A method and device for fast and high-precision ground calibration of a long-focus large-aperture optical camera

Abstract

The present invention discloses a method and device for fast and high-precision calibration of a telephoto large-aperture optical camera on the ground, including: combining a plane target with a collimator, supplemented by a beam splitter and a plane reflector to construct an infinitely distant target, and establishing a ground high-precision calibration mathematical model; deeply analyzing the method for positioning image feature points, and providing a mathematical solution for positioning uncertainty; adopting the construction of an optimal spatial calibration field and weighted nonlinear optimization based on image positioning uncertainty, which has the advantages of high calibration efficiency and avoiding falling into local minima; a multi-parameter decoupling method based on center standardization, which reduces the coupling of internal and external parameters and can effectively improve the calibration accuracy; establishing a calibration system device to achieve fast and high-precision calibration of telephoto large-aperture optical cameras on the ground. The present invention is suitable for fast and high-precision calibration of telephoto large-aperture optical cameras in a ground laboratory environment, and is of great significance for ensuring high-resolution imaging and high-precision positioning in fields such as aerospace remote sensing and mapping.

Description

Technical Field

[0001] The present invention relates to the technical field of optical sensor calibration, and in particular to a method and device for fast and high-precision ground calibration of a long-focus large-aperture optical camera. Background Art

[0002] Long-focus, large-aperture, high-resolution optical imaging technology has the characteristics of large range, long distance, and high precision. It is widely used in aerospace remote sensing, space-based surveillance, aerial mapping, border and coastal defense warning, urban security and other fields, and has extremely important strategic needs and broad market application prospects. However, "seeing clearly" does not mean "precise positioning". Due to the complexity of the observation environment and the limitations of the technical level, the current high-resolution satellite data quality, especially the geometric quality, is often difficult to meet the geometric positioning accuracy requirements in practical applications, which greatly restricts its practical application and service capabilities.

[0003] Due to the limitations of optical lens design and processing technology, the actual application of the lens cannot comply with the ideal imaging model to maintain linear transformation. There is always a certain amount of nonlinear distortion, which makes the actual imaging point deviate from the ideal imaging point. This error is particularly prominent for telephoto cameras, seriously reducing their spatial positioning accuracy. Therefore, before the telephoto optical camera is put into actual use, the camera internal parameters must be calibrated to ensure high-precision measurement. Especially for aviation and aerospace optical cameras, ground calibration before launch is the basis for high-precision detection of telephoto large-aperture optical systems. It is extremely important to ensure high-precision detection after entering orbit, and also provides initial parameters for subsequent on-orbit calibration. In order to improve its positioning accuracy, the first problem that needs to be solved is the high-precision calibration of the geometric distortion of the optical camera (Yang Junfeng, An Hong, Gao Lianyi. Contents and methods of calibration of aerospace optical measurement cameras [J]. Surveying and Mapping Science and Engineering, 2010, 000(001): 66-70.). However, as the focal length, aperture and volume of cameras become larger and larger, the existing methods have higher and higher requirements for test equipment and sites. The idea of ​​simply increasing the size and accuracy of the equipment to achieve high-precision geometric distortion calibration is no longer applicable.

[0004] At present, most telephoto cameras use the precision angle measurement method for ground calibration. Wu Guodong et al. (Wu Guodong, Han Bing, He Xu. Laboratory calibration method for geometric parameters of linear array CCD cameras based on precision angle measurement method [J]. Optics and Precision Engineering, 2007(10):1628-1632.) used a high-precision two-dimensional precision turntable and parallel light tube to calibrate the linear array camera with high precision. The calibration accuracy of the principal distance and principal point can reach um level. Yuan Guoqin et al. (Yuan Guoqin, Ding Yalin, Hui Shouwen, Liu Liguo, Yu Chunfeng. Grouped progressive calibration algorithm for surveying and mapping cameras based on precision angle measurement method [J]. Acta Optica Sinica, 2012, 32(01):134-139.) proposed a grouped asymptotic algorithm based on the precision angle measurement method. The grouped asymptotic approximation and weight theory were used to improve the precision angle measurement method, avoiding the influence of theoretical errors and sampling points on the algorithm accuracy, and improving the calibration accuracy. Foreign aerospace sensors, such as MISR and POLDER, also use the precision angle measurement method for calibration. The precision angle measurement method is intuitive, simple and easy to implement, but it has strict requirements on the calibration environment. The accuracy of the turntable directly determines the calibration accuracy. In addition, the production of a high-precision turntable is expensive and time-consuming. The precision angle measurement method requires a lot of manual intervention and adjustment, and it is impossible to achieve rapid calibration.

