Method for selecting strong points of space-based ISAL combining multi-dimensional features

Abstract

This invention provides a method for selecting strong scattering points in space-based ISAL systems by incorporating multi-dimensional features, relating to the field of ISAL image processing. This invention employs a multi-stage fusion processing strategy to form a complete link from image preprocessing to the final output of strong scattering points: pixel replication and extension are used to process convolutional boundary regions, significantly reducing boundary amplitude distortion caused by traditional zero-padding or periodic extension, and improving the signal-to-noise ratio of edge regions in the filtered amplitude map; the dynamic threshold generation process works synergistically with the preceding Gaussian kernel convolution preprocessing to suppress speckle noise while preserving the true amplitude characteristics of weakly scattering targets; amplitude-weighted averaging is used to calculate the centroid coordinates within the connected domain, overcoming the energy diffusion problem caused by the ISAL system's point spread function; transient interference points are eliminated based on the peak-to-mean ratio of multi-frame amplitude sequences, and the spatial distribution characteristics of the centroid position's spatial amplitude value and density constraints are combined to ensure that the output set of preferred scattering points simultaneously satisfies temporal stability, spatial saliency, and uniform distribution.

Description

Technical Field

[0001] This invention relates to the field of ISAL image processing, and specifically to a method for selecting strong points in space-based ISAL by combining multi-dimensional features. Background Technology

[0002] In inverse synthetic aperture lidar (ISAL) imaging systems, strong scattering points are important features of targets, and their accurate extraction is crucial for tasks such as motion compensation, attitude estimation, and target recognition.

[0003] In related technologies, ISAL scattering point extraction methods mainly include threshold segmentation, edge detection, and cluster analysis. These methods face several major challenges: severe laser speckle noise interference leads to random high-brightness areas in the target region, affecting the identification of true scattering points; the presence of system point diffusion effects causes energy diffusion in scattering points, resulting in insufficient localization accuracy for pixel-level extraction methods; furthermore, the time-varying scattering characteristics caused by target motion make it difficult for traditional single-frame analysis methods to maintain the temporal stability of scattering point information. While Gaussian filters can suppress noise during preprocessing, their boundary handling methods easily introduce edge localization errors, and the selection of filter parameters often requires a compromise between noise suppression and edge preservation, making it difficult to achieve an optimal balance. These problems severely restrict the accuracy of scattering point extraction, thus affecting the performance of subsequent critical tasks in the ISAL system.

[0004] In ISAL image processing, Gaussian filtering is widely used for noise suppression and feature enhancement. However, when dealing with image boundaries, traditional filling strategies (such as zero-fill and symmetrical fill) introduce significant artifacts or darkening effects, leading to a decrease in the accuracy of scattering point extraction near the boundaries. Furthermore, the parameter settings of the Gaussian filter itself have an inherent contradiction: a larger filter kernel can effectively suppress noise but blur target edges; a smaller filter kernel, while preserving details, has limited noise removal capabilities. This contradiction is particularly prominent in ISAL image processing, making it difficult for existing methods to simultaneously meet the requirements of high accuracy and high robustness in scattering point extraction, necessitating improvement. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a method for selecting strong points in space-based ISAL images by combining multi-dimensional features, which solves the technical problems of large boundary errors and difficulty in balancing noise suppression and edge preservation in traditional Gaussian filtering in ISAL image processing.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A method for selecting strong points in space-based ISAL by combining multi-dimensional features includes:

[0010] The modulo operation is performed on the input space-based ISAL complex matrix to generate an amplitude map, and then global normalization is performed.

[0011] The normalized amplitude map is convolved using a Gaussian kernel, and the convolution region that exceeds the image boundary is processed using a pixel copying and extension method to obtain the filtered amplitude map.

[0012] A dynamic threshold is generated based on the product of the scaling factor and the global maximum value. The filtered amplitude map is then binarized to obtain the candidate point mask matrix.

[0013] Perform connected component clustering analysis on the candidate point mask matrix, and filter valid connected components based on a preset area threshold;

[0014] A magnitude-weighted average is calculated on the pixel coordinates within the filtered connected domain to obtain the sub-pixel level coordinates of the centroid position of the strong scattering point;

[0015] Based on sub-pixel level coordinates, multi-frame amplitude sequences of each scattering point are extracted, and the ratio of peak value to mean value is calculated as a temporal stability index to screen the set of strong scattering points that meet the stability conditions.

[0016] A comprehensive scoring model is constructed by combining temporal stability index, spatial amplitude value and spatial distribution characteristics, and a preferred set of scattering points is selected from the set of strong scattering points;

[0017] Among them, the spatial amplitude value is the amplitude value corresponding to the centroid position of each strong scattering point in the normalized amplitude diagram, and the spatial distribution characteristics refer to the density distribution of each strong scattering point in two-dimensional space.

[0018] Preferably, the pixel replication and extension method includes:

[0019] When the upper boundary exceeds the limit, copy the pixel value of the corresponding column in the top row;

[0020] When the lower boundary exceeds the limit, copy the pixel value of the corresponding column in the lowest row.

[0021] When the left boundary exceeds the limit, copy the pixel values ​​of the corresponding row in the leftmost column;

[0022] When the right boundary exceeds the limit, copy the pixel value of the corresponding row in the rightmost column.

[0023] Preferably, the Gaussian kernel has a size of 3×3 and a standard deviation σ∈[0.4,0.6].

[0024] Preferably, the proportionality coefficient α is in the range of 0.25-0.35.

[0025] Preferably, the calculation method for the preset area threshold is as follows:

[0026]

[0027] in, This represents the minimum effective area after normalization. It is a rounding function; , These are the distance resolution and the azimuth resolution, respectively.

[0028] Preferably, the screening threshold for the set of strong scattering points is PMR ≥ 1.5; where PMR is the peak-to-mean ratio of an index used to characterize temporal stability.

[0029] Preferably, the comprehensive scoring model satisfies:

[0030]

[0031]

[0032] in, For the first A comprehensive score for each strong scattering point. For the first Temporal stability index of a strong scattering point For the first The amplitude value corresponding to the centroid position of the normalized amplitude map of each strong scattering point, with subscripts. Indicates the amplitude value, subscript Indicates normalization; and These are the weighting coefficients for the time-domain stability index value and the amplitude value, respectively. =0.5±0.12

[0033] A system for selecting strong points in space-based ISAL by combining multi-dimensional features, comprising:

[0034] The image amplitude calculation and normalization module is used to perform modulo operations on the input space-based ISAL complex matrix to generate an amplitude map and perform global normalization processing.

[0035] The image preprocessing module is used to perform convolution calculation on the normalized amplitude map using a Gaussian kernel, and to process the convolution region that exceeds the image boundary using the pixel copying and extension method to obtain the filtered amplitude map.

[0036] The amplitude threshold filtering module is used to generate a dynamic threshold based on the product of the scaling factor and the global maximum value, and to perform binarization processing on the filtered amplitude map to obtain the candidate point mask matrix.

[0037] The connected component clustering module is used to perform connected component clustering analysis on the candidate point mask matrix and filter valid connected components based on a preset area threshold.

[0038] The subpixel precision centroid estimation module is used to perform amplitude-weighted averaging calculations on the pixel coordinates within the filtered connected domains to obtain the subpixel-level coordinates of the centroid position of the strong scattering point.

[0039] The temporal stability analysis module is used to extract the multi-frame amplitude sequence of each scattering point based on sub-pixel level coordinates, calculate its peak value to mean ratio as a temporal stability index, and screen the set of strong scattering points that meet the stability conditions.

[0040] The weighted scoring and ranking selection module is used to construct a comprehensive scoring model by combining time-domain stability indicators, spatial domain amplitude values ​​and spatial distribution characteristics, and to select the preferred set of scattering points from the set of strong scattering points.

[0041] Among them, the spatial amplitude value is the amplitude value corresponding to the centroid position of each strong scattering point in the normalized amplitude diagram, and the spatial distribution characteristics refer to the density distribution of each strong scattering point in two-dimensional space.

