Point target wide-area high-precision positioning method, device and equipment combining imaging model fine-coarse calibration

Abstract

The application provides a point target wide-area high-precision positioning method, device and equipment combined with imaging model fine and coarse calibration, relates to the technical field of optical imaging precision measurement, and aims to realize point target wide-area high-precision positioning. The method comprises the following steps: collecting point target images located at different sub-pixel positions, calibrating the PSF of a first pixel on an image detector to obtain a reference PSF model; mapping the reference PSF model to a second pixel and fitting the imaging data of the second pixel to obtain an extended PSF model; wherein the extended PSF model represents the pixel response values corresponding to different positions of the center of the image spot from the center of the pixel; and positioning the wide-area point target images collected by the image detector by using the extended PSF model and the maximum likelihood method to obtain a positioning result.

Description

Technical Field

[0001] This application relates to the field of optical imaging precision measurement technology, and in particular to a method, apparatus and equipment for wide-area high-precision positioning of point targets by combining fine and coarse calibration of imaging models. Background Technology

[0002] In the field of optical imaging measurement, the imaging of any object can be viewed as a superposition of point target images, that is, a superposition of the optical system's PSF (Point Spread Function). Therefore, point targets are the most basic and typical optical targets. The calculation of the imaging position of a point target on an image detector is called point target localization technology, which directly determines the final measurement accuracy of optical instruments in many applications. In scenarios such as aerospace navigation and astronomical observation, for example, star sensors use star imaging localization and star catalog matching to complete spacecraft attitude calculation, and their attitude measurement accuracy is positively correlated with the star localization accuracy.

[0003] The core of high-precision point target localization is establishing an accurate imaging PSF model. For example, the Hubble Space Telescope uses multi-frame stellar images to invert experimental PSFs and combines least-squares fitting to reduce model errors. However, PSF calibration faces two major technical bottlenecks: First, traditional laboratory calibration requires generating sub-pixel displacements through precise motion devices to improve the spatial sampling rate. However, due to the non-uniformity of pixel response in image detectors, PSF models calibrated for a single pixel cannot be adapted to other pixels, resulting in poor model reusability. Second, to achieve wide-area high-precision localization, pixel-by-pixel calibration of the entire imaging area is required, generating massive amounts of model data and causing problems such as complex calibration processes. Therefore, limited by the non-uniformity of pixel response, existing PSF model calibration methods suffer from high calibration costs and small application areas, making it difficult to achieve wide-area high-precision localization of point targets. Summary of the Invention

[0004] In view of the above problems, embodiments of this application provide a method, apparatus and device for wide-area high-precision positioning of point targets that combines fine and coarse calibration of imaging models, so as to overcome the above problems or at least partially solve the above problems.

[0005] A first aspect of this application discloses a method for wide-area high-precision positioning of point targets combining fine and coarse calibration of imaging models, the method comprising:

[0006] Point target images located at different sub-pixel positions are acquired, and the point spread function (PSF) of the first pixel on the image detector is calibrated to obtain a baseline PSF model; wherein, the first pixel is any pixel on the image detector;

[0007] The baseline PSF model is mapped to the second pixel and fitted with the imaging data of the second pixel to obtain the extended PSF model; wherein, the second pixel is any pixel other than the first pixel on the image detector, the imaging data is the response value of the second pixel collected when the center of the imaging spot is located at multiple different positions in the image region near the center of the second pixel, and the extended PSF model represents the pixel response value corresponding to different positions of the spot center from the pixel center.

[0008] The wide-area point target image acquired by the image detector is located using the extended PSF model and the maximum likelihood method to obtain the location result.

[0009] Optionally, point target images located at different sub-pixel positions are acquired, and the point spread function (PSF) of the first pixel on the image detector is calibrated to obtain a baseline PSF model, including:

[0010] Multiple frames of point target images are acquired at different sub-pixel positions within the target window centered on the first pixel.

[0011] Based on the multi-frame point target image acquired at each sub-pixel position, the pixel response value corresponding to each sub-pixel position is calculated to obtain the imaging PSF sampling matrix of the first pixel. Each element in the imaging PSF sampling matrix represents the pixel response value corresponding to different positions of the image spot center from the center of the first pixel.

[0012] The imaging PSF sampling matrix is ​​interpolated to obtain a baseline PSF model.

[0013] Optionally, multiple frames of point target images are acquired at different sub-pixel positions within the target window centered on the first pixel, including:

[0014] The image detector is moved by a motion actuator to form a relative micro-motion between the point target and the image detector, and the image spot of the point target is controlled to perform sub-pixel-level micro-displacement within the target window centered on the first pixel.

[0015] Multi-frame point target image sampling is performed at each sub-pixel position to obtain multi-frame point target images acquired at different sub-pixel positions within the first pixel.

[0016] Optionally, the baseline PSF model is mapped to the second pixel and fitted with the imaging data of the second pixel to obtain an extended PSF model, including:

[0017] Move the target image spot within the target window centered on the second pixel, and collect imaging data at multiple target locations to obtain the imaging data of the second pixel;

[0018] The baseline PSF model is mapped to the second pixel, and the pixel response of the second pixel is predicted based on the baseline PSF model to obtain the predicted pixel response value of the second pixel;

[0019] The predicted pixel response value of the second pixel and the imaging data of the second pixel are fitted using the least squares method to obtain the extended PSF model.

