A two-dimensional adaptive constant false alarm rate detection method of grid complementary reference sliding window

Abstract

The application belongs to the technical field of target detection, and specifically discloses a two-dimensional adaptive constant false alarm rate detection method of a grid complementary reference sliding window, which comprises the following steps: constructing a grid complementary reference sliding window under a detection matrix and distributing reference units; initializing key parameters of the detection matrix, and converting an original image signal matrix into a signal power matrix; judging and assigning the coordinate features of the reference units based on a window type topological structure, and calculating the mean value after traversal to extract background clutter; calculating a second order statistic VI and a mean ratio MR, and calculating a detection threshold by means of three threshold factor calculation formulas and in combination with a judgment rule to select a constant false alarm rate algorithm. The two-dimensional adaptive constant false alarm rate detection method of the grid complementary reference sliding window has the advantages that the detection accuracy is improved by 49.4% under a multi-target and clutter interference environment, the threshold stability and boundary processing capacity are better under different backgrounds, the false alarm and missed alarm rates are effectively reduced, and the target detection performance is improved.

Description

Technical Field

[0001] This invention relates to the field of target detection technology, and in particular to a two-dimensional adaptive constant false alarm rate (CFAR) detection method using a gridded complementary reference sliding window. Background Technology

[0002] In the field of target detection, the Constant False Alarm Rate (CFAR) algorithm plays a crucial role in sonar, radar, and other detection systems. As a thresholding technique, CFAR dynamically adjusts the detection threshold by estimating background clutter power in real time, thereby ensuring a constant false alarm probability under complex and changing environmental conditions. The application of this algorithm not only significantly enhances the target recognition capability of detection systems under noise and clutter interference but also provides a solid guarantee for accurate target localization and tracking, playing an irreplaceable and vital role in various fields such as civilian surveillance. Due to the limitations of 2D image selection window types, most current methods for target detection are based on 2D cross-shaped windows, star-shaped windows, grid windows, and rectangular windows combined with unit-averaged constant false alarm rate (CFAR) detectors. Cross-shaped and star-shaped windows offer the advantage of rapid target detection, but their low sampling rate results in the loss of significant effective information and an inability to stably estimate background noise power levels, leading to poor target detection performance. CFAR detection methods based on rectangular and grid windows offer significant performance improvements over the first two, but these latter two window types are only found in traditional CFAR algorithms, with limited application examples. Grid-type reference windows offer better real-time data processing and robustness to complex environments than rectangular windows, and their deviation values ​​are also better, resulting in greater stability against uniform clutter backgrounds. In recent years, an adaptive CFAR algorithm (VI-CFAR) has become widely used. It integrates the characteristics of previous CFAR algorithms, adaptively selecting different CFAR processing algorithms to achieve better performance in uniform backgrounds, multi-target backgrounds, and clutter edge backgrounds. However, this adaptive constant false alarm rate (VI-CFAR) algorithm requires a reasonable sliding window configuration for both leading and trailing edges in the context of two-dimensional images to complete subsequent algorithmic tasks. Previously, VI-CFAR was applied to one-dimensional signal processing because setting the left and right windows is more convenient and computationally efficient in one dimension. Furthermore, if the window type is not chosen appropriately, it can lead to deviations in the target detection results, easily resulting in false alarms and false negatives, causing significant changes in the target detection probability.

[0003] The existing technology has the following defects and shortcomings:

[0004] (1) Fixed window type makes it difficult to balance computational efficiency and detection accuracy;

[0005] (2) The VI-CFAR algorithm lacks optimization in sliding window settings in two-dimensional images, resulting in inaccurate background noise estimation;

[0006] (3) Lacking an adaptive mechanism, it is impossible to dynamically adjust the detection strategy according to different backgrounds (uniform background, multi-target background and clutter interference background, etc.), which affects the stability and reliability of target detection. Summary of the Invention

[0007] The purpose of this invention is to provide a two-dimensional adaptive constant false alarm detection method with a gridded complementary reference sliding window, which can improve its detection performance under different backgrounds, such as uniform background, multi-target background and clutter interference background, and also reduce the possibility of false alarms and missed alarms.

