A weak signal detection method based on a multi-frame, multi-directional spiral vignetting effect correction model

Abstract

This invention aims to provide a weak signal detection method based on a multi-frame, multi-directional spiral vignetting effect correction model, comprising the following steps: A. Accumulating N frames of the original vignetting image to be processed, performing 25 multi-frame superpositions, and numbering the images from 1 to 25; B. Establishing a spiral model composed of 5×5 region blocks, where each region block is an L×L pixel box; C. Using each pixel point in each superimposed image as the center pixel point (0,0) of the spiral model, performing upward, downward, leftward, and right spiral perturbations to obtain the perturbation results for each pixel point in each superimposed image; D. Taking the minimum value obtained in step C, replacing the pixel value of the corresponding pixel point in the first original frame image to obtain the optimal background, and then performing differential processing with the first original frame image to obtain the final result. This invention aims to greatly improve the target detection capability.

Description

Technical Field

[0001] This invention relates to the field of computer image processing and is applicable to the tracking and detection of weak targets against non-uniform backgrounds. Specifically, it relates to a weak signal detection method based on a multi-frame, multi-directional spiral vignetting effect correction model. Background Technology

[0002] Infrared target detection is widely used in fields such as optical remote sensing, missile defense systems, and night navigation systems. In recent years, both infrared detection technology and infrared image processing technology have seen significant advancements. However, a challenge remains in the research of this core technology: at long distances, the target's pixel count is low, and the surrounding environment is noisy, factors that greatly affect the detection of small targets.

[0003] Against a strong clutter background, the distance between an object and the detector gradually decreases with increasing distance, causing the object's brightness to decrease non-linearly from the center to the periphery—a phenomenon known as "vignetting." Vignetting results in uneven distribution of the effective absorbed energy of the incident light by the imaging system, as well as uneven distribution of the output signal, leading to color and brightness distortions in the image and affecting the accuracy of image-based visual detection. Because most of the light energy is concentrated in the center of the image during propagation, the light energy diffuses outwards from the center like a point-diffusion energy distribution, forming a very noticeable central halo. Summary of the Invention

[0004] This invention aims to provide a weak signal detection method based on a multi-frame multi-directional spiral vignetting effect correction model. This method superimposes multiple frames of an image to enhance spatiotemporal information, obtains the neighborhood grayscale gradient based on the multi-frame multi-directional spiral perturbation model, and extracts the corresponding mean data using the optimal gradient model to replace the central pixel, thereby describing the vignetting distribution of the image. This achieves background suppression while effectively correcting the vignetting effect, greatly improving the detection capability of the target.

[0005] The technical solution of the present invention is as follows:

[0006] A. Accumulate N frames of the original image to be processed for the vignetting effect. Superimpose the original images from 1 to N frames to obtain the superimposed image f1, superimpose the images from 2 to (N+1) frames to obtain the superimposed image f2, and so on for 25 times to obtain 25 superimposed images. Number the superimposed images from 1 to 25.

[0007] B. Construct a spiral model consisting of 5×5 region blocks, where each region block is a 3×3 pixel bounding box; assign coordinate numbers to the center pixels of each region block in the spiral model, as shown in the table below:

[0008]

[0009]

[0010] C. Using each pixel point in each superimposed image as the center pixel point (0,0) of the spiral model, put it into the spiral model, and perform upward, downward, leftward and right spiral perturbations on each superimposed image based on the spiral model; obtain the results of each perturbation of each pixel point in each superimposed image;

[0011] D. Take the minimum value obtained in step C and replace the corresponding pixel value of that pixel in the first frame of the original image to obtain the optimal background. Then, perform differential processing on the optimal background and the first frame of the original image to obtain the final detection differential image.

[0012] In step C, the specific process of spiral perturbation for each pixel is as follows:

[0013] a. Calculate the average pixel value of each pixel in the region where the center coordinates (0, 0) of the first superimposed image are located, denoted as H. (x0,y0) ;

[0014] b. Upward spiral disturbance:

[0015] Upward spiral perturbation numbering: Each region block is numbered 1-25 in the following order: (0,0), (-1,0), (-1,-1), (0,-1), (1,-1), (1,0), (1,1), (0,1), (-1,1), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2), (2,-1), (2,0), (2,1), (2,2), (1,2), (0,2), (-1,2), (-2,2).

