Morphologically Enhanced Method for Detecting Weak Target Sequences

Abstract

A method for detecting faint target sequences based on morphological enhancement includes: preprocessing the original data to remove periodic texture noise and thermal noise; performing morphological enhancement on the preprocessed result; performing adaptive two-parameter constant false alarm rate (CFAR) calculation on the enhanced result to obtain a binary image; labeling the connected components of the binary image to remove false alarm targets and overexposed high-brightness stars; and performing multi-frame overlay on the obtained binary image, followed by continuous target search on the connected components of the overlay image to confirm moving targets. This disclosure designs a morphological enhancement calculation for faint targets in space. Through basic dilatation and erosion calculations, it achieves noise suppression around near-elliptical targets and enhancement of the target center. Furthermore, it removes false alarms and preserves moving target information through multi-frame target association detection after overlay. The design architecture is simple, computationally inexpensive, and effectively improves real-time computing performance.

Description

Technical Field

[0001] This invention relates to a method for detecting faint targets in space, particularly in the field of morphological enhancement-based methods for detecting faint target sequences. Background Technology

[0002] With the continuous maturation and expansion of space technology, the launch and operation missions of on-orbit spacecraft are becoming increasingly intensive, especially the planning of low-Earth orbit communication satellite constellations, leading to a rapid increase in man-made facilities in outer space. In recent years, the continuous occurrence of satellite collisions and disintegration incidents, and the increasing number of space targets, have posed a significant threat to space safety and high-value spacecraft. Space target surveillance has become crucial for ensuring the normal and reliable operation and maintenance of existing spacecraft. Real-time monitoring is a key early warning method for important spacecraft to avoid collisions, and 24-hour continuous monitoring is an urgent requirement for ensuring the safety of the space station.

[0003] The traditional space target detection process mainly includes three steps: First, based on the star catalog, high-brightness stars in the data are determined to form registration information between the foreground and background data; then, the same target location is searched in three or more frames and possible permutations and combinations are marked; finally, prediction and region search are performed based on multi-frame information to continuously eliminate marked suspected same targets and reduce false alarms.

[0004] For on-orbit targets, their wide-area and continuous characteristics in time and space make real-time detection and tracking difficult. At the same time, small-grained targets have poor light reflection capabilities and appear small and dim at long distances, with a signal-to-noise ratio often below 3, posing a significant challenge to effective target detection. Furthermore, the optical features of background stars and targets are similar, making it difficult to distinguish them solely from image features.

[0005] Traditional methods for detecting faint targets mainly employ linear and nonlinear techniques such as Gaussian filtering, bilateral filtering, and edge detection to differentiate the target body from the background. However, for faint point targets, these methods tend to cause the dispersion of the energy concentration of the point target, thereby reducing the strongest gray level of the target body after background separation and resulting in a decrease in the target signal-to-noise ratio. Summary of the Invention

[0006] This disclosure provides a morphological enhancement-based method for detecting faint target sequences, which can effectively improve the accuracy of detecting faint targets in space.

[0007] The method for detecting faint target sequences based on morphological enhancement disclosed herein includes the following steps:

[0008] S1: Preprocess the raw data to remove periodic texture noise and thermal noise;

[0009] S2: Morphological enhancement of the preprocessing results based on forward and inverse Top-Hat transform;

[0010] S3: Perform adaptive two-parameter constant false alarm rate calculation on the enhancement result to obtain a binary image;

[0011] S4: Perform connected component labeling on the binary image to remove false alarm targets and overexposed high-brightness stars;

[0012] S5: Perform multi-frame overlay on the binary image obtained in step S4, and perform continuous multi-frame target search on the labeled connected component set based on the overlaid image. Confirm whether there is a continuously moving target by confirming the associated trajectory of the multi-frame overlay.

[0013] Furthermore, step S1 specifically includes:

[0014] Gaussian filtering is used to remove periodic texture noise;

[0015] Median filtering is used to remove thermal noise.

[0016] Furthermore, step S2 specifically includes:

[0017] The preprocessed results are subjected to both forward and inverse Top-Hat transformations:

[0018] The positive Top-Hat transform result is added to the preprocessing result, and the inverse Top-Hat transform result is subtracted from the addition result to obtain the morphologically enhanced image.

[0019] Furthermore, step S3 specifically includes:

[0020] Based on the enhanced image data, a target window, a protection window, and a background window are set. The target window mainly includes the grayscale information of the target to be detected; the protection window mainly includes the grayscale information between the target and the background, which is used to protect the diffuse part of the target from being counted in the background window; the background window mainly covers the star noise information; the radius of the target window is determined by the brightness statistics value within the background window.