[0008] To achieve the above objectives, this invention provides a two-dimensional adaptive constant false alarm rate (CFAR) detection method using a gridded complementary reference sliding window, comprising the following steps:

[0009] S1. Construct a gridded complementary reference sliding window under the detection matrix and assign reference elements, including guard elements, left complementary window reference elements, right complementary window reference elements and detection elements (CUT);

[0010] S2. Initialize the key parameters of the detection matrix and convert the original image signal matrix into a signal power matrix using a power conversion algorithm;

[0011] S3. Based on the window topology of complementary reference sliding window, the coordinate characteristics of the left complementary window reference cell and the right complementary window reference cell are identified and assigned values. After traversing and identifying the assigned values, the mean value is calculated to extract background clutter.

[0012] S4. Calculate the second-order statistics VI and mean ratio MR of the left complementary window reference cell and the right complementary window reference cell. Calculate the detection threshold using the constant false alarm rate algorithm selected by combining the three threshold factor calculation formulas and the judgment rules.

[0013] Preferably, S1 includes the following steps:

[0014] S11. Set the reference sliding window to a regular window type in rectangular form;

[0015] S12. Extract the detection unit CUT and the protection unit separately to prevent target energy leakage;

[0016] S13. The complementary state of the grid window is marked with different colors. Color coding technology is used to distinguish the spatial distribution and signal power characteristics of the protection unit, the left complementary window reference unit, the right complementary window reference unit and the detection unit CUT. The functional classification of each color unit is marked with a color code legend, including the spatial distribution relationship of the left complementary window reference unit, the right complementary window reference unit and the detection unit CUT.

[0017] Preferably, in S2, the key parameters include the structural parameters of the detection matrix, the detection performance parameters, and the statistical number of left complementary window reference units. Statistical count of right complementary window reference units ;

[0018] The structural parameters of the detection matrix include the spatial proportion distribution of the guard cells, the left complementary window reference cells, and the right complementary window reference cells, as well as the topological configuration matrix of the left and right complementary windows; the detection performance parameters include the false alarm rate Pfa, the second-order statistic threshold Kvi, and the mean threshold Kmr.

[0019] Preferably, in S3, the coordinate positions of the left complementary reference unit in the detection matrix satisfy the characteristic that the horizontal and vertical coordinates are either both odd or both even, that is:

[0020] ;

[0021] The coordinate positions of the right complementary reference unit in the detection matrix satisfy the mathematical relationship that the horizontal and vertical coordinates are mutually exclusive, i.e.:

[0022] ;

[0023] in, This indicates that the coordinate value is odd. This indicates that the coordinate value is even. The x-coordinate represents the reference cell of the left complementary window. The vertical coordinate representing the reference cell of the left complementary window. The x-coordinate represents the reference cell of the right complementary window. The ordinate of the right complementary window reference cell.

[0024] Preferably, in S3, a traversal processing method is used to perform cell-level analysis on the detection matrix, including the following steps:

[0025] S31. Based on the preset window topology, perform coordinate feature discrimination on each unit:

[0026] If the cell position coordinates satisfy the spatial distribution characteristics of the left complementary window reference cell, then the power value of that cell is assigned as the left complementary window reference matrix. The corresponding element is incremented, and the left window reference cell counter is incremented; otherwise, the right complementary window reference matrix is ​​executed. The corresponding assignment and counting operations;

[0027] S32. After completing the traversal, refer to the left complementary window matrix. And right complementary window reference matrix The unassigned regions are initialized with zero values ​​to eliminate numerical errors caused by undefined variables in subsequent statistical calculations, and finally the matrix values ​​of the left complementary window reference cell and the right complementary window reference cell are obtained.

[0028] S33. Based on the signals in the power matrix, sum the left complementary window reference cell matrix, the right complementary window reference cell matrix, and the corresponding detected values, and record the result as follows: Then divide by the statistical number of left complementary window reference units and right complementary window reference units. The mean values ​​of the left complementary window reference cell matrix and the right complementary window reference cell matrix are obtained. .