[0020] Optionally, the predicted pixel response value of the second pixel includes predicted pixel response values ​​corresponding to multiple target locations; mapping the baseline PSF model to the second pixel and predicting the pixel response value of the second pixel based on the baseline PSF model to obtain the predicted pixel response value of the second pixel includes:

[0021] The reference PSF model is mapped to the second pixel, and the reference positions corresponding to the plurality of target positions are determined from the reference PSF model respectively;

[0022] The pixel response value corresponding to the reference position in the reference PSF model is used as the predicted pixel response value corresponding to the target position in the second pixel.

[0023] Optionally, the wide-area point target image acquired by the image detector is located using the extended PSF model and the maximum likelihood method to obtain the location result, including:

[0024] The imaging region is extracted from the wide-area point target image, and the image background value of the imaging region is removed to obtain the pixel response matrix of the imaging region;

[0025] Based on the extended PSF model, establish the joint probability distribution of the pixel response matrix of the imaging region;

[0026] Based on the maximum likelihood method, the location of the image spot with the highest probability of appearing in the pixel response matrix of the imaging region is calculated according to the joint probability distribution, and is used as the localization result.

[0027] Optionally, based on the extended PSF model, a joint probability distribution of the pixel response matrix of the imaging region is established, including:

[0028] The number of photons detected during the pixel response process is determined to follow a Poisson distribution, and the mean of the Poisson distribution is determined according to the extended PSF model.

[0029] The pixel response probability density function of the third pixel is determined based on the mean of the Poisson distribution, wherein the third pixel is any pixel in the imaging region;

[0030] The joint probability distribution is obtained by multiplying the pixel response probability density functions of all pixels within the imaging region.

[0031] Optionally, based on the maximum likelihood method and according to the joint probability distribution, the location of the image spot that has the highest probability of appearing in the pixel response matrix of the imaging region is calculated as the localization result, including:

[0032] Based on the joint probability distribution, construct the cost function;

[0033] The cost function is minimized using the iterative Newton-Rafaelson method, and the localization result is obtained when the iteration termination condition is met.

[0034] The cost function includes terms related to the location of the image spot. Represented as:

[0035] ,

[0036] Where i and j represent the row and column of the pixel, respectively, and K represents the pixel gain. Represents pixels The relative response intercept relative to the first pixel, This indicates the distance of the image spot from the center of the second pixel. pixel response value, This represents the actual pixel response value of the second pixel that was collected.

[0037] A second aspect of this application discloses a wide-area high-precision point target positioning device combining fine and coarse imaging model calibration, the device comprising:

[0038] The calibration module is used to acquire point target images located at different sub-pixel positions, calibrate the point spread function (PSF) of the first pixel on the image detector, and obtain a baseline PSF model; wherein, the first pixel is any pixel on the image detector;

[0039] The fitting module is used to map the baseline PSF model to the second pixel and fit it with the imaging data of the second pixel to obtain an extended PSF model; wherein, the second pixel is any pixel other than the first pixel on the image detector, the imaging data is the response value of the second pixel collected when the center of the imaging spot is located at multiple different positions in the image region near the center of the second pixel, and the extended PSF model represents the pixel response value corresponding to different positions of the spot center from the pixel center;

[0040] The localization module is used to locate the wide-area point target image acquired by the image detector using the extended PSF model and the maximum likelihood method, and obtain the localization result.

[0041] A third aspect of this application discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model described in the first aspect of this application.

[0042] A fourth aspect of this application discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model as described in the first aspect of this application.

[0043] A fifth aspect of this application discloses a computer program product, including a computer program that, when executed by a processor, implements the steps of the point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model as described in the first aspect of this application.

[0044] The embodiments of this application have the following advantages:

[0045] In this embodiment, point target images located at different sub-pixel positions are acquired. The point spread function (PSF) of the first pixel on the image detector is calibrated to obtain a finely calibrated baseline PSF model. Based on this finely calibrated baseline PSF model, the baseline PSF model is mapped to the second pixel and fitted with the imaging data of the second pixel to establish extended PSF models for other pixels. This extended PSF model characterizes the pixel response values ​​corresponding to different positions of the image spot center at different distances from the pixel center. Therefore, this method uses a coarsely calibrated extended PSF model as an approximation, reducing calibration costs while ensuring the accuracy of the PSF model and expanding the applicable area of ​​the accurate PSF model, thus making it suitable for practical engineering applications. Furthermore, by using the extended PSF model and the maximum likelihood method, the wide-area point target images acquired by the image detector are located, making full use of the distribution characteristics of random noise to achieve point target location measurement close to the theoretical accuracy limit, thereby improving the accuracy of point target location. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart illustrating the steps of a wide-area high-precision point target positioning method combining fine and coarse calibration of an imaging model, as provided in an embodiment of this application.

[0048] Figure 2 This is a schematic diagram of a pixel non-uniform coarse calibration result provided in an embodiment of this application;

[0049] Figure 3 This is a comparison chart of PSF coarse calibration results provided in an embodiment of this application;

[0050] Figure 4 This is a flowchart of another method for wide-area high-precision positioning of point targets that combines fine and coarse calibration of the imaging model, provided in an embodiment of this application.

[0051] Figure 5 This is a schematic diagram of a wide-area high-precision positioning result for a point target provided in an embodiment of this application;

[0052] Figure 6 This is a schematic diagram of the structure of a point target wide-area high-precision positioning device that combines fine and coarse calibration of an imaging model, provided in an embodiment of this application.