[0042] A storage medium storing a computer program, wherein the computer program causes a computer to perform the method for selecting strong points of space-based ISAL combining multidimensional features as described above.

[0043] An electronic device, the electronic device comprising:

[0044] Processor and memory;

[0045] The memory stores program instructions;

[0046] The processor is configured to run the program instructions to perform the method for selecting strong points of space-based ISAL by combining multi-dimensional features as described above.

[0047] (III) Beneficial Effects

[0048] This invention provides a method for selecting strong points in space-based ISAL systems by combining multi-dimensional features. Compared with existing technologies, it has the following advantages:

[0049] This invention employs a multi-stage fusion processing strategy to form a complete link from image preprocessing to the final strong scattering point output, effectively improving the extraction accuracy and robustness of strong scattering points. By processing the convolution boundary region using a pixel replication and extension method, the boundary amplitude distortion caused by traditional zero-padding or periodic extension is significantly reduced, improving the signal-to-noise ratio of the edge region of the filtered amplitude map. The dynamic threshold generation process works synergistically with the preceding Gaussian kernel convolution preprocessing to suppress speckle noise while preserving the true amplitude characteristics of weakly scattering targets. The amplitude-weighted averaging method is used to calculate the centroid coordinates within the connected domain, overcoming the energy diffusion problem caused by the ISAL system's point spread function. Transient interference points are eliminated based on the peak-to-mean ratio of multi-frame amplitude sequences, and the spatial distribution characteristics of the centroid position's spatial amplitude value and density constraints are combined to ensure that the output set of optimal scattering points simultaneously satisfies temporal stability, spatial saliency, and uniform distribution. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 A flowchart illustrating a method for selecting strong points of space-based ISAL by combining multi-dimensional features, provided in an embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram illustrating a multidimensional weighting method as provided in an embodiment of the present invention.

[0053] Figure 3 A schematic diagram of a satellite model provided for an embodiment of the present invention;

[0054] Figure 4 A schematic diagram of a detection result at a signal-to-noise ratio of 10dB provided in an embodiment of the present invention;

[0055] Figure 5 This is a schematic diagram illustrating the change in detection probability with signal-to-noise ratio, provided as an embodiment of the present invention.

[0056] Figure 6 This is a schematic diagram illustrating the variation of overall positioning accuracy with signal-to-noise ratio, provided as an embodiment of the present invention.

[0057] Figure 7 This is a schematic diagram illustrating the variation of false alarm probability with signal-to-noise ratio, provided as an embodiment of the present invention.

[0058] Figure 8 This is a schematic diagram illustrating the variation of overall detection performance with signal-to-noise ratio, provided by an embodiment of the present invention.

[0059] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0060] Component labeling explanation: 100 - electronic device, 101 - memory, 102 - processor, 103 - display. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] To address the three major challenges of laser speckle noise interference, point spread effect leading to positioning errors, and time-varying scattering causing feature instability, this invention proposes a multi-level fusion processing flow. Through the synergistic effect of multiple technologies, it achieves high-precision and robust extraction of strong scattering points in complex scenes. Specifically, this includes: Gaussian filtering noise reduction technology to suppress various interferences such as speckle noise and impulse noise; adaptive threshold segmentation technology for initial screening of strong scattering points based on image energy distribution; spatial clustering analysis technology for filtering true scattering regions based on the spatial connectivity of scattering points; temporal stability evaluation technology for filtering out strong scattering points that are stable in the time dimension; and sub-pixel centroid estimation technology for accurate positioning of strong scattering points. Through the synergistic effect of these technologies, the algorithm effectively solves the three major challenges faced by traditional methods, achieving high-precision and robust extraction of strong scattering points in complex scenes.

[0063] The method provided in this invention covers the entire process from the input of the original (space-based) ISAL complex image to the final output of the optimal strong scattering point, with each core step logically connected. Specifically, for ISAL complex images, the following processing flow is used to extract highly stable strong scattering points:

[0064] First, the image is preprocessed, including extracting amplitude information, normalization, and noise reduction.

[0065] Subsequently, candidate regions for strong scattering points were initially screened using amplitude thresholds.

[0066] Next, the connected component clustering method is used to screen out the real scattering regions that conform to the physical scattering characteristics, and the scattering points are accurately located by sub-pixel centroid estimation.

[0067] Based on this, temporal stability analysis is used to screen for temporally stable scattering points, and spatial distribution constraints are combined to eliminate points with unreasonable spatial locations.

[0068] Finally, the most representative strong scattering points are selected by weighted score sorting, and the target rotation center is estimated to provide support for subsequent processing.

[0069] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0070] Example 1:

[0071] like Figure 1 As shown, this embodiment of the invention provides a method for selecting strong points in space-based ISAL by combining multi-dimensional features, including:

[0072] S1. Perform modulo operation on the input space-based ISAL complex matrix to generate an amplitude map, and then perform global normalization.

[0073] S2. Perform convolution calculation on the normalized amplitude map using a Gaussian kernel, and process the convolution region that exceeds the image boundary using the pixel copying and extension method to obtain the filtered amplitude map.

[0074] S3. Generate a dynamic threshold based on the product of the scaling factor and the global maximum value, and perform binarization on the filtered amplitude map to obtain the candidate point mask matrix.

[0075] S4. Perform connected component clustering analysis on the candidate point mask matrix and filter the effective connected components according to the preset area threshold;

[0076] S5. Perform amplitude-weighted average calculation on the pixel coordinates within the filtered connected domain to obtain the sub-pixel level coordinates of the centroid position of the strong scattering point.

[0077] S6. Extract the multi-frame amplitude sequence of each scattering point based on sub-pixel level coordinates, calculate the peak-to-mean ratio as a temporal stability index, and screen the set of strong scattering points that meet the stability conditions.

[0078] S7. Combine time-domain stability index, spatial amplitude value and spatial distribution characteristics to construct a comprehensive scoring model, and select the preferred scattering point set from the set of strong scattering points.

[0079] Among them, the spatial amplitude value is the amplitude value corresponding to the centroid position of each strong scattering point in the normalized amplitude diagram, and the spatial distribution characteristics refer to the density distribution of each strong scattering point in two-dimensional space.

[0080] The strong scattering points extracted in this embodiment of the invention are representative, highly stable, and have a reasonable spatial distribution, which can effectively support the ISAL system in subsequent tasks such as target recognition, attitude estimation, and motion compensation.

[0081] The following will detail each step of the above technical solution:

[0082] In step S1, the input space-based ISAL complex matrix is ​​moduloed to generate an amplitude map, and global normalization is performed.

[0083] First, it should be clarified that the original image output by the ISAL imaging system is a complex domain matrix, which contains two core pieces of information: amplitude information and phase information. Amplitude information reflects the electromagnetic scattering intensity of each pixel on the target surface; the amplitude values ​​of strong scattering points are significantly higher than those of the background and weak scattering areas. Phase information mainly reflects the distance between the target and the radar and has no direct effect on the detection of strong scattering points. Therefore, in the extraction of strong scattering points in this embodiment of the invention, only amplitude information needs to be used, and phase information does not need to be considered. First, a modulo operation needs to be performed on the complex domain ISAL image to extract the amplitude information, resulting in an amplitude map.

[0084] Specifically, an ISAL image is an M×N dimensional complex matrix, where M is the number of pixels in the azimuth direction and N is the number of pixels in the range direction. Each element in the matrix corresponds to the complex data of a pixel, and its mathematical expression is:

[0085] (1)

[0086] in, Indicates that the index is pixel data, For azimuth index, For distance-oriented index, ∈[1,M], ∈[1,N], The imaginary unit, Let be the real part of the complex number. This represents the imaginary part of a complex number. The modulo operation converts a complex matrix into a real matrix, i.e., an amplitude diagram. The mathematical expression for the modulo operation is:

[0087] (2)

[0088] in, Indicates that the index is The magnitude (i.e., amplitude) of a pixel, subscript Represents amplitude value, operators For modulo operation, this operation can effectively extract the scattering intensity information of each pixel and form an amplitude map containing only amplitude information.