[0029] Preferably, in S4, each cell of each window is traversed. If the cell belongs to the left complementary window reference cell, the second-order statistic of the left window is calculated using the following formula:

[0030] ;

[0031] In the formula, Represents the values ​​of each cell in the left complementary window;

[0032] If the cell belongs to the right complementary window reference cell, the second-order statistic of the right window is calculated using the following formula:

[0033] ;

[0034] In the formula, Represents the values ​​of each cell in the right complementary window;

[0035] The mean values ​​of the left complementary window reference cell and the right complementary window reference cell are as follows:

[0036] .

[0037] Preferably, in S4, there are three forms for calculating the threshold factors of the left complementary reference window and the right complementary reference window, including methods that only require the number of reference cells in the left complementary reference window, methods that only require the number of reference cells in the right complementary reference window, and methods that require the total number of reference cells in both windows. The formulas are as follows:

[0038] ;

[0039] ;

[0040] ;

[0041] Based on the second-order statistics of the left complementary window reference unit and the right complementary window reference unit Compared to the mean, MR will be used in three scenarios: uniform background, multi-target background, and clutter edge background. The MR is compared with the second-order statistical threshold Kvi and the mean threshold Kmr to determine whether the background clutter is uniform and whether the mean ratio MR is the same. Based on the judgment results, different constant false alarm rate (CFAR) algorithms are used to calculate the threshold required for different detection units.

[0042] Preferably, the rules for determining whether background clutter is uniform and whether the mean ratio (MR) is the same are as follows:

[0043] ;

[0044] ;

[0045] .

[0046] Therefore, the present invention employs the above-mentioned two-dimensional adaptive constant false alarm rate (CFAR) detection method using a gridded complementary reference sliding window, and the beneficial effects are as follows:

[0047] (1) This invention uses a gridded complementary reference sliding window to replace the traditional sliding window. Through specific reference unit distribution rules and calculation methods, the target detection probability is effectively improved. Compared with the traditional adaptive constant false alarm algorithm (VI-CFAR), the detection probability is significantly improved in multi-target and clutter interference environments, reducing the possibility of false alarms and missed alarms, and achieving more accurate target recognition.

[0048] (2) The present invention has a dynamic adaptive mechanism, which can automatically select the appropriate traditional constant false alarm algorithm to calculate the detection threshold according to different environmental characteristics such as uniform background, multi-target background and clutter edge background, so as to ensure that the false alarm probability of the system remains constant under complex and ever-changing background conditions, effectively cope with various interferences, and stably and reliably complete the target detection task.

[0049] (3) The present invention performs well in threshold calculation and processing. The gridded complementary sliding window makes the reference unit distribution more uniform and reduces the impact of interference on adaptive threshold calculation. Under different signal-to-noise ratio conditions, the threshold change is more sensitive and stable, and the boundary processing capability is stronger. It can effectively avoid false alarms or missed detections caused by background clutter interference and improve the overall performance of the detection system.

[0050] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the distribution of gridded complementary reference sliding window cells under the detection matrix in an embodiment of the two-dimensional adaptive constant false alarm rate detection method of gridded complementary reference sliding window according to the present invention;

[0052] Figure 2This is a window diagram obtained by assigning values ​​to the left and right complementary windows of a two-dimensional adaptive constant false alarm rate detection method using a gridded complementary reference sliding window according to an embodiment of the present invention.

[0053] Figure 3 This is an overall flowchart of an embodiment of a two-dimensional adaptive constant false alarm rate (CFAR) detection method using a gridded complementary reference sliding window according to the present invention.

[0054] Figure 4 This is a target detection trajectory implemented according to the improved adaptive constant false alarm rate algorithm in an embodiment of the two-dimensional adaptive constant false alarm rate detection method of the gridded complementary reference sliding window of the present invention.

[0055] Figure 5 This invention presents a comparison of detection probabilities between two methods with different signal-to-noise ratios under uniform conditions using a gridded complementary reference sliding window for two-dimensional adaptive constant false alarm rate detection.