[0016] The upward spiral perturbation process is as follows: In the first perturbation, the (2-i)th superimposed image is used to extract region blocks based on their corresponding image numbers. The average pixel value of each pixel in this region block is calculated and denoted as H. i (r1,t1); Set each H... i (r1,t1) and The gradients are obtained by subtracting the gradients from the subtracted gradients. The superimposed image corresponding to the minimum gradient is selected, and the coordinates (x1, y1) of the center pixel of the region containing the minimum gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min1 ;

[0017] In the second perturbation, (x1, y1) is used as the center pixel. Each superimposed image is used to extract region blocks based on its image number. The average pixel value of each pixel in that region block is calculated and denoted as H. i (r2, t2); All H... i (r2, t2) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min2 ;

[0018] c. Downward spiral disturbance:

[0019] The numbering method for the downward spiral perturbation is as follows: each region block is numbered 1-25 in the following order: (0,0), (1,0), (1,1), (0,1), (-1,1), (-1,0), (-1,-1), (0,-1), (1,-1), (2,-1), (2,0), (2,1), (2,2), (1,2), (0,2), (-1,2), (-2,2), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2).

[0020] The downward spiral perturbation process is as follows: In the first perturbation, the (2-i)th superimposed image is used to extract region blocks based on their corresponding image numbers. The average pixel value of each pixel in this region block is calculated and denoted as H. i (r3, t3); All H's i (r3, t3) and The gradients are obtained by subtracting the gradients from each gradient. The superimposed image corresponding to the smallest gradient is selected, and the coordinates (x2, y2) of the center pixel of the region containing the smallest gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min3 And obtain the image number of the image;

[0021] In the second perturbation, (x2, y2) is used as the center pixel. Regions are extracted from each superimposed image using the index corresponding to its image number. The average pixel value of each pixel in that region is calculated and denoted as H. i (r4, t4); All H... i (r4,t4) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min4 ;

[0022] d. The numbering method for the left-hand spiral perturbation is as follows: each region block is numbered 1-25 in the following order: (0,0), (1,0), (1,1), (0,1), (-1,1), (-1,0), (-1,-1), (0,-1), (1,-1), (2,-1), (2,0), (2,1), (2,2), (1,2), (0,2), (-1,2), (-2,2), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2).

[0023] The process of left-spiral perturbation: In the first perturbation, the (2-i)th superimposed image is used to extract the region block according to the sequence number corresponding to its image number. The average pixel value of each pixel in the region block is calculated and denoted as H. i (r5, t5); Set each H... i (r5, t5) and The gradients are obtained by subtracting the gradients from the subtracted gradients. The superimposed image corresponding to the minimum gradient is selected, and the coordinates (x3, y3) of the center pixel of the region containing the minimum gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min5 And obtain the image number of the image;

[0024] In the second perturbation, with (x3, y3) as the center pixel, each superimposed image is used to extract region blocks based on its image number. The average pixel value of each pixel in that region block is calculated and denoted as H. i (r6,t6); Set each H... i (r6,t6) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min6 ;

[0025] e. Right-hand spiral perturbation:

[0026] The numbering method for the right-hand spiral perturbation is as follows: each region block is numbered 1-25 in the following order: (0,0), (0,1), (-1,1), (-1,0), (-1,-1), (0,-1), (1,-1), (1,0), (1,1), (1,2), (0,2), (-1,2), (-2,2), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2), (2,-1), (2,0), (2,1), (2,2).

[0027] Right-hand spiral perturbation implementation process: In the first perturbation, the (2-i)th superimposed image is used to extract the region block according to the sequence number corresponding to its image number. The average pixel value of each pixel in the region block is calculated and denoted as H. i (r7, t7); Set each H... i (r7, t7) and The gradients are obtained by subtracting the gradients from the subtracted gradients. The superimposed image corresponding to the minimum gradient is selected, and the coordinates (x4, y4) of the center pixel of the region containing the minimum gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min7 And obtain the image number of the image;

[0028] In the second perturbation, (x4, y4) is used as the center pixel. Regions are extracted from each superimposed image using the index corresponding to its image number. The average pixel value of each pixel in that region is calculated and denoted as H. i (r8, t8); Set each H... i (r8, t8) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min8 ;

[0029] D. Take the H value of each pixel in each superimposed image obtained in step C. min1 H min2 H min3 H min4 H min5 H min6 H min7 H min8 The minimum value in the first frame is used to replace the corresponding pixel value of that pixel in the original image to obtain the optimal background f'(x0,y0). Then, the optimal background f'(x0,y0) is differentially processed with the original image to obtain the final desired difference image.