[0053] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0054] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0055] Point target localization methods in related technologies can be divided into two categories. The first category is the centroid method, which calculates the first-order gray-scale moment of pixel regions in the star imaging area that meet a certain energy response threshold. This method is simple and computationally fast, but because it uses the centroid of the pixel to replace the centroid of the light intensity distribution within the pixel, it suffers from an S-shaped systematic error and has low accuracy. The second category is the fitting method, which fits different PSF functions to the measured pixel data to obtain the center position of the image spot. The most commonly used fitting method is the Gaussian fitting method, which assumes that the pixel data of star imaging can be approximated as a Gaussian function or an integral function of the Gaussian function. This method has a large computational load and has good accuracy in ideal cases where the actual PSF is not much different from the Gaussian function. However, for real non-ideal PSFs, the accuracy often drops significantly.

[0056] Therefore, the core of high-precision point target positioning is to establish an accurate imaging PSF model. However, due to the limitations of pixel response inhomogeneity, existing PSF model calibration methods suffer from high calibration costs and small application areas, making it difficult to achieve wide-area high-precision point target positioning. To overcome the lack of wide-area high-precision point target positioning methods in related technologies, this application provides a method for wide-area high-precision point target positioning that combines fine and coarse imaging model calibration. This method reduces calibration costs while ensuring PSF accuracy and expands the applicable area of ​​the accurate PSF model, thus making it suitable for practical engineering applications.

[0057] Reference Figure 1 As shown, Figure 1 This is a flowchart illustrating the steps of a wide-area high-precision point target localization method combining fine and coarse imaging model calibration, as provided in an embodiment of this application. Figure 1 As shown, the wide-area high-precision point target localization method combining fine and coarse calibration of the imaging model may include steps S110 to S130:

[0058] Step S110: Acquire point target images located at different sub-pixel positions, calibrate the point spread function (PSF) of the first pixel on the image detector, and obtain a baseline PSF model; wherein, the first pixel is any pixel on the image detector.

[0059] Here, a point target image refers to an image acquired for a point target (optical point target). The point target image contains image patches corresponding to the point targets, and the center of the image patch is the location of the point target. Specifically, any pixel on the image detector is designated as the first pixel, and this first pixel is divided into multiple ( The subpixel position grid can control the movement of the image spot at different subpixel positions within the first pixel, thereby acquiring point target images located at different subpixel positions.

[0060] Based on the point target image located at different sub-pixel positions, the pixel response corresponding to each sub-pixel position is calculated. Then, based on the pixel response corresponding to each sub-pixel position within the first pixel, a baseline PSF sampling matrix is ​​obtained. This sampling matrix is ​​then interpolated to obtain a baseline PSF model. This baseline PSF model is a pixel response value (PSF value) with continuous distance, meaning that there is a corresponding pixel response value at any position from the center of the image spot to the center of the first pixel (the baseline PSF model can characterize the precise correspondence between different pixel phases and pixel responses). Therefore, the baseline PSF model is a precisely calibrated baseline PSF model.

[0061] Step S120: Map the baseline PSF model to the second pixel and fit it with the imaging data of the second pixel to obtain an extended PSF model; wherein, the second pixel is any pixel other than the first pixel on the image detector, the imaging data is the response value of the second pixel collected when the center of the imaging spot is located at multiple different positions in the image region near the center of the second pixel, and the extended PSF model represents the pixel response value corresponding to different positions of the spot center from the pixel center.

[0062] The extended PSF model is a coarsely calibrated PSF model for pixel non-uniformity. Since the extended PSF model is obtained by fitting the finely calibrated baseline PSF model and the imaging data of the second pixel, the extended PSF model can be approximated as an accurate PSF model. The pixel response of any pixel other than the first pixel (the second pixel) on the image detector is calibrated using the method in step S120.

[0063] Specifically, a target window centered on the second pixel (e.g., centered on the second pixel) can be used. Imaging data is acquired at multiple target locations within a certain range. Specifically, a point target image is acquired at each target location, and the pixel response value at that target location is calculated based on the point target image. Then, the baseline PSF model is mapped to the second pixel to obtain the predicted pixel response. Based on the linear model of the pixel response, the imaging data of the predicted pixel response and the second pixel are fitted to obtain the extended PSF model.

[0064] Step S130: Using the extended PSF model and the maximum likelihood method, the wide-area point target image acquired by the image detector is located to obtain the location result.

[0065] Image detectors acquire wide-area point target images containing corresponding image patches, with the center of the image patch indicating the point target's location. For wide-area point target images, the imaging region (i.e., the region where the image patch is located) can be extracted, and the imaging region can be located using an extended PSF model and the maximum likelihood method to calculate the image patch's position and obtain the localization result.

[0066] Through the above implementation process, point target images located at different sub-pixel positions are acquired. The point spread function (PSF) of the first pixel on the image detector is calibrated to obtain a finely calibrated baseline PSF model. Based on this finely calibrated baseline PSF model, it is mapped to the second pixel and fitted with the imaging data of the second pixel to establish extended PSF models for other pixels. This extended PSF model characterizes the pixel response values ​​corresponding to different positions of the image spot center at different distances from the pixel center. Therefore, this method uses a coarsely calibrated extended PSF model as an approximation, reducing calibration costs while ensuring the accuracy of the PSF model and expanding the applicable area of ​​the accurate PSF model, thus making it suitable for practical engineering applications. Furthermore, by using the extended PSF model and the maximum likelihood method, the wide-area point target images acquired by the image detector are located. By fully utilizing the distribution characteristics of random noise, the point target location measurement is achieved close to the theoretical accuracy limit, improving the accuracy of point target location.