[0056] Figure 6 This invention presents a comparison of detection probabilities between two methods with different signal-to-noise ratios under uniform conditions using a gridded complementary reference sliding window for two-dimensional adaptive constant false alarm rate detection.

[0057] Figure 7 This invention provides a comparison of the detection probabilities of two methods under different signal-to-noise ratios in a multi-target environment under condition (1) of a two-dimensional adaptive constant false alarm rate detection method using a gridded complementary reference sliding window.

[0058] Figure 8 This is a comparison of the detection probabilities of two methods with different signal-to-noise ratios under conditions (2) in a multi-target environment of the present invention, which is an embodiment of a two-dimensional adaptive constant false alarm detection method using a gridded complementary reference sliding window;

[0059] Figure 9 This invention provides a comparison of the detection probabilities of two methods with different signal-to-noise ratios under conditions (3) in a multi-target environment, based on an embodiment of a two-dimensional adaptive constant false alarm detection method using a gridded complementary reference sliding window.

[0060] Figure 10 This invention presents a comparison of the detection probabilities of two methods with different signal-to-noise ratios under clutter interference environments using a two-dimensional adaptive constant false alarm rate detection method with a gridded complementary reference sliding window.

[0061] Figure 11 This is a three-dimensional visualization and top view of the threshold simulation results of the traditional and improved adaptive constant false alarm rate (CFAR) algorithm in a uniform environment according to an embodiment of the two-dimensional adaptive CFAR detection method with a gridded complementary reference sliding window of the present invention. (a) is a simulation comparison of the detection threshold in a uniform environment (3D visualization), and (b) is a simulation comparison of the detection threshold in a uniform environment (top view).

[0062] Figure 12This is a side view of the threshold simulation results of the traditional and improved adaptive constant false alarm rate (CFAR) algorithm under a uniform environment according to an embodiment of the two-dimensional adaptive CFAR detection method with a gridded complementary reference sliding window of the present invention. (a) is a comparison of the threshold simulation results in the x-axis direction under a uniform environment, and (b) is a comparison of the threshold simulation results in the y-axis direction under a uniform environment.

[0063] Figure 13 This is a three-dimensional visualization and top view of the threshold simulation results of the traditional and improved adaptive constant false alarm rate (CFAR) algorithm in a multi-target environment, based on an embodiment of the two-dimensional adaptive CFAR detection method using a gridded complementary reference sliding window according to the present invention. (a) is a simulation comparison of the detection threshold in a multi-target environment (3D visualization), and (b) is a simulation comparison of the detection threshold in a multi-target environment (top view).

[0064] Figure 14 This is a side view of the threshold simulation results of the traditional and improved adaptive constant false alarm rate (CFAR) algorithm in a multi-target environment according to an embodiment of the two-dimensional adaptive CFAR detection method with a gridded complementary reference sliding window of the present invention. (a) is a comparison of the threshold simulation results in the x-axis direction in a multi-target environment, and (b) is a comparison of the threshold simulation results in the y-axis direction in a multi-target environment.

[0065] Figure 15 This is a three-dimensional visualization and top view of the threshold simulation results of the traditional and improved adaptive constant false alarm rate (CFAR) algorithm in a clutter edge environment according to an embodiment of the two-dimensional adaptive CFAR detection method with a gridded complementary reference sliding window of the present invention. (a) is a simulation comparison of the detection threshold in the clutter edge environment (3D visualization), and (b) is a simulation comparison of the detection threshold in the clutter edge environment (top view).

[0066] Figure 16 This is a side view of the threshold simulation results of the traditional and improved adaptive constant false alarm rate (CFAR) algorithm in a clutter edge environment according to an embodiment of the two-dimensional adaptive CFAR detection method with a gridded complementary reference sliding window of the present invention. (a) is a comparison of the threshold simulation results in the x-axis direction in the clutter edge environment, and (b) is a comparison of the threshold simulation results in the y-axis direction in the clutter edge environment. Detailed Implementation

[0067] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0068] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0069] This invention discloses a two-dimensional adaptive constant false alarm rate (VI-CFAR) detection method (hereinafter referred to as VIGB-CFAR) using a meshed complementary front and rear edge reference sliding window, comprising the following steps:

[0070] S1, such as Figure 1 As shown, a gridded complementary reference sliding window under the detection matrix is ​​constructed and reference cells are assigned, including a guard cell, a left complementary window reference cell, a right complementary window reference cell, and a detection cell CUT, including the following steps:

[0071] S11. Set the reference sliding window to a regular rectangular window type to optimize the adaptability of the reference unit allocation and ensure its compatibility with the construction requirements of subsequent mesh complementary windows.