[0021] A binary image is created based on whether a target might exist within the target window: if so, all points within the target window are set to 1, and all other points are set to 0; the criteria for determining whether a target might exist within the target window are as follows:

[0022]

[0023] in, μ is the mean of the target window. B T represents the mean of the background window. Thr For comparison thresholds.

[0024] Furthermore, the comparison threshold is selected between 1 and 2.

[0025] Furthermore, step S4 specifically includes:

[0026] For binary image Y b Perform connected component labeling to obtain the set of connected components. P represents the number of regions. This represents the set of x-coordinates of the i-th region. This represents the set of y-coordinates of the i-th region;

[0027] The smaller areas in the annotation results are removed as false alarm targets, and the larger areas are removed as overexposed high-brightness star targets, resulting in the binary image Y′. b ,Right now

[0028]

[0029] Where φ represents empty space, S low and S high These represent the lower and upper limits of the area threshold, respectively. i Representing A i The area of ​​the region.

[0030] Furthermore, step S5 specifically includes:

[0031] S51, the binary image Y′ obtained in step S4 b Perform multi-frame overlay and discrimination operations:

[0032]

[0033] W represents the number of adjacent frames superimposed;

[0034] The set of connected components after superposition is

[0035] S52, for any region A′ i,j With its center of mass Using a circle as the center and radius r, where r is inversely proportional to the camera's field of view and directly proportional to the number of pixels, a search is performed on surrounding targets within the radius region. If a target is found in the next frame, it is added to the suspected vector set. The slope of the suspected vector is:

[0036]

[0037] S53, regarding the suspected vector set A i,j,s The region where each suspected vector is located Perform with center of mass Searching with a radius of r centered at r, if a target is found in the next frame, it is added to the suspected vector set. Similarly, the slope of the suspected vector is calculated as follows:

[0038]

[0039] for Any value in SL, if i,j,s If there are equal or similar values, they will continue to be retained as suspected target areas; otherwise, they will be removed as false alarm targets.

[0040] S54, regarding the reservation Continue the search as described above and calculate the slope, similarly. Each vector in and SL i,j,s Check if there are duplicate values; if so, keep the alert, otherwise remove it as a false alarm.

[0041] If there are confirmed targets in three or more consecutive frames, it is considered that there is a continuously moving target; otherwise, it is considered to be a stellar background.

[0042] Compared with the prior art, the beneficial effects of this disclosure are: (1) For weak targets in space, a morphology-based enhancement calculation is designed. Through basic dilatation and erosion calculation, noise suppression around the near-elliptical target and enhancement of the target center are achieved; (2) Based on morphology enhancement, the gray-scale difference between the target and the background can be reconstructed while maintaining the edge contour structure, improving the distinction between the target's internal texture and the background, and effectively improving the efficiency of subsequent detection processing; (3) A multi-frame superposition detection method for short-exposure moving targets is proposed. Noise interference is reduced by adaptive threshold segmentation, star background interference during inertial frame observation is eliminated by multi-frame superposition decision, and false alarms are eliminated and moving target information is retained by the target movement characteristics of the superimposed multi-frames; (4) The proposed design architecture is simple and has a small computational load, which can effectively improve real-time computing performance. Attached Figure Description

[0043] The above and other objects, features and advantages of this disclosure will become more apparent from the more detailed description of exemplary embodiments of this disclosure taken in conjunction with the accompanying drawings, in which the same reference numerals generally represent the same components.

[0044] Figure 1 The flowchart is shown below for the morphological enhancement-based method for detecting faint target sequences according to this disclosure.

[0045] Figure 2 Examples of image filtering and morphological enhancement processing results;

[0046] Figure 3 Example of adaptive two-parameter constant false alarm rate (CFAR) detection results;

[0047] Figure 4Example of comparing the target and noisy background before and after morphological enhancement;

[0048] Figure 5 This is an example of the target sequence trajectory detection result after multi-frame superposition decision. Detailed Implementation

[0049] Preferred embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0050] This disclosure proposes a morphology-based enhancement computation for targets with low spatial density. Through basic dilatation and erosion calculations, it achieves noise suppression around near-elliptical targets and enhancement of the target center. Furthermore, it eliminates false alarms and preserves moving target information by superimposing the target movement characteristics of multiple frames. The design architecture is simple and computationally inefficient, effectively improving real-time computing performance.