[0067] In conjunction with the above embodiments, in one embodiment, this application also provides a method for wide-area high-precision positioning of point targets combining fine and coarse calibration of imaging models. In this method, the step S110 above, "acquiring point target images located at different sub-pixel positions, calibrating the point spread function (PSF) of the first pixel on the image detector, and obtaining a reference PSF model," specifically includes sub-steps S110-1 to S110-3:

[0068] Step S110-1: Acquire multi-frame point target images at different sub-pixel positions within the target window centered on the first pixel.

[0069] In this embodiment of the application, the target window is centered on the first pixel. Divided into multiple ( The sub-pixel position grid controls the movement of the image spot at different sub-pixel positions, allowing the image spot position to traverse ( The system identifies 30 sub-pixel locations and generates a multi-frame (e.g., 30 frames) point target image at each sub-pixel location. The sub-pixel location grid, divided by a window centered on the first pixel, is determined based on the energy concentration of the PSF. For example, the first pixel can be divided into... Subpixel location network.

[0070] In some embodiments, acquiring multiple frames of point target images at different sub-pixel positions within a target window centered on the first pixel includes: moving the image detector using a motion actuator to form a relative micro-motion between the point target and the image detector, controlling the image spot of the point target to perform sub-pixel-level micro-displacement within the target window centered on the first pixel; sampling multiple frames of point target images at each sub-pixel position to obtain multiple frames of point target images acquired at different sub-pixel positions within the first pixel.

[0071] The target window is a region consisting of multiple pixels centered on the first pixel. For example, the target window... It can be 1 pixel or The area consists of pixels. The motion actuator is a high-precision turntable or a nanometer pressure radio, etc. Specifically, a high-precision turntable can be used to rotate the optical measurement sensor, or a nanometer pressure radio can be used to move the image detector, forming a relative micro-motion between the point target and the image detector, so that the image spot moves in a region consisting of a first pixel ( Centered on) Perform subpixel-level steps within the target window with a step size of The micro-displacement causes the image spot position to traverse A grid of sub-pixel positions is used to acquire multiple frames of point target images at different sub-pixel positions within the target window centered on the first pixel.

[0072] Step S110-2: Based on the multi-frame point target image acquired at each sub-pixel position, calculate the pixel response value corresponding to each sub-pixel position to obtain the imaging PSF sampling matrix of the first pixel. Each element in the imaging PSF sampling matrix represents the pixel response value corresponding to different positions of the image spot center from the center of the first pixel.

[0073] The imaging PSF sampling matrix of the first pixel is a point target located at the interval. The imaging PSF sampling matrix of the sub-pixel position grid (with a sub-pixel step size). The pixel response value of the target image in each frame can be calculated separately, and the mean of the pixel response values ​​can be calculated based on the pixel response values ​​of the target image in each frame as the pixel response value corresponding to the sub-pixel position; based on the pixel response value corresponding to each sub-pixel position, the imaging PSF sampling matrix of the first pixel is obtained.

[0074] For example, the imaging PSF sampling matrix of the first pixel can be represented as:

[0075]

[0076] in, , , Indicates the initial position of the image spot Distance from the center of the first pixel; Indicates the distance from the center of the image spot to the center of the pixel. The pixel response value at that time, when the average frame rate is high, the grayscale signal-to-noise ratio is high, approximately This is a sample value of the imaging PSF.

[0077] Step S110-3: Interpolate the imaging PSF sampling matrix to obtain the baseline PSF model.

[0078] The imaging PSF sampling matrix of the first pixel is a point target located at the interval. To obtain a baseline PSF model with continuous distances for the pixel response values ​​of the sub-pixel position grid, the imaging PSF sampling matrix of the first pixel is interpolated using a cubic spline interpolation technique with non-node boundary conditions. This results in a precisely calibrated baseline PSF model. .

[0079] Through the above implementation process, based on multi-frame point target images at different sub-pixel positions, the benchmark PSF model corresponding to the first pixel is calibrated. Sampling multi-frame point target images reduces the influence of random noise and removes image background values, thereby improving the accuracy of the benchmark PSF model and achieving accurate calibration of the benchmark PSF model.

[0080] In conjunction with the above embodiments, in one embodiment, this application also provides a method for wide-area high-precision positioning of point targets combining fine and coarse calibration of imaging models. In this method, the step S120 above, "mapping the reference PSF model to the second pixel and fitting it with the imaging data of the second pixel to obtain the extended PSF model," specifically includes sub-steps S120-1 to S120-3:

[0081] Step S120-1: Move the target image spot within the target window centered on the second pixel, and collect imaging data at multiple target locations to obtain the imaging data of the second pixel.

[0082] Wherein, the target window centered at the second pixel can be centered at the second pixel. A region consisting of several pixels is defined by moving the target image patch within the target window centered on the second pixel, thereby making the second pixel... The pixel response changes significantly, and imaging data is acquired at multiple (e.g., 2-3) target locations. Specifically, multiple frames (e.g., 30 frames) of point target images are acquired at each target location, and the pixel response value of each frame of the point target image is calculated. The average of the pixel response values ​​of the multiple frames of point target images is then used as the pixel response value acquired at the target location. For example, the imaging data of the second pixel... It can be represented as:

[0083]

[0084] in, This represents the pixel response value acquired at the first target location. This represents the pixel response value acquired at the a-th target location.