[0089] Because the transmit power and receive gain of the ISAL system vary with the imaging scene and imaging time, the absolute amplitude values ​​of different frames differ. Directly using the raw amplitude values ​​for subsequent thresholding leads to poor threshold adaptability and increased false alarm and false negative rates. Therefore, it is necessary to normalize the amplitude map to eliminate the influence of system gain differences and environmental factors, providing standardized input for subsequent adaptive thresholding, ensuring the algorithm is insensitive to absolute intensity, and improving the algorithm's versatility and adaptability.

[0090] This step uses the global maximum value normalization method. The specific steps are as follows: First, traverse all pixels in the amplitude image and obtain the global maximum value of the amplitude image. subscript This means taking the maximum value; then, the amplitude value of each pixel in the amplitude map is divided by the global maximum value to obtain the normalized amplitude map. subscript Normalization is represented by the mathematical expression:

[0091] (3)

[0092] The normalized amplitude values ​​range from [0,1], with the normalized amplitude value corresponding to strong scattering points close to 1 and the normalized amplitude value corresponding to the background region close to 0. This effectively highlights the difference between strong scattering points and the background, providing a stable input basis for subsequent adaptive thresholding segmentation. Furthermore, this normalization method is simple and efficient to calculate, requiring no complex parameter adjustments, meeting real-time processing needs, and effectively preserving the relative intensity relationship of strong scattering points without affecting the accuracy of subsequent feature extraction.

[0093] In step S2, the normalized amplitude map is convolved using a Gaussian kernel, and the convolution region that exceeds the image boundary is processed using a pixel copying and extension method to obtain the filtered amplitude map.

[0094] Image preprocessing is a fundamental step in the extraction of strong scattering points. Its core purpose is to suppress various types of noise in the image while preserving the structural and positional information of strong scattering points to the greatest extent possible, thus laying a good foundation for subsequent steps such as threshold segmentation, cluster analysis, and sub-pixel localization.

[0095] In response to the noise characteristics of ISAL images (mainly including Gaussian white noise, speckle noise, and impulse noise), for example, this step uses a 3×3 Gaussian kernel to smooth the amplitude map. Gaussian filtering has advantages such as isotropy, uniform smoothing effect, and effective matching of the ISAL system point spread function. It can suppress isolated noise points while improving the local clustering characteristics of strong scattering points and avoid false detection of scattering points caused by noise interference.

[0096] The core of Gaussian filtering is the Gaussian kernel function, which is essentially a normal distribution function. It possesses the characteristics of smoothness, continuity, and differentiability, and can perform weighted smoothing on pixels in an image. The weight is determined by the distance between the pixel and the kernel center; the closer the pixel, the greater the weight, and the farther the pixel, the smaller the weight. This suppresses noise while preserving image details. The mathematical expression of the 3×3 Gaussian kernel function used in this algorithm is:

[0097] (4)

[0098] in, It is a two-dimensional Gaussian kernel function. The pixel coordinates are within the Gaussian kernel, with the kernel center as the origin (0,0), and the kernel size is 3×3. , The values ​​are all -1, 0, and 1; Pi is a constant. σ is a natural constant; σ is the standard deviation of the Gaussian kernel, used to control the smoothness of Gaussian filtering. The larger σ is, the stronger the smoothing effect and the better the noise suppression ability, but the greater the blurring of image details. σ The smaller the value, the weaker the smoothing effect, the better the image detail is preserved, but the worse the noise suppression ability.

[0099] Preferably, the standard deviation σ ∈ [0.4, 0.6]. Extensive experimental verification has shown that when σ = 0.5, the 3×3 Gaussian kernel achieves the optimal balance between noise suppression and detail preservation. At this point, a normalized discrete kernel matrix can be obtained, where the sum of all elements in the entire matrix is ​​1, ensuring that the brightness of the filtered image does not shift overall. The discrete kernel matrix is ​​as follows:

[0100] (5)

[0101] The specific implementation process of Gaussian filtering is as follows: slide the 3×3 Gaussian kernel sequentially to each pixel of the amplitude map, multiply the amplitude value of each pixel in the kernel by the corresponding kernel weight, and then sum all the products to obtain the filtered amplitude value of the pixel. After traversing all pixels, the filtered amplitude map is obtained.

[0102] It should be noted that, in order to further improve the filtering effect and solve the error problem caused by improper image boundary processing, this step also optimized the boundary processing, noise suppression and edge preservation characteristics of Gaussian filtering. Specifically, these include boundary processing, noise suppression and edge preservation characteristics.

[0103] 1) Boundary processing

[0104] During Gaussian filtering, when the sliding window (Gaussian kernel) moves to the image boundary, the window may extend beyond the image range, preventing normal filtering calculations from being performed on boundary pixels. Traditional boundary processing methods mainly include zero-filling and symmetrical filling, but these methods have significant drawbacks: zero-filling leads to lower amplitude values ​​in the boundary region, resulting in edge darkening and interfering with the detection of strong scattering points at the boundary; symmetrical filling introduces false boundary information, producing artifacts and increasing the localization deviation of boundary scattering points.

[0105] To address the aforementioned issues, this algorithm employs the "replicate" mode in median filtering for boundary processing. The core principle of this mode is that when the sliding window extends beyond the image range, the pixel values ​​of the extended portion are filled with the nearest pixel values ​​on the image boundary. In other words, the boundary pixel values ​​are copied to supplement the missing pixels within the window, ensuring that the filtering calculations for the boundary pixels can proceed normally.

[0106] The specific implementation process is as follows: For the upper boundary of the image, when the upper edge of the window extends beyond the image, the pixel values ​​of the extended portion are copied from the corresponding column of the top row of the image; for the lower boundary, the extended portion is copied from the corresponding column of the bottom row of the image; for the left boundary, the extended portion is copied from the corresponding row of the leftmost column of the image; and for the right boundary, the extended portion is copied from the corresponding row of the rightmost column of the image. This boundary processing mode effectively avoids edge darkening caused by zero-fill and artifacts introduced by symmetrical filling, preserving information from strongly scattering points at the boundary to the maximum extent.

[0107] After practical testing, it was found that using the "replicate" mode for boundary processing reduced the boundary error by 62% compared to the zero-fill method. The positioning deviation of the strong scattering points at the boundary was controlled within 0.1 pixels, which effectively improved the accuracy of boundary scattering point extraction and ensured the consistency of strong scattering point extraction throughout the entire image.

[0108] 2) Noise suppression

[0109] Noise in ISAL images mainly includes Gaussian white noise, speckle noise (γ distribution), and impulse noise. These noises interfere with the detection of strong scattering points, leading to an increase in false high-brightness points. Therefore, noise suppression is one of the core objectives of Gaussian filtering. From a frequency domain perspective, the frequency domain response of the Gaussian kernel is a low-pass filter, which can suppress high-frequency noise in the image (all types of noise belong to high-frequency components) while preserving low-frequency components (the energy concentration region of strong scattering points belongs to low-frequency components), thereby achieving the purpose of noise suppression.

[0110] The mathematical expression for the frequency domain response function of the Gaussian kernel is:

[0111] (6)

[0112] in, It is a two-dimensional Gaussian kernel function. For frequency domain coordinates, For the frequency domain distance coordinates, Using the frequency domain azimuth coordinates, this low-pass filter attenuates to -20dB at the Nyquist frequency (i.e., half the image sampling frequency). At this time, high-frequency noise can be effectively suppressed, while avoiding excessive attenuation of low-frequency components, ensuring that the energy information of strong scattering points is not lost.