[0051] The flowcharts of exemplary embodiments according to this disclosure are attached. Figure 1 As shown, it includes the following steps:

[0052] Step 1: Preprocess the raw data by first performing Gaussian filtering to remove periodic texture noise, and then performing median filtering to remove thermal noise.

[0053] Define I as an image matrix of size M×N, G(u,v) as a 3×3 Gaussian function, and the output result as Y. g The Gaussian filtering process for R=9 levels is as follows:

[0054]

[0055] Gaussian filtering result Y g =Y g,8 ,like Figure 2 As shown in (b).

[0056] For Y g Perform Q=2 level median filtering calculation, and output the result Y. m The calculation process is as follows:

[0057]

[0058] Median filtering result Y m =Y m,1 .

[0059] Step 2: Perform morphological enhancement on the preprocessed filtering result, and simultaneously perform forward and inverse Top-Hat transforms. Add the forward Top-Hat transform result to the preprocessed filtering result, and subtract the inverse Top-Hat transform result from the addition result to obtain the morphologically enhanced image.

[0060] The specific calculations are as follows:

[0061] Morphological enhancement is performed on the preprocessed filtering results, along with forward and inverse Top-Hat transforms.

[0062]

[0063] Where B represents a 3×3 square all-1 structuring element, and OTH and CTH represent the forward and inverse Top-Hat transformations, respectively. Θ represents the expansion operation, and Θ represents the erosion operation.

[0064] Positive Top-Hat Transformation Result OTH B Add (x,y) to the preprocessed filtering result, and subtract the inverse Top-Hat transform result CTH from the addition result. B (x,y) yields the morphologically enhanced image Y. e .

[0065] Y e (x,y)=OTH B (x,y)+Y m (x,y)-CTH B (x,y) (4)

[0066] Figure 2 and Figure 4 These are image contrasts achieved through filtering, morphological enhancement, and contrasts between dark targets and noise.

[0067] Step 3: Perform adaptive two-parameter constant false alarm rate calculation on the enhancement result to obtain a binary image.

[0068] For the enhanced image data Y e Adaptive two-parameter constant false alarm rate (CFAR) detection is performed on (x,y). The detection process includes statistical analysis of information from three windows: target window T. a The radius parameters for the three types of windows—protection window P, background window B, and background window—are as follows: r P and r B The target window primarily includes the grayscale information of the target to be detected; the protection window primarily includes the grayscale information between the target and the background transition, used to protect the diffuse portion of the target from being counted in the background window; the background window primarily covers star noise information. The criterion for determining whether a target is detected within the target window is...

[0069]

[0070] in, μ is the mean of the target window. B T represents the mean of the background window. Thr =1.2 is the comparison threshold. Target window T a radius The size is determined by the brightness statistics within the background window B:

[0071]

[0072] The result of the binary image is:

[0073]

[0074] Figure 3 A binarized image of the adaptive two-parameter constant false alarm rate (CFAR) detection results.

[0075] Step 4: Perform connected component labeling on the binary image. In the labeling results, smaller areas are removed as false alarm targets, and larger areas are removed as overexposed high-brightness star targets.

[0076] The specific calculations are as follows:

[0077] For binary image Y b Connectivity component labeling yields a set of connected components. P represents the number of regions. This represents the set of x-coordinates of the i-th region. Let Y' represent the set of y-coordinates of the i-th region. In the annotation results, smaller areas are removed as false alarm targets, and larger areas are removed as overexposed, high-brightness star targets, resulting in a binary image Y′. b .

[0078]

[0079] Where φ represents empty space, S low =5 and S high =100 represents the lower and upper limits of the area threshold, respectively. i Representing A i The area of ​​the region.

[0080] Step 5: Overlay the binary images obtained in Step 4. After overlay, search for surrounding targets within the radius of the labeled target in the image. Catalog the searched targets, calculate the slope and distance, and search for targets in all cataloged targets according to the slope direction and distance. If there are three or more consecutive targets, it is considered that there is a continuously moving target; otherwise, it is considered to be a stellar background.

[0081] The specific calculations are as follows:

[0082] The binary image Y′ obtained in step 4 b Perform multi-frame overlay and discrimination operations

[0083]

[0084] W represents the number of adjacent frames superimposed.

[0085] The set of connected components after superposition is For any region A′ i，j With its center of mass Using a circle as the center and a radius of r (r is inversely proportional to the camera's field of view and directly proportional to the number of pixels), a search for surrounding targets is performed within the radius region. If a target is found in the next frame, it is incorporated into a potential vector. The slope of the suspected vector is:

[0086]

[0087] For the suspected vector A i,j,s Each region in Perform with center of mass Searching with a radius of r centered at r, if a target is found in the next frame, a potential vector is encoded. Similarly, the slope of the suspected vector is calculated as follows:

[0088]

[0089] for Any value in SL, if i,j,s If there are equal values ​​(with small errors), the target area will continue to be retained; otherwise, it will be removed as a false alarm target.