[0030] Preferably, in step A, N = 50, 100, or 200.

[0031] The advantages of this invention compared to existing technologies are:

[0032] The method of this invention uses a spiral model in four directions to perturb the image. During the perturbation process, neighborhood block data on different frames are acquired to establish spatiotemporal information. A vignetting effect correction model is constructed based on the optimal gradient value to describe the distribution of the image, thereby reducing noise and enabling the method to still have good detection performance in strong clutter backgrounds. Attached Figure Description

[0033] Figure 1Example of overlay images;

[0034] Figure 2 Based on Figure 1 Example of upward spiral region block numbering;

[0035] Figure 3 Based on Figure 1 The upward spiral perturbation model;

[0036] Figure 4 Based on Figure 1 The downward spiral perturbation model;

[0037] Figure 5 Based on Figure 1 A left-spiral perturbation model;

[0038] Figure 6 Based on Figure 1 The right-handed spiral perturbation model;

[0039] Figure 7 Comparison of results from different algorithms for scenario 1;

[0040] Figure 8 Comparison of results from different algorithms in scenario 2;

[0041] Figure 9 Comparison of results from different algorithms for scenario 3. Detailed Implementation

[0042] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0043] Example 1

[0044] The weak signal detection method based on the multi-frame multi-directional spiral vignetting effect correction model in this embodiment includes the following steps:

[0045] A. Accumulate N frames of the original image to be processed for the vignetting effect. Superimpose frames 1-N to obtain superimposed image f1, frames 2-(N+1) to obtain superimposed image f2, and so on, performing this multi-frame superposition 25 times to obtain 25 superimposed images. Number each superimposed image from 1 to 25. See the example of the superimposed images below. Figure 1

[0046] B. Construct a spiral model consisting of 5×5 region blocks, where each region block is a 3×3 pixel bounding box; assign coordinate numbers to the center pixels of each region block in the spiral model, as shown in the table below:

[0047] (-2，-2) (-2，-1) (-2，0) (-2，1) (-2，2) (-1，-2) (-1，-1) (-1，0) (-1，1) (-1，2) (0，-2) (0，-1) (0，0) (0，1) (0，2) (1，-2) (1，-1) (1，0) (1，1) (1，2) (2，-2) (2，-1) (2，0) (2，1) (2，2)

[0048] C. Using each pixel point in each superimposed image as the center pixel point (0,0) of the spiral model, put them into the spiral model, and perform upward, downward, leftward and right spiral perturbations on each superimposed image based on the spiral model.

[0049] The specific process of spiral perturbation for each pixel is as follows:

[0050] a. Calculate the average pixel value of each pixel in the region where the center coordinates (0, 0) of the first superimposed image are located, denoted as .

[0051] b. Upward spiral disturbance:

[0052] Upward spiral perturbation numbering: Each region block is numbered 1-25 sequentially according to the coordinate order (0,0), (-1,0), (-1,-1), (0,-1), (1,-1), (1,0), (1,1), (0,1), (-1,1), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2), (2,-1), (2,0), (2,1), (2,2), (1,2), (0,2), (-1,2), (-2,2); see specific numbering example. Figure 2 The upward spiral disturbance line is shown below. Figure 3 ;

[0053] The upward spiral perturbation process is as follows: In the first perturbation, the (2-i)th superimposed image is used to extract region blocks based on their corresponding image numbers. The average pixel value of each pixel in this region block is calculated and denoted as H. i (r1,t1); Set each H... i (r1,t1) and The gradients are obtained by subtracting the gradients from the subtracted gradients. The superimposed image corresponding to the minimum gradient is selected, and the coordinates (x1, y1) of the center pixel of the region containing the minimum gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min1 ;

[0054] In the second perturbation, (x1, y1) is used as the center pixel. Each superimposed image is used to extract region blocks based on its image number. The average pixel value of each pixel in that region block is calculated and denoted as H. i (r2, t2); All H... i (r2, t2) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min2 ;