[0113] To quantify the noise suppression effect of Gaussian filtering, Peak Signal-to-Noise Ratio (PSNR) is used as the evaluation metric. A higher PSNR value indicates better noise suppression and superior image quality. Through extensive experimental testing, the 3×3 Gaussian filter used in this algorithm (…) The suppression effect on various types of noise is shown in Table 1, which can effectively suppress the main noise types in ISAL images.

[0114] Table 1. Suppression levels for different noise levels

[0115]

[0116] As shown in Table 1, Gaussian filtering exhibits the best suppression effect on Gaussian white noise, improving PSNR by 8-12 dB, effectively eliminating the interference of Gaussian white noise on the detection of strong scattering points. Its suppression effect on speckle noise is second best, improving PSNR by 6-9 dB, significantly reducing false high-brightness points caused by speckle noise. It also has a certain suppression effect on impulse noise, improving PSNR by 3-5 dB, reducing the impact of isolated impulse noise points. Overall, the design of these Gaussian filter parameters meets the noise suppression requirements of ISAL images, providing high-quality image input for subsequent strong scattering point extraction.

[0117] 3) Edge protection characteristics

[0118] While suppressing noise, preserving the edge and position information of strong scattering points to the greatest extent possible is another core requirement of Gaussian filtering. Poor edge preservation characteristics can lead to blurred edges and positional shifts of strong scattering points, affecting the accuracy of subsequent sub-pixel localization. To quantify the edge preservation characteristics of Gaussian filtering, this algorithm analyzes the response function of an ideal step edge, calculates the edge localization error, and compares it with traditional median filtering to verify the edge preservation advantages of Gaussian filtering.

[0119] The mathematical expression for the edge response function (response after Gaussian filtering) of an ideal step edge is:

[0120] (7)

[0121] in, The edge response function of an ideal step edge after Gaussian filtering. Let be the error function. The coordinates are the coordinates of the edge normal direction. The standard deviation of the Gaussian kernel is given. Edge positioning error refers to the deviation between the actual position of the filtered edge and the ideal edge position. It is calculated by finding the coordinates corresponding to the 50% amplitude point of the edge response function and the coordinates of the ideal edge.

[0122] Table 2 shows the positioning deviations and causes of different types of edges, which can intuitively reflect the edge preservation characteristics of Gaussian filtering.

[0123] Table 2 Comparison of Edge Positioning Deviation

[0124]

[0125] To further verify the edge-preserving advantage of Gaussian filtering, it was compared with traditional median filtering. Three core indicators were selected for evaluation: edge preservation index, corner retention rate, and texture distortion. The specific comparison results are shown in Table 3.

[0126] Table 3 Comparison of Gaussian filtering and median filtering

[0127]

[0128] The comparison results above show that the edge preservation index of Gaussian filtering is significantly higher than that of median filtering, indicating that it can better preserve edge information; the corner retention rate is higher than that of median filtering, which can effectively preserve the corner-like strong scattering points of the target surface; and the texture distortion is significantly lower than that of median filtering, indicating that it has less damage to image texture and can better preserve the spatial structure of strong scattering points.

[0129] Gaussian filtering, as a core preprocessing step, suppresses noise while preserving the scattering point structure to the maximum extent through isotropic smoothing. Its technical advantages are specifically reflected in the following four aspects, ensuring the superiority of the preprocessing effect and its engineering adaptability:

[0130] i) Kernel function design

[0131] A 3×3 kernel with σ=0.5 achieves the optimal balance between noise suppression and detail preservation. Specifically, Gaussian filtering with this parameter can achieve an 8dB noise suppression effect, while keeping the edge blurring level within 0.15 pixels. This effectively suppresses noise without excessively blurring the edge information of strong scattering points, ensuring the accuracy of subsequent positioning.

[0132] Boundary processing: The 'replicate' mode can effectively reduce the loss of boundary information, reducing the loss ratio by more than 60%, avoiding the omission of strong scattering points at the boundary and the positioning deviation, and ensuring the consistency of strong scattering point extraction throughout the entire image range.

[0133] ii) Physical compatibility

[0134] Gaussian distribution can match the point spread function (PSF) characteristics of ISAL system well. The point spread function of ISAL system is essentially approximately Gaussian distributed. Therefore, using Gaussian filtering for preprocessing can better compensate for the scattering point diffusion caused by the point spread effect and improve the positioning accuracy of strong scattering points.

[0135] iii) Computational efficiency

[0136] Gaussian filtering is achieved by using a separable convolution algorithm, which decomposes the 3×3 Gaussian kernel into two 1×3 one-dimensional kernels (horizontal and vertical) and performs convolution operations separately. Compared with directly performing convolution calculations with 3×3 kernels, this method can speed up the process by 33%, meeting the real-time processing requirements of the space-based ISAL system.

[0137] After practical testing, this preprocessing step reduced the error rate of subsequent threshold segmentation and clustering analysis by more than 40%, effectively reducing false detections of spurious scattering points and missed detections of real scattering points, laying a solid foundation for high-precision scattering point extraction.

[0138] In step S3, a dynamic threshold is generated based on the product of the scaling factor and the global maximum value, and the filtered amplitude map is binarized to obtain the candidate point mask matrix.

[0139] Amplitude threshold screening is the core step in the initial screening of strong scattering points. Its core principle is to utilize the difference in radiation intensity between the target strong scatterer and the background and weak scattering regions. By setting a reasonable amplitude threshold, the candidate regions of strong scattering points are initially screened, retaining only high-energy candidate regions and eliminating low-energy background and weak scattering regions. This reduces the amount of data for subsequent processing, improves the computational efficiency of the algorithm, and reduces the processing difficulty of subsequent steps.

[0140] Traditional thresholding methods often use a fixed threshold, where a fixed amplitude value is set as the threshold. Pixels above the threshold are considered candidate strong scattering points, while those below are considered background. However, the amplitude value of ISAL images varies with the imaging scene, system parameters, and environmental factors. A fixed threshold cannot adapt to the needs of different scenarios and is prone to problems such as missed detections (when the threshold is too high, some weak and strong scattering points are mistakenly identified as background) or false alarms (when the threshold is too low, some background noise is mistakenly identified as candidate strong scattering points).

[0141] To address the aforementioned issues, this step employs an adaptive thresholding method. The threshold is dynamically adjusted based on the maximum intensity of the image, ensuring its adaptability. The core idea is to use the global maximum value of the normalized amplitude map as a benchmark, setting a scaling factor α. The threshold is the product of the global maximum value and the scaling factor α, expressed mathematically as follows:

[0142] (8)

[0143] Where T is the adaptive threshold. α This is the proportionality coefficient. The global maximum value of the normalized amplitude plot is 1. Therefore, the actual value of the threshold T is equal to the scaling factor α.

[0144] Preferably, the scaling factor α ranges from 0.25 to 0.35. The value of scaling factor α is crucial, directly affecting the effectiveness of threshold screening, and requires optimization through extensive experiments. This step uses Monte Carlo experiments to determine the optimal value of α. The specific experimental process is as follows: ISAL image datasets with different signal-to-noise ratios (SNR = 3dB~10dB) and different target types are constructed. For each dataset, different α values ​​(0.1~0.5) are selected for threshold screening. The false negative rate and false alarm rate corresponding to each α value are statistically analyzed. Finally, the α value that achieves a balance between a false negative rate <5% and a false alarm rate <10% is selected as the optimal parameter. Experimental verification shows that when α = 0.3, the threshold screening effect is optimal, enabling effective initial screening of strong scattering points in different scenarios while controlling the false negative rate and false alarm rate within a reasonable range.

[0145] The setting of this adaptive threshold has a clear physical basis: In ISAL images, the energy of strong scattering points usually occupies the upper 30% of the energy distribution of the entire image, while the energy of the background and weak scattering areas occupies the lower 70%. Therefore, selecting a 30% scaling factor α can effectively distinguish strong scattering points from the background and weak scattering areas, ensuring that the candidate area mainly contains strong scattering points, while eliminating most of the background noise and weak scattering points.