[0085] It is understandable that multiple target locations for imaging data acquisition can be determined based on energy values. That is, the energy peak location, the location with higher energy, and the location with lower energy can each be considered a target location. For example, if imaging data is acquired at three target locations, the point target can be positioned at the center of the second pixel, the edge of the second pixel, and any pixel farther from the second pixel, respectively. In this case, the acquired data... For the energy peak sampling point, For sampling points with higher energy, This is a sampling point with very low energy.

[0086] Step S120-2: Map the reference PSF model to the second pixel, and predict the pixel response of the second pixel according to the reference PSF model to obtain the predicted pixel response value of the second pixel.

[0087] The predicted pixel response value of the second pixel includes predicted pixel response values ​​corresponding to multiple target locations. Specifically, mapping the reference PSF model to the second pixel and predicting the pixel response value of the second pixel based on the reference PSF model to obtain the predicted pixel response value of the second pixel includes: mapping the reference PSF model to the second pixel, determining the reference positions corresponding to the multiple target locations from the reference PSF model, and using the pixel response value corresponding to the reference position in the reference PSF model as the predicted pixel response value corresponding to the target position in the second pixel.

[0088] For example, the second pixel Predicted pixel response value It can be represented as:

[0089]

[0090] in, This represents the predicted pixel response value corresponding to the first target location. Indicates the first The predicted pixel response value corresponding to each target location.

[0091] Step S120-3: Fit the predicted pixel response value of the second pixel and the imaging data of the second pixel using the least squares method to obtain the extended PSF model.

[0092] The fitting equation (i.e., the linear model of pixel response) can be expressed as:

[0093]

[0094] By fitting the predicted pixel response value of the second pixel to the imaging data of the second pixel, the solution is obtained. and ;in, Indicates the second pixel Relative to the first pixel ( The relative response slope of ) Indicates the second pixel Relative to the first pixel ( The relative response intercept, and the fitting results are as follows: Figure 2 As shown, the extended PSF model for coarse calibration of pixel non-uniformity can be obtained by fitting the predicted pixel response values ​​and the collected pixel response values ​​at multiple target locations.

[0095] For example, the extended PSF model can be represented as:

[0096]

[0097] Thus, based on the baseline PSF model with fine calibration of the first pixel, coarse calibration of other pixels is achieved using imaging data sampled at multiple target sampling points (e.g., 3), resulting in an extended PSF model for coarse calibration of pixel non-uniformity. Figure 3 As shown, after coarse calibration, the PSF (pixel response) of different pixels is close to the finely calibrated baseline PSF model. Therefore, the extended PSF model can be approximated as an accurate PSF model. This method reduces the calibration cost from sub-pixel sampling points (e.g., The number of calibration steps has been reduced to three, which reduces calibration costs while ensuring the accuracy of the PSF model and expands the applicable area of ​​the accurate PSF model, making it suitable for practical engineering applications.

[0098] In conjunction with the above embodiments, in one embodiment, this application also provides a method for wide-area high-precision localization of point targets combining fine and coarse calibration of imaging models. In this method, the step S130 above, "localizing the wide-area point target image acquired by the image detector using the extended PSF model and the maximum likelihood method to obtain the localization result," specifically includes sub-steps S130-1 to S130-3:

[0099] Step S130-1: Extract the imaging region from the wide-area point target image and remove the image background value of the imaging region to obtain the pixel response matrix of the imaging region.

[0100] Wide-area point target images typically refer to large-scale images containing multiple point targets (such as stars or microscopic fluorescent markers). Since actual point targets only occupy a portion of the wide-area point target image (e.g., a star's image patch in a star sensor), it is necessary to extract the imaging region from the wide-area point target image. Specifically, methods such as threshold segmentation, edge detection, or template matching can be used to extract the effective imaging region of the target, thereby eliminating redundant areas with no signal or interference and narrowing the scope of subsequent processing.

[0101] Image background values ​​include non-target signals such as detector dark current, ambient stray light, and electronic noise, which can affect the accuracy of localization. Therefore, it is necessary to remove the image background values ​​of the imaging area. One approach is to select neighboring pixels outside the imaging area to calculate the average background value, or to extract low-frequency background components through low-pass filtering, and then subtract the estimated background value from the original grayscale values ​​of each pixel in the imaging area to eliminate baseline shift and highlight the true response of the target signal.

[0102] After removing the image background values, the remaining value of each pixel within the imaging region (ROI) represents the detector's response intensity to the point target's light signal, forming a two-dimensional matrix, namely the imaging region pixel response matrix. , Indicates the third pixel of the imaging area The pixel response value. The pixel response matrix of this imaging region eliminates non-target interference and directly reflects the energy distribution characteristics of the point target. Therefore, the target position can be located based on the pixel response matrix of the imaging region.

[0103] Step S130-2: Based on the extended PSF model, establish the joint probability distribution of the pixel response matrix of the imaging region.

[0104] Specifically, the number of photons detected during the pixel response process is determined to follow a Poisson distribution, and the mean of the Poisson distribution is determined according to the extended PSF model; the pixel response probability density function of the third pixel is determined according to the mean of the Poisson distribution, where the third pixel is any pixel in the imaging region; the pixel response probability density functions of all pixels in the imaging region are multiplied together to obtain the joint probability distribution.

[0105] The number of photons detected during pixel response follows a Poisson distribution, i.e. It follows a Poisson distribution, and the mean of the Poisson distribution is... Where K is the pixel gain, i.e., the gain coefficient for converting photon counts into pixel values. Specifically, K can be determined by the photon transfer method or by consulting a handbook. Therefore, the pixel response probability density function It can be represented as:

[0106]

[0107] in, Indicates the location of the image spot.