[0005] The calibration method based on the three-dimensional calibration field is often used for on-orbit calibration of remote sensing cameras. Lin Zongjian et al. (Lin Zongjian, Xuan Wenling, Sun Jie, Wang Dong. A wide-angle aerial camera composed of a small array CCD camera [J]. Surveying and Mapping Science, 2005(01):94-98+112.) used a standard three-dimensional calibration field to measure the internal orientation elements and optical distortion coefficients of the array CCD camera and obtained satisfactory results. Feng Wenhao (Feng Wenhao, Shang Haoliang, Hou Wenguang. Digital distortion model of images [J]. Journal of Wuhan University (Information Science Edition), 2006 (02): 99-103.), Wang Dong (Wang Dong. CCD camera calibration based on multi-slice spatial resection [D]. Shandong University of Science and Technology, 2003.) et al. established indoor calibration fields for close-range photogrammetry cameras at Wuhan University and Shandong University of Science and Technology, realizing the precise determination of camera internal orientation elements and optical distortion parameters; the three-dimensional control fields used for camera geometric calibration internationally mainly include the Fredrikstad test field in Norway, the SSC test field of NASA in the United States, the Vaihningen / Enz test field in Germany, the Sjokulla test field in Finland, and the Elchingen and Herbrugg geometric calibration fields of aerial photography manufacturers. Although the three-dimensional calibration field method can guarantee calibration accuracy, it is expensive and difficult to establish a high-precision three-dimensional calibration field, and it is easily limited by the site and experimental conditions.

[0006] At present, the key technical difficulties that restrict the ground calibration of telephoto and large-aperture optical cameras mainly include: (1) Complex calibration mode: The existing laboratory high-precision calibration method based on a three-axis turntable has strict requirements on the test equipment and test environment, is time-consuming to set up, has low calibration efficiency, and has uncontrollable system / manual errors. In addition, the angle measurement accuracy directly determines the calibration accuracy. (2) Uneven accuracy of image centroid extraction: The target imaging quality is uneven, and the image positioning accuracy is uneven. The longer the focal length, the more sensitive the calibration result is to image noise, that is, a small positioning deviation can lead to a large calibration error. (3) Slow optimization convergence speed: The optimization process is prone to gradient vanishing, falling into the local minimum, and unable to obtain the global optimal solution. (4) Multi-parameter coupling: The coupling between internal and external multi-parameters is serious, which affects the calibration accuracy. (5) The cost of telephoto and large-aperture camera calibration is high: The calibration of telephoto and large-aperture cameras requires a larger turntable and a collimator with a longer focal length and a larger field of view. Therefore, overcoming the above key technical difficulties and studying a fast and high-precision calibration method for long-focus and large-aperture optical cameras on the ground is of great significance to ensuring high-resolution imaging and high-precision positioning of optical payloads in aerospace and other fields, and is an important key technology that needs to be solved urgently.

[0007] As the focal length, aperture and volume of cameras become larger and larger, the requirements for test equipment and sites are also getting higher and higher. It is difficult to achieve high-precision camera distortion calibration by simply increasing the volume and accuracy of the equipment. Therefore, it is urgent to break through the traditional camera ground calibration technology based on the turntable precision angle measurement method, and study a fast and high-precision ground calibration system and method suitable for long-focus large-aperture optical cameras, fill the gap in the relevant system in this field in my country, and provide technical support for high-resolution and high-precision long-distance detection and application. Summary of the invention

[0008] In order to solve the above technical problems, the present invention provides a method and device for fast and high-precision ground calibration of a telephoto large-aperture optical camera, which has the characteristics of compact structure, high calibration efficiency, strong versatility, simple system composition, etc., and can realize fast and high-precision ground calibration of a telephoto large-aperture optical camera in a laboratory.