[0146] After determining the adaptive threshold T, the normalized amplitude image is binarized to generate a candidate point mask matrix. The specific rule is as follows: for each pixel in the normalized amplitude image, if its amplitude value is greater than or equal to the threshold T, it is determined as a candidate strong scattering point, and the corresponding position in the mask matrix is ​​set to 1; if its amplitude value is less than the threshold T, it is determined as background or a weak scattering point, and the corresponding position in the mask matrix is ​​set to 0. The mathematical expression for binarization is:

[0147] (9)

[0148] in, The candidate point mask matrix takes values ​​of 0 or 1. Regions with a value of 1 in the mask matrix are candidate regions for strong scattering points. This binarization process separates candidate regions for strong scattering points from the background, forming a clear candidate region mask. This provides a clear processing target for subsequent connected component clustering analysis, while significantly reducing the amount of data to be processed and improving the computational efficiency of the algorithm.

[0149] In step S4, connected component clustering analysis is performed on the candidate point mask matrix, and valid connected components are selected based on a preset area threshold.

[0150] After amplitude thresholding, the resulting candidate point mask matrix contains not only real strong scattering point regions but also a small number of isolated noise points (such as speckle noise and impulse noise not completely suppressed by Gaussian filtering). These isolated noise points typically appear as a single or a few adjacent pixels with a value of 1, and do not have any actual physical scattering significance. Furthermore, real strong scattering points, due to the point diffusion effect, have their energy diffused spatially, appearing as multiple adjacent pixels with a value of 1, forming a connected region, corresponding to a real scattering center on the target surface.

[0151] Therefore, connected component clustering analysis is needed to eliminate isolated noise points and filter out the connected regions of true strong scattering points. Connected component clustering is essentially an electromagnetic scattering spatial coherence detector. Its core principle is that true strong scattering points possess spatial coherence, meaning that strong scattering points corresponding to the same scattering center are spatially adjacent, while isolated noise points lack spatial coherence. Therefore, by labeling adjacent high-amplitude pixel blocks (connected components) and using features such as the area of ​​connected regions to filter out significant areas, the distinction between true strong scattering point regions and isolated noise points can be achieved.

[0152] Preferably, this step uses 8-connectivity for connected component labeling. 8-connectivity means that pixels in the eight neighborhood directions (up, down, left, right, top-left, top-right, bottom-left, bottom-right) of a pixel are considered adjacent pixels. This connectivity rule conforms to the expansion characteristics of strong scattering points in ISAL images—due to the point diffusion effect, the energy of a strong scattering point will diffuse in eight neighborhood directions, forming an 8-connected region. The mathematical model of an 8-connected region is as follows:

[0153] (10)

[0154] in, For the first One connected component; These are the pixel coordinates in the image coordinate system. This indicates that the pixel is a candidate point in the binary mask; This indicates the existence of a connected component. pixels in , It is an existential quantifier; Represents pixels and The maximum coordinate difference between them does not exceed 1, meaning they are 8-neighbors. To find the maximum value operator, For the absolute value operator; Represents the logical AND operator.

[0155] Understandably, this formula defines a spatially aggregated physical scattering unit, which means that when multiple candidate pixel points are spatially adjacent in their 8-neighborhood, they are aggregated into a physically meaningful scattering structure, corresponding to a specific scattering center on the target. Essentially, this connected component clustering process imposes dual constraints on the region in terms of energy and spatial continuity. For strong scattering points, both the energy constraint (amplitude value above a threshold) and the spatial structure continuity constraint (8-connectivity) must be satisfied simultaneously to be identified as a potential scattering point region, thus effectively distinguishing real scattering points from isolated noise points.

[0156] After labeling 8 connected components, multiple connected components are obtained. At this point, it is necessary to further filter out noise regions that do not conform to the physical scattering characteristics (i.e., small connected components corresponding to isolated noise points). The true scattering center is affected by the point spread function (PSF) and occupies a specific size in the image. Isolated noise points are usually isolated points, and the corresponding connected components have extremely small areas. Therefore, by setting a minimum effective size threshold, connected components with excessively small areas can be eliminated, while connected components with areas that conform to the physical characteristics can be retained.

[0157] The determination model for connected component area screening is as follows:

[0158] (11)

[0159] in, For the filtered first One valid connected component; For the first The area of ​​each connected region (the number of pixels within the region); The minimum effective area threshold (i.e., the minimum effective area after normalization) is derived based on the point spread function characteristics of the ISAL system.

[0160] Specifically, for the ISAL image after imaging processing, its two-dimensional point spread function is theoretically modeled as the sinc function (Singer function), with the mathematical expression as follows:

[0161] (12)

[0162] in, For two-dimensional point spread function, , Here, x represents the range resolution and azimuth resolution of the ISAL system, respectively, where x is the range coordinate and y is the azimuth coordinate. Based on the properties of the sinc function, its first main lobe width (the width corresponding to the first null point) is approximately:

[0163] (13)

[0164] That is, the width of the first main lobe of the point spread function in the range direction. for In the azimuth width for The energy of a strong scattering point is mainly concentrated in the first main lobe of the point spread function. Therefore, the area of ​​the connected region corresponding to a strong scattering point should be approximately equal to the area of ​​the first main lobe.

[0165] Distance resolution and azimuth resolution Normalized to the pixel dimension (assuming 1 pixel corresponds to 1 resolution unit), the area of ​​the first main lobe is: Therefore, the minimum effective area after normalization for:

[0166] (14)

[0167] in, The function rounds up to the nearest integer, ensuring the minimum effective area is an integer pixel. By filtering using this minimum effective area threshold, small connected regions (with areas less than a certain value) corresponding to isolated noise points can be effectively removed. Preserve the connected components (areas greater than or equal to) corresponding to the real strong scattering points. This further improves the purity of candidate strong scattering point regions, laying the foundation for subsequent sub-pixel localization.

[0168] In step S5, the pixel coordinates within the filtered connected domain are calculated by amplitude-weighted averaging to obtain the sub-pixel level coordinates of the centroid position of the strong scattering point.

[0169] After connected component clustering and filtering, the connected components corresponding to the true strong scattering points are obtained. At this point, it is necessary to determine the center position of each connected component, i.e., the precise coordinates of the strong scattering points. Traditional pixel-level positioning methods (such as taking the center pixel coordinates of the connected components) have low positioning accuracy, only achieving a positioning error of 1 pixel, which cannot meet the requirements of subsequent motion compensation and attitude estimation for sub-pixel-level positioning accuracy (target positioning error < 0.1 pixels).

[0170] To address the aforementioned issues, this step employs a weighted average method based on pixel intensity within a region to calculate centroid coordinates, achieving sub-pixel-level feature point location estimation. The core principle of this method is that the energy distribution of strong scattering points follows the characteristics of a point spread function (PSF). Different pixels within a connected region have different amplitude values ​​(scattering intensities). Pixels with higher intensities are closer to the true center of the strong scattering point. Therefore, assigning greater weight to pixels with higher intensities and calculating centroid coordinates through weighted averaging achieves higher accuracy than pixel-level localization.

[0171] The mathematical model for weighted centroid estimation is as follows:

[0172] (15)

[0173] in, For the first The centroid coordinates of each connected region (strong scattering point), i.e., the sub-pixel level coordinates of the strong scattering point; For connected components Pixel coordinates within; The amplitude value of the pixel (unnormalized) is used as a weighting coefficient; the numerator is the sum of the products of the pixel coordinates and the corresponding amplitude values, and the denominator is the sum of the amplitude values ​​of all pixels in the connected domain. By calculating this weighted average, we can obtain coordinates that are closer to the true center of the strong scattering point.

[0174] To verify the accuracy advantage of the weighted centroid estimation method, it was compared with traditional pixel-level positioning methods and Gaussian fitting positioning methods. Three core indicators were selected for evaluation: theoretical accuracy, computational complexity, and noise resistance. The specific comparison results are shown in Table 4.