[0108] The joint probability distribution is obtained by multiplying the pixel response probability density functions of all pixels within the imaging region. It can be represented as:

[0109]

[0110] Step S130-3: Based on the maximum likelihood method, calculate the image spot position that has the highest probability of appearing in the pixel response matrix of the imaging region according to the joint probability distribution, and use it as the localization result.

[0111] In this embodiment, the location of the image patch with the highest probability of appearing in the pixel response matrix of the imaging region is taken as the target localization result. This is done to maximize the probability of the pixel response matrix appearing in the imaging region. The solution is obtained by constructing a cost function, and the solution result is used as the localization result.

[0112] Specifically, based on the maximum likelihood method, the location of the image spot with the highest probability of appearing in the pixel response matrix of the imaging region is calculated according to the joint probability distribution, and used as the localization result. This includes: constructing a cost function according to the joint probability distribution; minimizing the cost function using the iterative Newton-Rafaelson method; and obtaining the localization result when the iteration termination condition is met.

[0113] The cost function is the negative natural logarithm of the joint probability distribution. It can be represented as:

[0114]

[0115] The cost function includes terms related to the location of the image spot. Represented as:

[0116]

[0117] Where i and j represent the row and column of the pixel, respectively, and K represents the pixel gain. Represents pixels The relative response intercept relative to the first pixel, This indicates the distance of the image spot from the center of the second pixel. pixel response value, This represents the actual pixel response value of the second pixel that was collected.

[0118] The initial position of the image spot can be calculated using the centroid method. The process of minimizing the cost function using the iterative Newton-Rafaelson method is as follows:

[0119]

[0120]

[0121] Where H and J are the cost functions, respectively. The Jacobian and Hessian matrices are, specifically:

[0122]

[0123]

[0124] in,

[0125]

[0126] Specifically, the iteration termination condition can be that the change in the iteration is less than the change threshold, or the number of iterations is greater than the number of iterations threshold. That is, when the change in the target position during the iteration is less than the change threshold, or the number of iterations is greater than the number of iterations threshold, the iteration ends and the high-precision target positioning result is output. In some embodiments, the iteration change threshold is set to 0.001 pixels and the number of iterations threshold is set to 10. In this embodiment, the algorithm typically converges after 2 to 4 iterations.

[0127] Through the above implementation process, by extending the PSF model and using the maximum likelihood method, the wide-area point target images acquired by the image detector are located. By making full use of the distribution characteristics of random noise, the point target is located and measured close to the theoretical accuracy limit, thus improving the accuracy of point target positioning.

[0128] The following specific embodiment illustrates the wide-area high-precision point target positioning method combining fine and coarse calibration of the imaging model according to the embodiments of this application. For example... Figure 4 As shown, the wide-area high-precision point target localization method combining fine and coarse calibration of the imaging model may include steps S410 to S490:

[0129] Step S410: Acquire multi-frame point target images at different sub-pixel positions within the target window centered on the first pixel.

[0130] Step S420: Based on the multi-frame point target image acquired at each sub-pixel position, calculate the pixel response value corresponding to each sub-pixel position to obtain the imaging PSF sampling matrix of the first pixel. Each element in the imaging PSF sampling matrix represents the pixel response value corresponding to different positions of the image spot center from the center of the first pixel.

[0131] Step S430: Interpolate the imaging PSF sampling matrix to obtain the baseline PSF model.

[0132] Step S440: Move the point target within the target window centered on the second pixel, and collect imaging data at multiple target locations to obtain the imaging data of the second pixel.

[0133] Step S450: Map the reference PSF model to the second pixel, and predict the pixel response of the second pixel according to the reference PSF model to obtain the predicted pixel response value of the second pixel.

[0134] Step S460: Fit the predicted pixel response value of the second pixel and the imaging data of the second pixel using the least squares method to obtain the extended PSF model.

[0135] Step S470: Extract the imaging region from the wide-area point target image and remove the image background value of the imaging region to obtain the pixel response matrix of the imaging region.

[0136] Step S480: Based on the extended PSF model, establish the joint probability distribution of the pixel response matrix of the imaging region.

[0137] Step S490: Based on the maximum likelihood method, calculate the image spot position that has the highest probability of appearing in the pixel response matrix of the imaging region according to the joint probability distribution, and use it as the localization result.

[0138] In this embodiment, a precisely calibrated baseline PSF model is used as the core of the system model. The maximum likelihood method is employed to fully utilize the distribution characteristics of random noise, achieving point target positioning measurements close to the theoretical accuracy limit. This represents a significant improvement in accuracy compared to traditional centroid methods and Gaussian fitting methods. Furthermore, using a coarsely calibrated extended PSF model as an approximation greatly simplifies the PSF calibration process, eliminating the need for additional calibration equipment such as integrating spheres. This also reduces the amount of data in the PSF model, facilitating model writing and making it suitable for practical engineering applications.

[0139] Furthermore, to better illustrate the method implemented in this application, the point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model in this application is compared with the maximum likelihood method of single-pixel fine calibration and the traditional centroid method positioning method. The error calculation results are as follows: Figure 5 As shown, the average centering error of the traditional centroid method is 0.0501 pixels, the average centering error of the maximum likelihood method for single-pixel fine calibration is 0.0083 pixels, and the average centering error of the wide-area high-precision point target positioning method combining fine and coarse calibration of the imaging model in this application is 0.0040 pixels. Therefore, the wide-area high-precision point target positioning method combining fine and coarse calibration of the imaging model provided in this application can significantly improve the positioning accuracy of point targets compared with existing positioning methods.