[0009] In order to achieve the above object, the present invention adopts the following technical scheme:

[0010] A method for fast and high-precision ground calibration of a telephoto large-aperture optical camera comprises the following steps:

[0011] Step a: In a ground laboratory environment, a high-precision ground calibration hardware device for a telephoto large-aperture optical camera is constructed, and only a plane target is used in combination with a collimator, supplemented by a beam splitter and a plane reflector to achieve high-precision ground calibration of the telephoto large-aperture optical camera;

[0012] Step b: Modeling and extracting image feature points, using a high-precision image feature point extraction method based on an optimal scale to achieve high-precision extraction of feature points under different scale conditions;

[0013] Step c, by solving the uncertainty modeling based on the analysis of the corresponding internal parameters in the Jacobi matrix, the influence of the target space calibration field on the internal parameter calibration accuracy is obtained, and the optimal space calibration field is selected and constructed;

[0014] Step d, by establishing a mathematical model for local positioning of feature points, a mathematical solution between the uncertainty of image feature point positioning and image noise is obtained;

[0015] Step e: using image feature point positioning uncertainty weighting and center normalization method to avoid falling into local minimum and reduce coupling between multiple parameters;

[0016] Step f, swing the plane reflector to change the incident direction of the second group of light rays, cover the entire large-aperture optical camera field of view for imaging, and substitute the above steps in sequence to finally solve and obtain the internal parameters of the optical camera.

[0017] Furthermore, in step a, the infinite imaging target is designed as a miniature plane target installed on the focal plane of the collimator to generate an infinite plane target, which is divided into two groups of infinite target rays by a beam splitter. The first group directly enters the camera to be calibrated, and the second group enters the field of view of the camera to be calibrated synchronously after being reflected by a reflector.

[0018] Furthermore, in the step b, an adaptive multi-scale centroid extraction method is used to extract the centroid of feature points, so that the centroid extraction of different image feature points is completed under the optimal scale condition, thereby improving the accuracy of feature point centroid extraction.

[0019] Furthermore, in the step d, the positioning deviation of each feature point is quantitatively analyzed, and the relationship between the image feature point positioning accuracy and the image noise is solved by analyzing the local feature point positioning of the image and the image noise model, and the mathematical solution of the image feature point positioning uncertainty is obtained, laying the foundation for the subsequent image positioning uncertainty weighting operation.

[0020] Furthermore, in the step e, the image positioning uncertainty is weighted by the Jacobi matrix in the camera intrinsic parameter calibration optimization to solve the problems of gradient vanishing, slow convergence speed and easy falling into local minimum; the central normalization method is introduced to reduce the coupling of internal and external parameters in the calibration process, so as to achieve fast and high-precision calibration of the telephoto camera intrinsic parameters.

[0021] The present invention also provides a fast and high-precision ground calibration device for a telephoto large-aperture optical camera, comprising: an image analysis and calculation module, a target imaging and image acquisition module, and an infinite target module;

[0022] The image analysis and calculation module includes a computer, which is used to process the images of the target imaging and image acquisition modules, and to optimize and solve the camera internal parameters according to the camera internal parameter model;

[0023] The target imaging and image acquisition module includes a high-resolution camera and a control box, wherein the high-resolution camera includes a CCD imaging large target surface and a telephoto large-aperture optical lens, and the control box triggers the high-resolution camera to image the infinitely distant target;

[0024] The infinity target module includes a plane target, a collimator, a beam splitter prism and a plane reflector. The plane target is installed at the focal plane of the collimator, and after being illuminated by a backlight source, it passes through the collimator and enters the beam splitter prism. One beam of light directly enters a telephoto large-aperture optical lens, and the other beam of light passes through the plane reflector and enters the telephoto large-aperture optical lens.