[0072] S12. The detection unit CUT and the protection unit are extracted separately. The signals in these two units are not calculated. The purpose of extraction is to prevent the target energy leakage from interfering with the background noise estimation and thus improve the robustness of the detection threshold.

[0073] S13. Following the grid window method, use different colors to mark the complementary states of the grid windows. Figure 1 The spatial distribution and signal power characteristics of the protection unit, left complementary window reference unit, right complementary window reference unit and detection unit CUT are distinguished by color coding technology on the left.

[0074] Figure 1 The right side uses color charts to indicate the functional categories of each color unit, including the spatial distribution of the left complementary window reference unit, the right complementary window reference unit, and the detection unit CUT.

[0075] The distribution of the gridded complementary reference sliding window cells has now been obtained. The main steps of the implementation process of the two-dimensional adaptive constant false alarm rate (CFAR) detection algorithm based on the gridded complementary reference sliding window cells in this invention are as follows:

[0076] S2. In the implementation of the adaptive constant false alarm rate (VI-CFAR) algorithm, the key parameters of the detection matrix need to be initialized in advance. The key parameters include the structural parameters of the detection matrix, the detection performance parameters, and the statistical number of left complementary window reference cells. Statistical count of right complementary window reference units .

[0077] The structural parameters of the detection matrix include the spatial proportions of the guard cells, left complementary window reference cells, and right complementary window reference cells, as well as the topological configuration matrices of the left and right complementary windows. Detection performance parameters include the false alarm rate (Pfa), the second-order statistic threshold (Kvi), and the mean threshold (Kmr). The original image signal matrix is ​​transformed into a signal power matrix using a power conversion algorithm. This initialization process strictly adheres to the Neyman-Pearson criterion, ensuring that an optimal detection benchmark is established before algorithm execution, providing a reliable parameter basis for subsequent adaptive detection.

[0078] S3, based on Figure 1 The window topology of the complementary reference sliding window is shown. The reference cells of the left and right complementary reference sliding windows have specific spatial distribution patterns. The coordinate characteristics of the reference cells of the left and right complementary windows are identified and assigned values. After traversing and identifying the assigned values, the mean value is calculated to extract background clutter.

[0079] like Figure 1 The coordinate position of the left complementary reference cell No. 1 in the detection matrix satisfies the property that the horizontal and vertical coordinates are either both odd or both even, that is:

[0080] ;

[0081] Figure 1 The coordinates of the second right complementary reference unit in the detection matrix satisfy the mathematical relationship that the horizontal and vertical coordinates are mutually exclusive, i.e.:

[0082] ;

[0083] in, This indicates that the coordinate value is odd. This indicates that the coordinate value is even. The x-coordinate represents the reference cell of the left complementary window. The vertical coordinate representing the reference cell of the left complementary window. The x-coordinate represents the reference cell of the right complementary window. The ordinate of the right complementary window reference cell.

[0084] This geometric constraint ensures the symmetry and uniformity of the reference cells in the gridded complementary distribution, providing a structured sampling basis for subsequent statistical estimation of background noise.

[0085] In the algorithm implementation, a traversal processing method is used to perform cell-level analysis on the detection matrix, including the following steps:

[0086] S31. Based on the preset window topology, perform coordinate feature discrimination on each unit:

[0087] If the cell position coordinates satisfy the spatial distribution characteristics of the left complementary window reference cell, then the power value of that cell is assigned as the left complementary window reference matrix. The corresponding element is incremented, and the left window reference cell counter is incremented; otherwise, the right complementary window reference matrix is ​​executed. The corresponding assignment and counting operations.