[0090] For the reserved Continue the search as described above and calculate the slope, similarly. Each vector in and SL i,j,s Check if there are duplicate values; if so, keep the target; otherwise, discard it as a false alarm. If there are three or more consecutive confirmed targets, it is considered that a continuously moving target exists; otherwise, it is considered a stellar background. Figure 5 A multi-frame superimposed binary image (a) and a multi-frame confirmed associated trajectory image (b) are presented.

[0091] The above technical solutions are merely exemplary embodiments of the present invention. For those skilled in the art, based on the application methods and principles disclosed in the present invention, it is easy to make various types of improvements or modifications, and not limited to the methods described in the specific embodiments of the present invention. Therefore, the methods described above are merely preferred and not restrictive.

Claims

1. A method for detecting faint target sequences based on morphological enhancement, comprising the following steps: S1: Preprocess the raw data to remove periodic texture noise and thermal noise; S2: Morphological enhancement of the preprocessing results based on forward and inverse Top-Hat transform; S3: Perform adaptive two-parameter constant false alarm rate calculation on the enhancement result to obtain a binary image; S4: Perform connected component labeling on the binary image to remove false alarm targets and overexposed high-brightness stars; S5: Perform multi-frame overlay on the binary image obtained in step S4, and perform continuous multi-frame target search on the labeled connected component set based on the overlaid image. Confirm whether there is a continuously moving target by confirming the associated trajectory of the multi-frame image. Step S4 specifically includes: For binary images Perform connected component labeling to obtain the set of connected components. , Indicates the number of regions. Indicates the first each region The set of coordinates Indicates the first each region A set of coordinates; In the annotation results, smaller areas are removed as false alarm targets, and larger areas are removed as overexposed, high-brightness star targets, resulting in a binary image. ,Right now in, Indicates empty, and These represent the lower and upper limits of the area threshold, respectively. represent The area of ​​the region; Step S5 specifically includes: S51, the binary image obtained in step S4 Perform multi-frame overlay and discrimination operations: Indicates the number of adjacent frames superimposed; The set of connected components after superposition is ; S52, for any region With its center of mass With the center as the center, For radius, The search for surrounding targets within a radius region is inversely proportional to the camera's field of view and directly proportional to the number of pixels. If a target is found in the next frame, it is added to the suspected vector set. The slope of the suspected vector is: S53, on suspected vector sets The region where each suspected vector is located Perform with center of mass With the center, and For radius search, if a target is found in the next frame, it is included in the potential vector set. Similarly, the slope of the suspected vector is: for Any value in, if in If there are equal or similar values, they will continue to be retained as suspected target areas; otherwise, they will be removed as false alarm targets. S54, regarding the reservation Continue the search as described above and calculate the slope, similarly. Each vector in and Check if there are duplicate values; if so, keep the alert, otherwise remove it as a false alarm. If there are confirmed targets in three or more consecutive frames, it is considered that there is a continuously moving target; otherwise, it is considered to be a stellar background.

2. The method according to claim 1, characterized in that, Step S1 specifically includes: Gaussian filtering is used to remove periodic texture noise; Median filtering is used to remove thermal noise.

3. The method according to claim 1 or 2, characterized in that, Step S2 specifically includes: The preprocessed results are subjected to both forward and inverse Top-Hat transformations: The positive Top-Hat transform result is added to the preprocessing result, and the inverse Top-Hat transform result is subtracted from the addition result to obtain the morphologically enhanced image.

4. The method according to claim 1, characterized in that, Step S3 specifically includes: Based on the enhanced image data, a target window, a protection window, and a background window are set. The target window mainly includes the grayscale information of the target to be detected; the protection window mainly includes the grayscale information between the target and the background, which is used to protect the diffuse part of the target from being counted in the background window; the background window mainly covers the star noise information; the radius of the target window is determined by the brightness statistics value within the background window. A binary image is created based on whether a target might exist within the target window: if so, all points within the target window are set to 1, and all other points are set to 0; the criteria for determining whether a target might exist within the target window are as follows: in, The mean of the target window. This is the average value of the background window. For comparison thresholds.

5. The method according to claim 4, characterized in that, The comparison threshold is selected between 1 and 2.

Images