[0140] Based on the same technical concept, this application also provides a point target wide-area high-precision positioning device that combines fine and coarse calibration of imaging models, referring to... Figure 6 As shown, Figure 6 This is a schematic diagram of a point target wide-area high-precision positioning device combining fine and coarse calibration of an imaging model, provided in an embodiment of this application. The device includes:

[0141] The calibration module 610 is used to acquire point target images located at different sub-pixel positions, calibrate the point spread function (PSF) of the first pixel on the image detector, and obtain a baseline PSF model; wherein, the first pixel is any pixel on the image detector;

[0142] The fitting module 620 is used to map the baseline PSF model to the second pixel and fit it with the imaging data of the second pixel to obtain an extended PSF model; wherein, the second pixel is any pixel other than the first pixel on the image detector, the imaging data is the response value of the second pixel collected when the center of the imaging spot is located at multiple different positions in the image region near the center of the second pixel, and the extended PSF model represents the pixel response value corresponding to different positions of the spot center from the pixel center;

[0143] The positioning module 630 is used to locate the wide-area point target image acquired by the image detector using the extended PSF model and the maximum likelihood method, and obtain the positioning result.

[0144] In one optional embodiment, the calibration module includes:

[0145] The image acquisition module is used to acquire multi-frame point target images at different sub-pixel positions within the target window centered on the first pixel.

[0146] The response calculation module is used to calculate the pixel response value corresponding to each sub-pixel position based on the multi-frame point target image acquired at each sub-pixel position, and obtain the imaging PSF sampling matrix of the first pixel. Each element in the imaging PSF sampling matrix represents the pixel response value corresponding to different positions of the image spot center from the center of the first pixel.

[0147] The interpolation processing module is used to perform interpolation processing on the imaging PSF sampling matrix to obtain a reference PSF model.

[0148] In an optional embodiment, the image acquisition module is specifically used to move the image detector using a motion actuator to form a relative micro-motion between the point target and the image detector, control the image spot of the point target to perform sub-pixel-level micro-displacement within the target window centered on the first pixel; and perform multi-frame point target image sampling at each sub-pixel position to obtain multi-frame point target images acquired at different sub-pixel positions within the first pixel.

[0149] In one optional embodiment, the fitting module includes:

[0150] The data acquisition module is used to move the target image patch within the target window centered on the second pixel and acquire imaging data at multiple target locations to obtain the imaging data of the second pixel;

[0151] The response prediction module is used to map the reference PSF model to the second pixel and predict the pixel response of the second pixel according to the reference PSF model to obtain the predicted pixel response value of the second pixel.

[0152] The data fitting module is used to fit the predicted pixel response value of the second pixel and the imaging data of the second pixel using the least squares method to obtain the extended PSF model.

[0153] In one optional embodiment, the predicted pixel response value of the second pixel includes predicted pixel response values ​​corresponding to multiple target locations; the response prediction module is specifically used to map the reference PSF model to the second pixel, determine the reference positions corresponding to the multiple target locations from the reference PSF model respectively; and use the pixel response value corresponding to the reference position in the reference PSF model as the predicted pixel response value corresponding to the target position in the second pixel.

[0154] In one optional embodiment, the positioning module includes:

[0155] The region extraction module is used to extract the imaging region from the wide-area point target image and remove the image background value of the imaging region to obtain the pixel response matrix of the imaging region.

[0156] The distribution establishment module is used to establish the joint probability distribution of the pixel response matrix of the imaging region based on the extended PSF model;

[0157] The location calculation module is used to calculate the location of the image spot with the highest probability of occurrence in the pixel response matrix of the imaging region based on the maximum likelihood method and the joint probability distribution, as the localization result.

[0158] In an optional embodiment, the distribution establishment module is specifically used to determine the number of photons detected during the pixel response process as following a Poisson distribution, and to determine the mean of the Poisson distribution according to the extended PSF model; to determine the pixel response probability density function of the third pixel according to the mean of the Poisson distribution, wherein the third pixel is any pixel in the imaging region; and to multiply the pixel response probability density functions of all pixels in the imaging region to obtain the joint probability distribution.

[0159] This application also provides an electronic device, see embodiments thereof. Figure 7 , Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. For example... Figure 7 As shown, the electronic device 700 includes a memory 710 and a processor 720. The memory 710 and the processor 720 are connected via a bus for communication. The memory 710 stores a computer program that can run on the processor 720 to implement the steps of the point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model described in the embodiments of this application.

[0160] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model described in this application.

[0161] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model described in this application.

[0162] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0163] This application describes embodiments of methods and apparatus according to flowchart illustrations and / or block diagrams. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0164] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0165] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0166] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0167] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0168] The above provides a detailed description of a point target wide-area high-precision positioning method, apparatus, and device combining fine and coarse calibration of imaging models provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for wide-area high-precision positioning of point targets combining fine and coarse calibration of imaging models, characterized in that, The method includes: Point target images located at different sub-pixel positions are acquired, and the point spread function (PSF) of the first pixel on the image detector is calibrated to obtain a baseline PSF model; wherein, the first pixel is any pixel on the image detector; The baseline PSF model is mapped to the second pixel and fitted with the imaging data of the second pixel to obtain the extended PSF model; wherein, the second pixel is any pixel other than the first pixel on the image detector, the imaging data is the response value of the second pixel collected when the center of the imaging spot is located at multiple different positions in the image region near the center of the second pixel, and the extended PSF model represents the pixel response value corresponding to different positions of the spot center from the pixel center. The wide-area point target image acquired by the image detector is located using the extended PSF model and the maximum likelihood method to obtain the location result. This includes: extracting the imaging region from the wide-area point target image and removing the image background value of the imaging region to obtain the pixel response matrix of the imaging region; establishing the joint probability distribution of the pixel response matrix of the imaging region according to the extended PSF model; and calculating the spot position that has the highest probability of appearing in the pixel response matrix of the imaging region based on the maximum likelihood method and the joint probability distribution, as the location result.