[0175] Table 4 Comparison of Gaussian filtering and median filtering

[0176]

[0177] The comparison results above show that the weighted centroid estimation method has significant advantages: its theoretical accuracy (0.1-0.3 pixels) is much higher than that of pixel-level positioning methods, which can meet the needs of sub-pixel-level positioning; its computational complexity is O(N), which is much lower than that of Gaussian fitting methods (O(N²)), making it computationally efficient and able to meet the needs of real-time processing; and it has excellent noise resistance, as the influence of noisy pixels can be suppressed by amplitude weighting, ensuring the stability of the positioning results.

[0178] Compared to Gaussian fitting, the weighted centroid estimation method, while theoretically having slightly lower accuracy, significantly reduces computational complexity and offers better noise resistance, making it more suitable for the real-time processing requirements of space-based ISAL systems. Compared to pixel-level positioning methods, its accuracy is significantly improved, meeting the high-precision requirements of subsequent tasks. Therefore, this algorithm selects the weighted centroid estimation method to achieve sub-pixel-level positioning of strong scattering points, balancing positioning accuracy, computational efficiency, and noise resistance.

[0179] In step S6, the amplitude sequence of each scattering point is extracted based on sub-pixel level coordinates, and the ratio of its peak value to mean value is calculated as a temporal stability index to screen the set of strong scattering points that meet the stability conditions.

[0180] After subpixel centroid estimation, the precise coordinates of strong scattering points in a single frame image are obtained. However, ISAL imaging is typically performed across multiple consecutive frames. During target motion, the surface scattering characteristics dynamically change with attitude and viewing angle. Some scattering points may experience energy decay, disappearance, or new appearance. These time-varying scattering points have poor stability and cannot be used as reference points for subsequent tasks such as motion compensation, target tracking, and attitude estimation. Therefore, a temporal stability analysis step is needed to screen out strong scattering points that are stable in the time dimension, thereby improving the reliability and stability of strong scattering points.

[0181] It should be noted that the core idea of ​​temporal stability analysis is to extract the time series of the distance cell corresponding to each candidate strong scattering point (i.e., the amplitude value change sequence of the distance cell in multiple consecutive frames of images), and quantify the temporal stability of the candidate strong scattering point by calculating the peak-to-mean ratio (PMR) of the time series. The higher the PMR value, the more stable the amplitude value of the scattering point is in the time dimension, and the more likely it is to be a real strong scattering point; the lower the PMR value, the greater the amplitude value fluctuation of the scattering point, the worse the stability, and the more likely it is to be a time-varying scattering point or a noise point.

[0182] Specifically, the distance unit for each candidate strong scattering point is first determined. (i.e., centroid coordinates) Distance coordinates in (corresponding distance cell); then extract continuous This distance unit in the frame image amplitude value sequence ,in The number of consecutive imaging frames (usually taken as...) (To ensure the representativeness of the time series), then calculate the peak value of the time series. (Maximum value and mean in a time series) (The average value of the time series); finally, calculate the peak-to-mean ratio (PMR), the mathematical expression of which is:

[0183] (16)

[0184] in, , This refers to the azimuth time (i.e., the frame number). .

[0185] The physical significance of time-domain stability analysis lies in the fact that PMR is essentially a time-varying measure of the scattering cross section (RCS). True strong scattering points correspond to the fixed physical structure of the target surface, and their scattering cross section is relatively stable in the time dimension. Therefore, the corresponding amplitude value fluctuates less in the time series, and the ratio of the peak value to the mean is larger, that is, the PMR value is high. On the other hand, time-varying scattering points or noise points have large fluctuations in their scattering cross section, and the corresponding amplitude value fluctuates violently in the time series, and the ratio of the peak value to the mean is smaller, that is, the PMR value is low.

[0186] To achieve the screening of stable strong scattering points, a PMR threshold needs to be set. After extensive experimental verification, when The optimal screening effect is achieved when the retention rate of stable strong scattering points is ≥90%, and the rejection rate of time-varying scattering points is ≥85%. The specific screening rule is: if the PMR value of a candidate strong scattering point is greater than or equal to... If the PMR value is less than 1, it is determined to be a time-domain stable strong scattering point and is retained; if the PMR value is less than 1, it is considered a strong scattering point and is retained. If a point is found to be unstable in the time domain, it will be discarded.

[0187] By performing a time-domain stability analysis step, time-varying scattering points and noise points can be effectively eliminated, while retaining strong scattering points that are stable in the time dimension. This ensures that the extracted strong scattering points have good stability, which can meet the needs of subsequent tasks such as continuous imaging, motion compensation, and target tracking, and further improves the robustness and reliability of the algorithm.

[0188] In step S7, a comprehensive scoring model is constructed by combining the temporal stability index, spatial amplitude value and spatial distribution characteristics, and a preferred set of scattering points is selected from the set of strong scattering points.

[0189] After temporal stability analysis, a set of temporally stable strong scattering points was obtained. However, this set may contain multiple strong scattering points with varying amplitudes, intensities, and temporal stability. Not all scattering points can provide effective support for subsequent tasks. Therefore, a comprehensive evaluation model needs to be constructed to calculate the comprehensive weighted score of each strong scattering point. The optimal N strong scattering points are selected by ranking the scores to ensure that the extracted strong scattering points are representative, stable, and energy significant, thus maximizing their support for subsequent tasks.

[0190] like Figure 2As shown, this step fully considers the temporal stability, spatial intensity, and spatial distribution characteristics of the target in the ISAL image, and constructs a comprehensive evaluation model. The calculation of the comprehensive weighted score combines the temporal stability index (PMR value) and spatial intensity index (normalized amplitude value) of each strong scattering point, while implicitly incorporating the constraints of spatial distribution characteristics. Its mathematical expression is:

[0191] (17)

[0192] in, For the first The overall score of each strong scattering point; For the first The temporal stability index (PMR value) of a strong scattering point is dimensionless. For the first The normalized intensity of a strong scattering point (i.e., the normalized amplitude value corresponding to its centroid position) is dimensionless and ranges from [0,1]. and The weighting coefficients for the PMR value and the normalized intensity are respectively, satisfying... .

[0193] It is important to note that while the comprehensive weighted score superficially includes two indicators—temporal stability and spatial intensity—it implicitly incorporates spatial distribution characteristics. The reason spatial distribution constraints are implicit in the comprehensive evaluation model is that high-scoring, strong scattering points typically correspond to the core scattering structures on the target surface. These core scattering structures themselves exhibit a certain degree of uniformity on the target, avoiding excessive concentration in any particular area. Therefore, there is no need to include spatial distribution characteristics as an independent indicator in the weighted calculation; further optimization can be achieved through a spatial distribution filtering step.

[0194] Preferred, =0.5±0.12. The weighting coefficients were set based on the requirements of subsequent tasks. Since temporal stability and spatial intensity have equally important effects on subsequent tasks, after experimental optimization, the following settings were adopted: and This ensures that temporal stability and spatial intensity have equal weight in the overall score, avoiding both the retention of unstable scattering points caused by considering only spatial intensity and the retention of low-energy scattering points caused by considering only temporal stability.

[0195] The above definition of the comprehensive score implicitly includes amplitude normalization, that is, using the normalized amplitude value as the spatial intensity index, avoiding the situation where absolute intensity dominates the comprehensive score, ensuring the comparability of the comprehensive scores of strong scattering points in different frames and different scenes, and improving the versatility of the comprehensive evaluation model.

[0196] The selection strategy of strong scattering points is divided into two steps to ensure the rationality and adaptability of the selection results:

[0197] The first step: Score ranking.

[0198] Calculate the comprehensive score of each strong scattering point according to the comprehensive evaluation model , and sort all strong scattering points in descending order of scores. The higher the score of a strong scattering point, the better its time-domain stability, the higher its spatial domain intensity, and the more representative it is.

[0199] The second step: Adaptive selection.