[0025] The beneficial effects of the present invention compared with the prior art are:

[0026] The present invention proposes a method and device for fast and high-precision calibration of a telephoto large-aperture optical camera on the ground. This project intends to adopt a telephoto camera ground calibration mode based on a plane target combined with a collimator to construct an infinite imaging target, and designs a large-field-of-view imaging and calibration system that combines a beam splitter and a plane reflector. The measurement system consists of an infinite plane target, a beam splitter and a reflector. By freely moving the plane reflector, an infinite target imaging covering the entire field of view can be constructed. Therefore, the system has the characteristics of compact structure, high calibration efficiency, strong versatility, and simple system composition. It is very suitable for realizing high-precision calibration of telephoto large-aperture optical cameras on the ground in the laboratory. The calibration method proposed in this project not only provides an effective means for high-precision calibration of telephoto large-aperture optical cameras on the ground, but also has important theoretical significance and practical value for promoting the verification and calibration of large-scale telephoto optical systems and further improving the processing and quality of optical systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a flow chart of a method for fast and high-precision ground calibration of a telephoto and large-aperture optical camera according to the present invention;

[0028] Figure 2 Schematic diagram for fast and high-precision calibration of telephoto and large-aperture optical cameras on the ground;

[0029] Figure 3 Schematic diagram of the optimal calibration field distribution of the spatial target; among them, Figures (a), (b), (c), (e), (f), (h), and (i) are images within the optimal calibration field, and Figures (d), (g), and (j) are images within the non-optimal calibration field. DETAILED DESCRIPTION

[0030] The basic idea of ​​the present invention is to combine a plane target with a collimator, and construct an infinite target with the aid of a beam splitter and a plane reflector, and establish a ground high-precision calibration mathematical model; analyze the image feature point positioning method, and give a mathematical solution to the positioning uncertainty; adopt the construction of an optimal spatial calibration field and weighted nonlinear optimization based on image positioning uncertainty; a multi-parameter decoupling method based on center standardization can reduce the coupling of internal and external parameters, and can effectively improve the calibration accuracy; and by establishing a calibration system prototype, the ground fast and high-precision calibration of a telephoto large-aperture optical camera can be achieved.

[0031] like Figure 1 As shown, the present invention provides a method for fast and high-precision ground calibration of a telephoto and large-aperture optical camera, comprising the following steps:

[0032] Step 11: In a ground laboratory environment, build a high-precision ground calibration hardware device for a telephoto and large-aperture optical camera.

[0033] like Figure 2 As shown, the device for implementing the present invention includes an image analysis and calculation module, a target imaging and image acquisition module and an infinite target module. The image analysis and calculation module includes a computer for processing the image of the target imaging and image acquisition module, and optimizing and solving the camera internal parameters according to the camera internal parameter model. The target imaging and image acquisition module includes a high-resolution camera and a control box, wherein the high-resolution camera includes a CCD imaging large target surface and a telephoto large-aperture optical lens, and the control box triggers the high-resolution camera to image the infinite target. The infinite target module includes a plane target, a collimator, a beam splitter prism and a plane reflector. The plane target is installed at the focal plane of the collimator, and after being illuminated by a backlight source, it passes through the collimator and enters the beam splitter prism. One beam of light directly enters the telephoto large-aperture optical lens, and the other beam of light passes through the plane reflector and enters the telephoto large-aperture optical lens. The device adopts a plane target in combination with a collimator, supplemented by a beam splitter prism and a plane reflector to achieve high-precision ground calibration of a telephoto large-aperture optical camera. The infinite imaging target is designed as a miniature plane target installed on the focal plane of the collimator to generate an infinite plane target. The infinite target rays are divided into two groups by the beam splitter. The first group directly enters the camera to be calibrated, and the second group enters the field of view of the camera to be calibrated synchronously after being reflected by the reflector. The swinging plane reflector can change the incident direction of the second group of light rays, which can cover the entire field of view of the large-aperture optical camera.

[0034] Step 12: Image feature point modeling and extraction: A high-precision image feature point extraction method based on the optimal scale is used to achieve high-precision extraction of feature points under different scale conditions.

[0035] Let I(u,v) be the grayscale distribution function of the image, and any pixel p in the image = [u,v] T The Hessian matrix of is,

[0036]

[0037] Among them, r uu , r uv , r vv is the second-order partial derivative after convolution of I(u,v) with Gaussian kernel, obtained by convolution of I(u,v) with the second-order derivative convolution kernel of Gaussian kernel. The optimal value of Gaussian convolution kernel is determined according to the size of light spot image to ensure the center positioning accuracy of light spot image. The present invention uses coordinate normalization to obtain the normalization operator:

[0038]

[0039] in, r is the spot radius, in pixels. According to the normalized operator C spatial distribution, it can be concluded that the spot center condition is that C(u,v) is greater than 0 and is a local maximum. The actual image is affected by noise, and the precise spot radius cannot be accurately obtained. Therefore, in order to improve the accuracy of spot center positioning, the maximum value C(u,v) corresponding to different C is selected to determine the optimal spot radius σ opt .