[0055] c. Downward spiral disturbance:

[0056] The downward spiral perturbation is numbered as follows: each region block is numbered 1-25 sequentially according to its coordinates: (0,0), (1,0), (1,1), (0,1), (-1,1), (-1,0), (-1,-1), (0,-1), (1,-1), (2,-1), (2,0), (2,1), (2,2), (1,2), (0,2), (-1,2), (-2,2), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2); the downward spiral perturbation line is shown in [reference needed]. Figure 4 ;

[0057] The downward spiral perturbation process is as follows: In the first perturbation, the (2-i)th superimposed image is used to extract region blocks based on their corresponding image numbers. The average pixel value of each pixel in this region block is calculated and denoted as H. i (r3, t3); All H's i (r3, t3) and The gradients are obtained by subtracting the gradients from each gradient. The superimposed image corresponding to the smallest gradient is selected, and the coordinates (x2, y2) of the center pixel of the region containing the smallest gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min3 And obtain the image number of the image;

[0058] In the second perturbation, (x2, y2) is used as the center pixel. Regions are extracted from each superimposed image using the index corresponding to its image number. The average pixel value of each pixel in that region is calculated and denoted as H. i (r4, t4); All H... i (r4,t4) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min4 ;

[0059] d. The numbering method for the left-spiral perturbation is as follows: Each region block is numbered 1-25 sequentially according to the coordinate order (0,0), (1,0), (1,1), (0,1), (-1,1), (-1,0), (-1,-1), (0,-1), (1,-1), (2,-1), (2,0), (2,1), (2,2), (1,2), (0,2), (-1,2), (-2,2), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2); the left-spiral perturbation line is shown in the figure. Figure 5 ;

[0060] The process of left-spiral perturbation: In the first perturbation, the (2-i)th superimposed image is used to extract the region block according to the sequence number corresponding to its image number. The average pixel value of each pixel in the region block is calculated and denoted as H. i (r5, t5); Set each H... i (r5, t5) and The gradients are obtained by subtracting the gradients from the subtracted gradients. The superimposed image corresponding to the minimum gradient is selected, and the coordinates (x3, y3) of the center pixel of the region containing the minimum gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min5 And obtain the image number of the image;

[0061] In the second perturbation, with (x3, y3) as the center pixel, each superimposed image is used to extract region blocks based on its image number. The average pixel value of each pixel in that region block is calculated and denoted as H. i (r6,t6); Set each H... i (r6,t6) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min6 ;

[0062] e. Right-hand spiral perturbation:

[0063] The numbering method for the right-hand spiral perturbation is as follows: each region block is numbered 1-25 sequentially according to the coordinate order (0,0), (0,1), (-1,1), (-1,0), (-1,-1), (0,-1), (1,-1), (1,0), (1,1), (1,2), (0,2), (-1,2), (-2,2), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2), (2,-1), (2,0), (2,1), (2,2); the right-hand spiral perturbation line is shown in [reference needed]. Figure 6 ;

[0064] Right-hand spiral perturbation implementation process: In the first perturbation, the (2-i)th superimposed image is used to extract the region block according to the sequence number corresponding to its image number. The average pixel value of each pixel in the region block is calculated and denoted as H. i (r7, t7); Set each H... i (r7, t7) and The gradients are obtained by subtracting the gradients from the subtracted gradients. The superimposed image corresponding to the minimum gradient is selected, and the coordinates (x4, y4) of the center pixel of the region containing the minimum gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min7 And obtain the image number of the image;

[0065] In the second perturbation, (x4, y4) is used as the center pixel. Regions are extracted from each superimposed image using the index corresponding to its image number. The average pixel value of each pixel in that region is calculated and denoted as H. i (r8, t8); Set each H... i (r8, t8) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min8 ;

[0066] D. Take the H value of each pixel in each superimposed image obtained in step C. min1 H min2 H min3 H min4 H min5 H min6 H min7 H min8 The minimum value in the first frame is used to replace the corresponding pixel value of the original image to obtain the optimal background f'(x0,y0). Then, the optimal background f'(x0,y0) is differentially processed with the original image to obtain the final detection difference image.