[0200] According to the actual task requirements, set the number N of strong scattering points to be selected, and select the top N points with the highest scores from the sorted strong scattering points as the final strong scattering points. At the same time, introduce an adaptive mechanism. When the number of effective strong scattering points is less than N (that is, after screening through all previous steps, the number of stable strong scattering points < N), automatically reduce the selection amount, and use all the actually searched stable strong scattering points as the selection results to avoid misselection of false scattering points caused by forcibly selecting N points.

[0201] Through the weighted score and ranking selection steps, the optimal strong scattering points can be screened out from the set of stable strong scattering points. These strong scattering points have good time-domain stability and high spatial domain intensity, and are highly representative. They can provide high-quality reference points for subsequent tasks such as rotation center estimation, motion compensation, target recognition, and attitude estimation, ensuring the performance of subsequent tasks.

[0202] So far, the entire process of the method for selecting strong points in space-based ISAL combined with multi-dimensional features has been fully introduced. In summary, the embodiments of the present invention form a complete link from image preprocessing to the output of final strong scattering points through a multi-stage fusion processing strategy. Each stage is closely linked and collaboratively optimized. The specific processing process is as follows: First, preprocess the ISAL complex image, extract the amplitude information and perform normalization and noise reduction processing; then, initially screen the candidate regions of strong scattering points through amplitude threshold screening; next, screen out the real scattering regions that meet the physical scattering characteristics through connected domain clustering analysis; then, accurately locate the strong scattering points through sub-pixel centroid estimation; then, screen out the stable strong scattering points in the time dimension through time-domain stability analysis; subsequently, eliminate unreasonable scattering points through spatial distribution constraints; finally, select the optimal strong scattering points through weighted score ranking and estimate the target rotation center to provide support for subsequent tasks. The finally extracted strong scattering points are characterized by strong representativeness, high stability, and reasonable spatial distribution, and can effectively support subsequent tasks such as target recognition, attitude estimation, and motion compensation of the ISAL system.

[0203] To further help understand the superiority of the embodiments of the present invention, the following provides specific simulation verification:

[0204] The detection success rate and false alarm rate under different signal-to-noise ratios were simulated. The target was a satellite model, which included 17 strong point targets, such as... Figures 3-4 As shown.

[0205] Detection performance such as Figures 4-8 As shown, when the echo signal-to-noise ratio is -10dB or higher, it can achieve 100% detection of strong point targets.

[0206] Example 2:

[0207] This invention provides a system for selecting strong points in space-based ISAL systems by combining multi-dimensional features, characterized by comprising:

[0208] The image amplitude calculation and normalization module is used to perform modulo operations on the input space-based ISAL complex matrix to generate an amplitude map and perform global normalization processing.

[0209] The image preprocessing module is used to perform convolution calculation on the normalized amplitude map using a Gaussian kernel, and to process the convolution region that exceeds the image boundary using the pixel copying and extension method to obtain the filtered amplitude map.

[0210] The amplitude threshold filtering module is used to generate a dynamic threshold based on the product of the scaling factor and the global maximum value, and to perform binarization processing on the filtered amplitude map to obtain the candidate point mask matrix.

[0211] The connected component clustering module is used to perform connected component clustering analysis on the candidate point mask matrix and filter valid connected components based on a preset area threshold.

[0212] The subpixel precision centroid estimation module is used to perform amplitude-weighted averaging calculations on the pixel coordinates within the filtered connected domains to obtain the subpixel-level coordinates of the centroid position of the strong scattering point.

[0213] The temporal stability analysis module is used to extract the multi-frame amplitude sequence of each scattering point based on sub-pixel level coordinates, calculate its peak value to mean ratio as a temporal stability index, and screen the set of strong scattering points that meet the stability conditions.

[0214] The weighted scoring and ranking selection module is used to construct a comprehensive scoring model by combining time-domain stability indicators, spatial domain amplitude values ​​and spatial distribution characteristics, and to select the preferred set of scattering points from the set of strong scattering points.

[0215] Among them, the spatial amplitude value is the amplitude value corresponding to the centroid position of each strong scattering point in the normalized amplitude diagram, and the spatial distribution characteristics refer to the density distribution of each strong scattering point in two-dimensional space.

[0216] The strong scattering points extracted in this embodiment of the invention are representative, highly stable, and have a reasonable spatial distribution, which can effectively support the ISAL system in subsequent tasks such as target recognition, attitude estimation, and motion compensation.

[0217] In an optional implementation, the pixel replication extension method includes:

[0218] When the upper boundary exceeds the limit, copy the pixel value of the corresponding column in the top row;

[0219] When the lower boundary exceeds the limit, copy the pixel value of the corresponding column in the lowest row.

[0220] When the left boundary exceeds the limit, copy the pixel values ​​of the corresponding row in the leftmost column;

[0221] When the right boundary exceeds the limit, copy the pixel value of the corresponding row in the rightmost column.

[0222] In an alternative implementation, the Gaussian kernel is 3×3 in size and has a standard deviation σ∈[0.4,0.6].

[0223] In an optional implementation, the scaling factor α ranges from 0.25 to 0.35.

[0224] In an optional implementation, the preset area threshold is calculated as follows:

[0225]

[0226] in, This represents the minimum effective area after normalization. It is a rounding function; , These are the distance resolution and the azimuth resolution, respectively.

[0227] In an optional implementation, the screening threshold for the set of strong scattering points is PMR ≥ 1.5; where PMR is the peak-to-mean ratio of an index used to characterize temporal stability.

[0228] In an optional implementation, the comprehensive scoring model satisfies:

[0229]

[0230]

[0231] in, For the first A comprehensive score for each strong scattering point. For the first Temporal stability index of a strong scattering point For the first The amplitude value corresponding to the centroid position of the normalized amplitude map of each strong scattering point, with subscripts. Indicates the amplitude value, subscript Indicates normalization; and These are the weighting coefficients for the time-domain stability index value and the amplitude value, respectively. =0.5±0.12.

[0232] Example 3:

[0233] This invention provides a storage medium storing a computer program that causes a computer to execute a method for selecting strong points in space-based ISAL combining multidimensional features, as provided in any embodiment of this invention.

[0234] In embodiments of the present invention, any combination of one or more storage media may be used. The storage medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, RAM, ROM, an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium that contains or stores a program that may be used by or in connection with an instruction execution system, apparatus, or device.

[0235] Example 4:

[0236] This invention provides an electronic device. Figure 9 The diagram shown is a structural schematic of the electronic device 100 provided in an embodiment of the present invention. In some embodiments, the electronic device may be a mobile phone, tablet computer, wearable device, in-vehicle device, augmented reality (AR) / virtual reality (VR) device, laptop computer, ultra-mobile personal computer (UMPC), netbook, personal digital assistant (PDA), or other terminal device. Furthermore, the space-based ISAL strong point selection method combining multi-dimensional features provided in this embodiment can also be applied to databases, servers, and service response systems based on terminal artificial intelligence. This embodiment does not limit the specific application scenarios of the space-based ISAL strong point selection method combining multi-dimensional features.

[0237] like Figure 9 As shown, the electronic device 100 provided in this embodiment of the invention includes a memory 101 and a processor 102.

[0238] The memory 101 is used to store computer programs; preferably, the memory 101 includes various media that can store program code, such as ROM, RAM, magnetic disk, USB flash drive, memory card or optical disk.

[0239] Specifically, memory 101 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. Electronic device 100 may further include other removable / non-removable, volatile / non-volatile computer system storage media. Memory 101 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0240] The processor 102 is connected to the memory 101 and is used to execute the computer program stored in the memory 101 so that the electronic device 100 performs the method for selecting strong points of space-based ISAL combining multi-dimensional features provided in any embodiment of the present invention.

[0241] In an optional implementation, the processor 102 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, evaluation index gate or transistor logic devices, or evaluation index hardware components.