[0040] Step 13: Construct the optimal spatial calibration field, which specifically includes the following steps:

[0041] Step 131: For camera intrinsic calibration, the covariance matrix of the intrinsic parameters directly reflects the uncertainty of each solution parameter, that is, J T The k×k submatrix in the upper left corner of J, where J is the Jacobian matrix composed of the partial derivatives of the back-projection error of the image feature points with respect to the camera intrinsic parameters. The information matrix H = J T J can be decomposed into, Where U is a k×k symmetric matrix, W is k×3m, and V is a diagonal matrix. -1 The upper left submatrix of is;

[0042] ∑=(U-WV -1 W T ) + (3)

[0043] If the global optimization method is used, it is necessary to calculate ∑ and its trace for multiple sets of assumed next imaging positions multiple times, where the superscript + means the matrix pseudo-inverse. Let image m+1 be the next target imaging position. In order to solve the optimal target imaging position, the first image to the mth image has been obtained and the intrinsic and extrinsic parameters have been calculated. Then:

[0044]

[0045] Among them, A is the optimal target field criterion matrix, i is the image number, Vi Matrix V computed for image i.

[0046] At this time, for the image m+1 at the next candidate target position, we only need to calculate:

[0047]

[0048] The optimal imaging position of the next target can be determined by continuous incremental method, and the optimal spatial calibration data field can be constructed by analogy. Figure 3 As shown, the images in (a), (b), (c), (e), (f), (h), and (i) are the constructed optimal calibration field data, and the images in (d), (g), and (j) are the eliminated non-optimal calibration field data.

[0049] Step 14: Determine the uncertainty of the positioning of the image feature points, which specifically includes the following steps:

[0050] Let p = (u0, v0) be the pixel-level coordinates of the target feature point, and p1 = (u1, v1) be the sub-pixel-level coordinates. Take the ideal positioning point of the target feature point in the noise-free image as the coordinate origin (0, 0), and u and v as the coordinate axes to establish the O-st coordinate system. In this coordinate system, the sub-pixel-level coordinates of the target feature point p1 are (s0, t0). Therefore, the gray value of the point after convolution can be expressed by the second-order Taylor expansion at (0, 0), that is:

[0051]

[0052] in, I(s,t) in the new coordinate system with a standard deviation of σ g The gray value at (0,0) after convolution of the Gaussian template, are the first-order partial derivatives at (0,0), are the second-order partial derivatives at (0,0). As mentioned above, they can be decomposed into where f ss , f st , f tt They are the second-order partial derivatives of the ideal image after convolution, n ss , n st , n tt are the second-order partial derivatives after image noise convolution. According to the properties of (s0, t0), the first-order derivative of formula (3) at the center of the feature point relative to s and t is zero, that is:

[0053]

[0054] This gives the exact location of the feature point:

[0055]

[0056] Since the grayscale distribution of the light spot in the ideal image satisfies the Gaussian distribution, it can be expressed as where σ w is the standard deviation, is the maximum grayscale value of the image, and M is the grayscale coefficient. s (0,0)=0,f s (0,0) is the first-order partial derivative of s at (0,0) after Gaussian convolution of the ideal image, and n ss (0,0)<<f ss (0,0). According to the Gaussian distribution characteristics, in the convolution function, at (s=0, t=0), u and v satisfy the exchangeability, f ss (0,0)=f tt (0,0), f st (0,0) = 0. Therefore, the positioning variance of the center point coordinates can be expressed as:

[0057]

[0058] From the above content, we can see that K is the maximum grayscale value of the image. is the image noise variance;

[0059] The Gaussian optimal standard deviation σ g Next but:

[0060]

[0061] So we have σ g =σ w Substituting into equation (6), the center point coordinate positioning variance can be expressed as:

[0062]

[0063] It can be seen that when choosing the best scale, that is, the Gaussian best standard deviation σ g When performing convolution calculations, it is possible to avoid solving the inaccurate σ of the actual image. w value, ensuring that the actual uncertainty region is only related to the signal and noise levels. is the variance of the noise n(s,t), K is The maximum grayscale value. That is, the value calculated by formula (11) Defined as the target feature point image positioning uncertainty U d .