[0067] Example 2

[0068] The method of Example 1 is compared with the prior art method, and the prior art method being compared is as follows:

[0069] For details on the multi-directional gradient algorithm, please refer to the existing technical literature: ZLLi, TSSheng, and SLLou, “Correction method of vignetting effect in infrared system based on polynomial approximation,” Infr. Laser Eng., vol. 45, no. S1, 2016, Art. no. S104002.

[0070] For details of the two-point correction algorithm, please refer to the existing technical literature: Xu Xinxing, Yang Hongbo, Gao Yunguo, et al. Design of automatic switching device for non-uniformity correction of infrared camera [D]. Applied Optics. 2013.34(4):658-662;

[0071] For details on the spiral gradient algorithm, please refer to the existing technical literature: F. Xiangsuo, M. Lei*, L. Juliu, G. Hongwei, F. Linping and X. Zhiyong, Dim and Small Target Detection Based on SpiralGradient Optimization Estimation and High-Order Correlation Enhancement, IEEE Access, 2022, 10(1): 14767-14778.

[0072] The comparison results are shown below. Figure 7 , Figure 8 and Figure 9 ,like Figure 7 , Figure 8 and Figure 9 visible, Figure 7 , Figure 8 and Figure 9 In the diagram, Figure (a) represents the signal-to-noise ratio (SNR) of the original image, Figure (b) represents the background evaluation image of the proposed algorithm, Figure (c) represents the SNR of the difference image from the multi-directional gradient algorithm, Figure (d) represents the SNR of the difference image from the two-point correction algorithm, Figure (e) represents the SNR of the difference image from the spiral gradient algorithm, and Figure (f) represents the SNR of the difference image from the proposed algorithm. Compared with other algorithms, the proposed algorithm achieves enhanced SNR in the above three scenarios, resulting in clearer target detection and better detection performance.

Claims

1. A weak signal detection method based on a multi-frame, multi-directional spiral vignetting effect correction model, characterized in that, Includes the following steps: A. Accumulate N frames of the original image to be processed for the vignetting effect. Superimpose the original images from 1 to N frames to obtain the superimposed image f1, superimpose the images from 2 to (N+1) frames to obtain the superimposed image f2, and so on for 25 times to obtain 25 superimposed images. Number the superimposed images from 1 to 25. B. Construct a spiral model consisting of 5×5 region blocks, where each region block is a 3×3 pixel bounding box; assign coordinate numbers to the center pixels of each region block in the spiral model, as shown in the table below: C. Using each pixel point in each superimposed image as the center pixel point (0,0) of the spiral model, put it into the spiral model, and perform upward, downward, leftward, and right spiral perturbations on each superimposed image based on the spiral model; obtain the perturbation results of each pixel point in each superimposed image. D. Take the minimum value obtained in step C and replace the corresponding pixel value of that pixel in the first frame of the original image to obtain the optimal background. Then, perform differential processing on the optimal background and the first frame of the original image to obtain the final detection differential image.