[0242] In an optional embodiment of the invention, the electronic device 100 may further include a display 103. The display 103 is communicatively connected to the memory 101 and the processor 102, and is used to display a GUI interface related to the method for selecting strong points of space-based ISAL combining multi-dimensional features.

[0243] It is understood that the system, storage medium, and electronic device for space-based ISAL strong point selection combining multi-dimensional features provided in the embodiments of the present invention correspond to the method for space-based ISAL strong point selection combining multi-dimensional features provided in the embodiments of the present invention. The explanations, examples, and beneficial effects of the relevant contents can be referred to the corresponding parts of the method, and will not be repeated here.

[0244] In summary, compared with existing technologies, it has the following beneficial effects:

[0245] This invention addresses three core challenges in strong scattering point extraction for space-based ISAL systems: large interference from laser speckle noise, positioning deviation caused by point diffusion effect, and feature instability caused by time-varying scattering. It proposes a strong scattering point selection method that combines multi-dimensional features and full-process collaborative optimization. Through multi-technology fusion and parameter optimization, it achieves high-precision and robust extraction of strong scattering points in complex scenarios, fully meeting the technical requirements of subsequent core tasks of space-based ISAL systems such as motion compensation, attitude estimation, target recognition, and target reconstruction.

[0246] This method's entire workflow design forms a complete link from the original ISAL complex image input to the optimal strong scattering point output. Each link is interconnected and synergistically empowers the process. Each processing step includes targeted innovative designs with no redundant modules. Specific core highlights are as follows: Image amplitude calculation and normalization adopt global maximum value normalization to eliminate system gain differences and provide standardized input for subsequent adaptive threshold selection; the image preprocessing stage optimizes the design with 3×3 Gaussian filtering (σ=0.5), combined with a replicate boundary processing mode, to achieve the optimal balance between noise suppression and detail preservation, while adapting to the point spread function characteristics of the ISAL system; amplitude threshold selection abandons the traditional hard threshold and adopts CA-CFAR constant false alarm rate technology to achieve pixel-by-pixel adaptive threshold adjustment, solving the problem of poor adaptability of traditional methods; connected component clustering adopts 8 connectivity rules and minimum effective area constraints to accurately distinguish between real scattering areas and isolated noise points, conforming to the spatial diffusion characteristics of strong scattering points; sub-pixel accuracy centroid estimation adopts the intensity-weighted average method, breaking through the pixel-level positioning bottleneck and achieving the goal of high-precision positioning error at the sub-pixel level.

[0247] In summary, the method provided by this invention effectively solves the application pain points of traditional methods in space-based ISAL scenarios through full-process collaborative optimization and detailed innovation in each link. It takes into account positioning accuracy, robustness, real-time performance and engineering feasibility. The modular design facilitates subsequent engineering implementation and parameter debugging. It can provide core technical support for improving the performance of space-based ISAL systems, and has important theoretical significance and broad engineering application value. It can be further extended to related fields such as complex airspace target monitoring and high-precision laser imaging.

[0248] 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 apparatus 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 apparatus. 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 apparatus that includes said element.

[0249] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for space-based ISAL hot spot selection incorporating multi-dimensional features, characterized in that, include: The modulo operation is performed on the input space-based ISAL complex matrix to generate an amplitude map, and then global normalization is performed. The normalized amplitude map is convolved using a Gaussian kernel, and the convolution region that exceeds the image boundary is processed using a pixel copying and extension method to obtain the filtered amplitude map. A dynamic threshold is generated based on the product of the scaling factor and the global maximum value. The filtered amplitude map is then binarized to obtain the candidate point mask matrix. Perform connected component clustering analysis on the candidate point mask matrix, and filter valid connected components based on a preset area threshold; A magnitude-weighted average is calculated on the pixel coordinates within the filtered connected domain to obtain the sub-pixel level coordinates of the centroid position of the strong scattering point; Based on sub-pixel level coordinates, multi-frame amplitude sequences of each scattering point are extracted, and the ratio of peak value to mean value is calculated as a temporal stability index to screen the set of strong scattering points that meet the stability conditions. A comprehensive scoring model is constructed by combining temporal stability index, spatial amplitude value and spatial distribution characteristics, and a preferred set of scattering points is selected from the set of strong scattering points; Among them, the spatial amplitude value is the amplitude value corresponding to the centroid position of each strong scattering point in the normalized amplitude diagram, and the spatial distribution characteristics refer to the density distribution of each strong scattering point in two-dimensional space.

2. The method of claim 1, wherein, The pixel replication and extension method includes: When the upper boundary exceeds the limit, copy the pixel value of the corresponding column in the top row; When the lower boundary exceeds the limit, copy the pixel value of the corresponding column in the lowest row. When the left boundary exceeds the limit, copy the pixel values ​​of the corresponding row in the leftmost column; When the right boundary exceeds the limit, copy the pixel value of the corresponding row in the rightmost column.

3. The method of claim 1, wherein, The Gaussian kernel has a size of 3×3 and a standard deviation σ∈[0.4,0.6].

4. The method of claim 1, wherein, The proportionality coefficient α ranges from 0.25 to 0.

35.

5. The method of claim 1, wherein, The calculation method for the preset area threshold is as follows: in, This represents the minimum effective area after normalization. It is a rounding function; , These are the distance resolution and the azimuth resolution, respectively.

6. The method of claim 1, wherein, The screening threshold for the set of strong scattering points is PMR≥1.5; where PMR is the peak-to-mean ratio of the time-domain stability index.

7. The method of claim 1, wherein, The comprehensive scoring model satisfies: in, For the first A comprehensive score for each strong scattering point. For the first Temporal stability index of a strong scattering point For the first The amplitude value corresponding to the centroid position of the normalized amplitude map of each strong scattering point, with subscripts. Indicates the amplitude value, subscript Indicates normalization; and These are the weighting coefficients for the time-domain stability index value and the amplitude value, respectively. =0.5±0.

12.

8. A system for space-based ISAL hot spot selection incorporating multi-dimensional features, characterized by, include: The image amplitude calculation and normalization module is used to perform modulo operations on the input space-based ISAL complex matrix to generate an amplitude map and perform global normalization processing. The image preprocessing module is used to perform convolution calculation on the normalized amplitude map using a Gaussian kernel, and to process the convolution region that exceeds the image boundary using the pixel copying and extension method to obtain the filtered amplitude map. The amplitude threshold filtering module is used to generate a dynamic threshold based on the product of the scaling factor and the global maximum value, and to perform binarization processing on the filtered amplitude map to obtain the candidate point mask matrix. The connected component clustering module is used to perform connected component clustering analysis on the candidate point mask matrix and filter valid connected components based on a preset area threshold. The subpixel precision centroid estimation module is used to perform amplitude-weighted averaging calculations on the pixel coordinates within the filtered connected domains to obtain the subpixel-level coordinates of the centroid position of the strong scattering point. The temporal stability analysis module is used to extract the multi-frame amplitude sequence of each scattering point based on sub-pixel level coordinates, calculate its peak value to mean ratio as a temporal stability index, and screen the set of strong scattering points that meet the stability conditions. The weighted scoring and ranking selection module is used to construct a comprehensive scoring model by combining time-domain stability indicators, spatial domain amplitude values ​​and spatial distribution characteristics, and to select the preferred set of scattering points from the set of strong scattering points. Among them, the spatial amplitude value is the amplitude value corresponding to the centroid position of each strong scattering point in the normalized amplitude diagram, and the spatial distribution characteristics refer to the density distribution of each strong scattering point in two-dimensional space.

9. A storage medium, characterized by It stores a computer program, wherein the computer program causes a computer to perform the method for selecting strong points of space-based ISAL by incorporating multidimensional features as described in any one of claims 1 to 7.

10. An electronic device, comprising: The electronic device includes: Processor and memory; The memory stores program instructions; The processor is configured to run the program instructions to perform the method for selecting strong points of space-based ISAL combining multidimensional features as described in any one of claims 1 to 7.

Images