[0064] Step 15: Weighted nonlinear optimization of image feature point positioning uncertainty. The Jacobian matrix J in the nonlinear optimization objective function contains the partial derivatives of the objective function residual with respect to the optimization parameters. It is a block matrix. The first column is the partial derivative with respect to the camera intrinsic parameters, and the remaining part is the partial derivative with respect to the extrinsic parameters. J is a sparse matrix.

[0065] Let the number of parameters in the camera be k, then J can be expressed as,

[0066]

[0067] Among them, A i is 2n×k, containing the partial derivatives of the residual with respect to the k intrinsic camera parameters, B i is 2n×3, containing the partial derivatives of the residual with respect to the external parameters, are the back-projection errors of the image feature points respectively; then the information matrix can be expressed as J T J:

[0068]

[0069] Similar to J, J T J is highly sparse, and its inverse matrix (J T J) -1 Provides estimates of the covariance matrix of the estimated intrinsic and extrinsic parameters.

[0070] In order to avoid falling into the local minimum during the optimization solution process, the present invention extracts the calculated image feature point positioning uncertainty σ ij Embedded in J T J to form a new matrix H,

[0071] H=J T diag(σ 11 ,σ 12 ,…,σ 1n ,σ 21 ,…,σ mn )J (14)

[0072] Among them, diag() means taking the diagonal elements of the matrix, and the superscript T means the transpose of the matrix.

[0073] The information matrix based on the weighted uncertainty of image feature point positioning can be used for optimization solution using Gauss-Newton or Levenberg-Marquardt optimization algorithm and quantitative analysis of internal parameter uncertainty.

[0074] Step 16: Center normalization and multi-parameter decoupling. The calibration process of telephoto cameras involves solving and optimizing the internal and external parameters of the camera at the same time, which can easily lead to problems such as large differences in the partial derivatives of the variables and strong correlations between the parameters, resulting in serious coupling in the multi-parameter solution process and reduced solution accuracy. By centering the coefficient matrix, the correlation strength between the elements can be weakened, which is important for accurately solving the internal parameters.

[0075] Suppose there are p observations X1, X2, …, X p The properties and units of measurement of the Jacobian matrix (i.e., the partial derivatives of the corresponding parameters) can be different, so different independent variables need to be standardized after being centered. Let,

[0076]

[0077] Import,

[0078]

[0079] Among them, R j is the mean square error of the jth observation, X ij represents the i-th measurement value of the j-th observation, X ij (i=1,...,n,j=1,...,p) means a total of n measurements, p observations, β j represents the jth coefficient to be estimated, is the average of p observations, is the average value of j observations, where the matrix X * The sum of the squares of the elements in each column is 1, called X * After center standardization, X *' X * It becomes a symmetrical matrix with 1 on the main diagonal. In this way, the normal equation has strong constraints, the solution is relatively stable, and the interference of accidental observation errors is relatively small. Another advantage of central standardization is that the square matrix X *' X * (i,j) is the independent variable X i and X j To evaluate the calculation accuracy of the solution, let the mean square error matrix MSE be:

[0080]

[0081] in, where f x 、f yare the effective focal lengths of the camera intrinsic parameters to be calibrated, u0 and v0 are the principal point coordinate components of the camera intrinsic parameters to be calibrated, k1, k2, k3 are the radial distortion coefficients of the camera intrinsic parameters to be calibrated, p1 and p2 are the tangential distortion coefficients of the camera intrinsic parameters to be calibrated, σ 2 Solve the normal equation The variance of the normal equation, t is the number of eigenvalues, λ i is the eigenvalue of the normal equation. Obviously, if λ i The bigger, The smaller it is, the higher the accuracy of the solution.

[0082] Step 17: Solve the intrinsic parameters of the optical camera. The beam splitter is used to divide the target rays at infinity into two groups. The first group directly enters the camera to be calibrated, and the second group enters the field of view of the camera to be calibrated synchronously after being reflected by the reflector. The swinging plane reflector can change the incident direction of the second group of light rays, which can cover the imaging in the field of view of the entire large-aperture optical camera. The image analysis and calibration module is the system software structure, which analyzes the collected target image data, extracts image features and combines the camera and target parameters. The minimum back-projection error of the target feature points is established as the objective function, and the camera intrinsic parameters to be calibrated are obtained through nonlinear optimization.