2. The weak signal detection method based on a multi-frame, multi-directional spiral vignetting effect correction model as described in claim 1, characterized in that: In step C, the specific process of spiral perturbation for each pixel is as follows: a. Calculate the average pixel value of each pixel in the region where the center coordinates (0, 0) of the first superimposed image are located, denoted as . b. Upward spiral disturbance: Upward spiral perturbation numbering: Each region block is numbered 1-25 in the following order: (0,0), (-1,0), (-1,-1), (0,-1), (1,-1), (1,0), (1,1), (0,1), (-1,1), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2), (2,-1), (2,0), (2,1), (2,2), (1,2), (0,2), (-1,2), (-2,2). The upward spiral perturbation process is as follows: In the first perturbation, the (2-i)th superimposed image is used to extract region blocks based on their corresponding image numbers. The average pixel value of each pixel in this region block is calculated and denoted as H. i (r1,t1); Set each H... i (r1,t1) and The gradients are obtained by subtracting the gradients from the subtracted gradients. The superimposed image corresponding to the minimum gradient is selected, and the coordinates (x1, y1) of the center pixel of the region containing the minimum gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min1 ; In the second perturbation, (x1, y1) is used as the center pixel. Each superimposed image is used to extract region blocks based on its image number. The average pixel value of each pixel in that region block is calculated and denoted as H. i (r2, t2); All H... i (r2, t2) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min2 ; c. Downward spiral disturbance: The numbering method for the downward spiral perturbation is as follows: each region block is numbered 1-25 in the following order: (0,0), (1,0), (1,1), (0,1), (-1,1), (-1,0), (-1,-1), (0,-1), (1,-1), (2,-1), (2,0), (2,1), (2,2), (1,2), (0,2), (-1,2), (-2,2), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2). The downward spiral perturbation process is as follows: In the first perturbation, the (2-i)th superimposed image is used to extract region blocks based on their corresponding image numbers. The average pixel value of each pixel in this region block is calculated and denoted as H. i (r3, t3); All H's i (r3, t3) and The gradients are obtained by subtracting the gradients from each gradient. The superimposed image corresponding to the smallest gradient is selected, and the coordinates (x2, y2) of the center pixel of the region containing the smallest gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min3 And obtain the image number of the image; In the second perturbation, (x2, y2) is used as the center pixel. Regions are extracted from each superimposed image using the index corresponding to its image number. The average pixel value of each pixel in that region is calculated and denoted as H. i (r4, t4); All H... i (r4,t4) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min4 ; d. The numbering method for the left-hand spiral perturbation is as follows: each region block is numbered 1-25 in the following order: (0,0), (1,0), (1,1), (0,1), (-1,1), (-1,0), (-1,-1), (0,-1), (1,-1), (2,-1), (2,0), (2,1), (2,2), (1,2), (0,2), (-1,2), (-2,2), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2). The process of left-spiral perturbation: In the first perturbation, the (2-i)th superimposed image is used to extract the region block according to the sequence number corresponding to its image number. The average pixel value of each pixel in the region block is calculated and denoted as H. i (r5, t5); Set each H... i (r5, t5) and The gradients are obtained by subtracting the gradients from the subtracted gradients. The superimposed image corresponding to the minimum gradient is selected, and the coordinates (x3, y3) of the center pixel of the region containing the minimum gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min5 And obtain the image number of the image; In the second perturbation, with (x3, y3) as the center pixel, each superimposed image is used to extract region blocks based on its image number. The average pixel value of each pixel in that region block is calculated and denoted as H. i (r6,t6); Set each H... i (r6,t6) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min6 ; e. Right-hand spiral perturbation: The numbering method for the right-hand spiral perturbation is as follows: each region block is numbered 1-25 in the following order: (0,0), (0,1), (-1,1), (-1,0), (-1,-1), (0,-1), (1,-1), (1,0), (1,1), (1,2), (0,2), (-1,2), (-2,2), (-2,1), (-2,0), (-2,-1), (-2,-2), (-1,-2), (0,-2), (1,-2), (2,-2), (2,-1), (2,0), (2,1), (2,2). Right-hand spiral perturbation implementation process: In the first perturbation, the (2-i)th superimposed image is used to extract the region block according to the sequence number corresponding to its image number. The average pixel value of each pixel in the region block is calculated and denoted as H. i (r7, t7); Set each H... i (r7, t7) and The gradients are obtained by subtracting the gradients from the subtracted gradients. The superimposed image corresponding to the minimum gradient is selected, and the coordinates (x4, y4) of the center pixel of the region containing the minimum gradient in the superimposed image are recorded, along with the average value H of all pixels in that region. min7 And obtain the image number of the image; In the second perturbation, (x4, y4) is used as the center pixel. Regions are extracted from each superimposed image using the index corresponding to its image number. The average pixel value of each pixel in that region is calculated and denoted as H. i (r8, t8); Set each H... i (r8, t8) and The subtraction of the gradients is performed to obtain new gradients. The superimposed image corresponding to the minimum gradient is selected, and the average value H of all pixels in the region containing the minimum gradient in the superimposed image is recorded. min8 ; D. Take the H value of each pixel in each superimposed image obtained in step C. min1 H min2 H min3 H min4 H min5 H min6 H min7 H min8 The minimum value in the first frame is used to replace the corresponding pixel value of the original image to obtain the optimal background f'(x0,y0). Then, the optimal background f'(x0,y0) is differentially processed with the original image to obtain the final detection difference image.

3. The weak signal detection method based on a multi-frame, multi-directional spiral vignetting effect correction model as described in claim 1, characterized in that: In step A, N = 50, 100, or 200.

Images