional laser center line, the camera parameter matrix and the collinearity condition equation to construct the external parameter calibration equation for a large field of view multi-camera array based on the extrapolation of the straight line B2, and solves the initial value of the external parameters of the camera array in the first frame. b3 uses the image-side line of the camera array and the reprojection spatial distance error to construct a function that minimizes the reprojection error of the camera array line; b4 employs the Levenberg-Marquardt nonlinear optimization method to iteratively optimize the extrinsic parameters of the camera array, obtaining accurate extrinsic parameter values ​​for the first frame of the large field-of-view camera array; The slope clustering method calculates the minimum variance and determines the optimal split point through dynamic recursion, using the following formula: in, Indicates will The slope is divided into kind, It is a dynamic recursive process. Initial value for temporary split point, yes The slope variance at time is calculated through dynamic recursion. If the sum of variances over time is the minimum sum of variances, then the split point corresponding to the minimum sum of variances is the optimal split point. The method for extracting single laser centerline image points based on the Steger algorithm constrained by the deviation of normal direction angle and slope direction specifically includes: in, For the first The centerline point set of the laser beam For the first Image point normal direction, The threshold for determining the angle of the normal direction. For the first The slope of a pixel, For slope clusters, The slope direction deviation threshold; Normal direction angle threshold Deviation threshold from slope direction The calculation formula is as follows: in, For pixel spacing, For maximum curvature; Based on the RANSAC-improved least squares method, the laser center point set is... By performing a line fitting, the two-dimensional centerline equation of a single laser beam is obtained: 。 2. The method for accurate initial value calibration of camera array extrinsic parameters for linear extrapolation of a spatial laser beam according to claim 1, characterized in that, In step a2, the intersection point determination method uses a connectivity function A(P) and a neighborhood density function B(P) for determination. When both functions A(P) and B(P) satisfy a set condition, it is determined to be an intersection point. in, It is the set of grayscale points of the laser beam. It is an optimized binarized point set of the laser beam, optimized using the target background method connected component analysis and skeleton refinement method. It is a connectivity function. It is the neighborhood density function. It's like a dot. The surrounding 8 neighboring image points, when and The point is determined to be an intersection.

3. The method for accurate initial value calibration of camera array extrinsic parameters for linear extrapolation of a spatial laser beam according to claim 1, characterized in that, Construct a function to minimize the line reprojection error of the camera array : in, It is the number of paired laser lines.

4. The method for accurate initial value calibration of camera array extrinsic parameters for linear extrapolation of a spatial laser beam according to claim 1, characterized in that, The rotation matrix is ​​iteratively optimized using the Levenberg-Marquardt nonlinear optimization method. Translation vector : in, It is a rotation matrix The Lie algebraic form of the rotation vector increment, It is the translation vector increment. It is the Jacobian matrix of the error with respect to the optimization variables. It is the step size, if the first Gradient error of the next iteration Less than When the iteration stops, output the optimal values ​​of the extrinsic parameters of the camera array in the first frame, and the rotation matrix. Translation vector .

Images