[0083] like Figure 3 The figure shows a schematic diagram of the optimal calibration field distribution of the spatial target; among them, Figures (a), (b), (c), (e), (f), (h), and (i) are images within the optimal calibration field, and Figures (d), (g), and (j) are images within the non-optimal calibration field.

Claims

1. A fast and high-precision ground calibration method for a long-focus large-aperture optical camera, characterized in that: The steps include: Step a: In a ground laboratory environment, a high-precision ground calibration hardware device for a telephoto large-aperture optical camera is constructed, and only a plane target is used in combination with a collimator, supplemented by a beam splitter and a plane reflector to achieve high-precision ground calibration of the telephoto large-aperture optical camera; The infinite imaging target is designed as a miniature plane target installed on the focal plane of the collimator to generate an infinite plane target, which is divided into two groups of infinite target rays by a beam splitter. The first group directly enters the camera to be calibrated, and the second group enters the field of view of the camera to be calibrated synchronously after being reflected by the reflector. Step b: Modeling and extracting image feature points, using a high-precision image feature point extraction method based on an optimal scale to achieve high-precision extraction of feature points under different scale conditions; Step c, by solving the uncertainty modeling based on the analysis of the corresponding internal parameters in the Jacobi matrix, the influence of the target space calibration field on the internal parameter calibration accuracy is obtained, and the optimal space calibration field is selected and constructed; Step d, by establishing a mathematical model for local positioning of feature points, a mathematical solution between the uncertainty of image feature point positioning and image noise is obtained; Step e, using image feature point positioning uncertainty weighting and center normalization method to avoid falling into local minimum and reduce coupling between multiple parameters; weighting the image positioning uncertainty of the Jacobi matrix in the camera intrinsic parameter calibration optimization to solve the problems of gradient vanishing, slow convergence speed and easy falling into local minimum; introducing the center normalization method to reduce the coupling of internal and external parameters in the calibration process, so as to achieve fast and high-precision calibration of telephoto camera intrinsic parameters; Step f, swing the plane reflector to change the incident direction of the second group of light rays, cover the entire large-aperture optical camera field of view for imaging, and substitute the above steps in sequence to finally solve and obtain the internal parameters of the optical camera.

2. The method for fast and high-precision ground calibration of a long-focus large-aperture optical camera according to claim 1, characterized in that: In the step b, an adaptive multi-scale centroid extraction method is used to extract the centroid of feature points, so that the centroid extraction of different image feature points is completed under the optimal scale condition, thereby improving the accuracy of feature point centroid extraction.

3. The method for fast and high-precision ground calibration of a long-focus large-aperture optical camera according to claim 2, characterized in that: In the step d, the positioning deviation of each feature point is quantitatively analyzed, and the relationship between the image feature point positioning accuracy and the image noise is solved by analyzing the local feature point positioning of the image and the image noise model, and the mathematical solution of the image feature point positioning uncertainty is obtained, laying the foundation for the subsequent image positioning uncertainty weighting operation.

4. A device for implementing the method for fast and high-precision ground calibration of a telephoto and large-aperture optical camera according to any one of claims 1 to 3, characterized in that: include: Image analysis and calculation module, target imaging and image acquisition module and infinite target module; The image analysis and calculation module includes a computer, which is used to process the images of the target imaging and image acquisition modules, and to optimize and solve the camera internal parameters according to the camera internal parameter model; The target imaging and image acquisition module includes a high-resolution camera and a control box, wherein the high-resolution camera includes a CCD imaging large target surface and a telephoto large-aperture optical lens, and the control box triggers the high-resolution camera to image the infinitely distant target; The infinity target module includes a plane target, a collimator, a beam splitter prism and a plane reflector. The plane target is installed at the focal plane of the collimator, and after being illuminated by a backlight source, it passes through the collimator and enters the beam splitter prism. One beam of light directly enters a telephoto large-aperture optical lens, and the other beam of light passes through the plane reflector and enters the telephoto large-aperture optical lens.

Images