2. The point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model according to claim 1, characterized in that, Images of point targets located at different sub-pixel positions are acquired. The point spread function (PSF) of the first pixel on the image detector is calibrated to obtain a baseline PSF model, including: Multiple frames of point target images are acquired at different sub-pixel positions within the target window centered on the first pixel. Based on the multi-frame point target image acquired at each sub-pixel position, the pixel response value corresponding to each sub-pixel position is calculated to obtain the imaging PSF sampling matrix of the first pixel. Each element in the imaging PSF sampling matrix represents the pixel response value corresponding to different positions of the image spot center from the center of the first pixel. The imaging PSF sampling matrix is ​​interpolated to obtain a baseline PSF model.

3. The point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model according to claim 2, characterized in that, Multiple frames of point target images are acquired at different sub-pixel positions within the target window centered on the first pixel, including: The image detector is moved by a motion actuator to form a relative micro-motion between the point target and the image detector, and the image spot of the point target is controlled to perform sub-pixel-level micro-displacement within the target window centered on the first pixel. Multi-frame point target image sampling is performed at each sub-pixel position to obtain multi-frame point target images acquired at different sub-pixel positions within the first pixel.

4. The point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model according to any one of claims 1-3, characterized in that, The baseline PSF model is mapped to the second pixel and fitted with the imaging data of the second pixel to obtain the extended PSF model, including: Move the target image spot within the target window centered on the second pixel, and collect imaging data at multiple target locations to obtain the imaging data of the second pixel; The baseline PSF model is mapped to the second pixel, and the pixel response of the second pixel is predicted based on the baseline PSF model to obtain the predicted pixel response value of the second pixel; The predicted pixel response value of the second pixel and the imaging data of the second pixel are fitted using the least squares method to obtain the extended PSF model.

5. The point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model according to claim 4, characterized in that, The predicted pixel response value of the second pixel includes predicted pixel response values ​​corresponding to multiple target locations; mapping the baseline PSF model to the second pixel, and predicting the pixel response value of the second pixel based on the baseline PSF model to obtain the predicted pixel response value of the second pixel includes: The reference PSF model is mapped to the second pixel, and the reference positions corresponding to the plurality of target positions are determined from the reference PSF model respectively; The pixel response value corresponding to the reference position in the reference PSF model is used as the predicted pixel response value corresponding to the target position in the second pixel.

6. The point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model according to claim 1, characterized in that, Based on the extended PSF model, the joint probability distribution of the pixel response matrix of the imaging region is established, including: The number of photons detected during the pixel response process is determined to follow a Poisson distribution, and the mean of the Poisson distribution is determined according to the extended PSF model. The pixel response probability density function of the third pixel is determined based on the mean of the Poisson distribution, wherein the third pixel is any pixel in the imaging region; The joint probability distribution is obtained by multiplying the pixel response probability density functions of all pixels within the imaging region.

7. The point target wide-area high-precision positioning method combining fine and coarse calibration of the imaging model according to claim 1, characterized in that, Based on the maximum likelihood method, and according to the joint probability distribution, the location of the image spot with the highest probability of appearing in the pixel response matrix of the imaging region is calculated as the localization result, including: Based on the joint probability distribution, construct the cost function; The cost function is minimized using the iterative Newton-Rafaelson method, and the localization result is obtained when the iteration termination condition is met. The cost function includes terms related to the location of the image spot. Represented as: ， Where i and j represent the row and column of the pixel, respectively, and K represents the pixel gain. This indicates the distance of the image spot from the center of the second pixel. pixel response value, This represents the actual pixel response value of the second pixel that was collected.

8. A wide-area high-precision point target positioning device combining fine and coarse calibration of an imaging model, characterized in that, The device includes: The calibration module is used to acquire point target images located at different sub-pixel positions, calibrate the point spread function (PSF) of the first pixel on the image detector, and obtain a baseline PSF model; wherein, the first pixel is any pixel on the image detector; The fitting module is used to map the baseline PSF model to the second pixel and fit it with the imaging data of the second pixel to obtain an extended PSF model; wherein, the second pixel is any pixel other than the first pixel on the image detector, the imaging data is the response value of the second pixel collected when the center of the imaging spot is located at multiple different positions in the image region near the center of the second pixel, and the extended PSF model represents the pixel response value corresponding to different positions of the spot center from the pixel center; The localization module is used to locate the wide-area point target image acquired by the image detector using the extended PSF model and the maximum likelihood method, and to obtain the localization result. This includes: extracting the imaging region from the wide-area point target image and removing the image background value of the imaging region to obtain the pixel response matrix of the imaging region; establishing a joint probability distribution of the pixel response matrix of the imaging region based on the extended PSF model; and calculating the location of the spot that has the highest probability of appearing in the pixel response matrix of the imaging region based on the joint probability distribution using the maximum likelihood method, as the localization result.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the point target wide-area high-precision positioning method combining imaging model fine and coarse calibration as described in any one of claims 1-7.

Images