[0088] S32. After completing the traversal, refer to the left complementary window matrix. And right complementary window reference matrix The unassigned regions in the two reference matrices are initialized with zero values. This operation aims to eliminate numerical errors caused by undefined variables in subsequent statistical calculations, ensure the mathematical rigor of background noise estimation, and finally obtain the matrix values ​​of the left complementary window reference cell and the right complementary window reference cell.

[0089] S33. Based on the signals in the power matrix, sum the left complementary window reference cell matrix, the right complementary window reference cell matrix, and the corresponding detected values, and record the result as follows: Then divide by the statistical number of left complementary window reference units and right complementary window reference units. The mean values ​​of the left complementary window reference cell matrix and the right complementary window reference cell matrix are obtained. .

[0090] The purpose of this is to facilitate the subsequent calculation of the second-order statistic VI and mean ratio MR in the adaptive constant false alarm rate (VI-CFAR). The left and right complementary window plots obtained after successive assignments of the reference sliding window are shown below. Figure 2 As shown.

[0091] S4. Calculate the second-order statistics VI and mean ratio MR of the left complementary window reference cell and the right complementary window reference cell. Calculate the detection threshold using the constant false alarm rate algorithm selected by combining the three threshold factor calculation formulas and the judgment rules.

[0092] Calculating the second-order statistics VI and mean ratio MR of the left and right complementary window reference cells requires the reference cell matrix of each of the left and right complementary reference sliding windows from step S3. The number of reference cells in the left complementary reference sliding window and the right complementary reference sliding window. And the mean values ​​of the reference cells of the left complementary reference sliding window and the right complementary reference sliding window. .

[0093] Iterate through each cell of each window. If the cell belongs to the reference cell of the left complementary window, calculate the second-order statistic of the left window using the following formula:

[0094] ;

[0095] In the formula, Represents the values ​​of each cell in the left complementary window;

[0096] If the cell belongs to the right complementary window reference cell, the second-order statistic of the right window is calculated using the following formula:

[0097] ;

[0098] In the formula, Represents the values ​​of each cell in the right complementary window;

[0099] The mean values ​​of the left complementary window reference cell and the right complementary window reference cell are as follows:

[0100] .

[0101] There are three forms for calculating the threshold factors of the left complementary reference sliding window and the right complementary reference sliding window: methods that only require the number of reference cells in the left complementary reference window, methods that only require the number of reference cells in the right complementary reference window, and methods that require the total number of reference cells in both windows. The formulas are as follows:

[0102] ;

[0103] ;

[0104] ;

[0105] Based on the second-order statistics of the left complementary window reference unit and the right complementary window reference unit Compared to the mean, MR will be used in three scenarios: uniform background, multi-target background, and clutter edge background. The mean ratios MR and MR are compared with the second-order statistic threshold Kvi and the mean threshold Kmr, respectively, to determine whether the background clutter is uniform and whether the mean ratios MR are the same. The judgment rules are as follows:

[0106] ;

[0107] ;

[0108] .

[0109] Based on the judgment results, different traditional constant false alarm rate (CFAR) algorithms were used to calculate the thresholds required for different detection units, resulting in the adaptive CFAR threshold selection table shown in Table 1.

[0110] Table 1 Summary of Adaptive Threshold Selection

[0111]

[0112] The adaptive threshold calculation formula is as follows:

[0113] ①: When the left complementary reference window is uniform, the right complementary reference window is uniform, and the ratio of the means of the left and right complementary reference windows is the same, the cell average constant false alarm rate (CA) algorithm is used, and the adaptive threshold is calculated as follows:

[0114] ;

[0115] The power values ​​of all reference units within the left and right complementary reference sliding windows are averaged, and the detection threshold is calculated by combining the preset false alarm rate and other parameters.

[0116] ②: When the left complementary reference window and the right complementary reference window are both uniform, but the mean ratios of the left and right complementary reference windows are different, the maximum selection constant false alarm rate (GO) algorithm is used, and the adaptive threshold is calculated as follows:

[0117] ;

[0118] That is, the maximum sum of the power of the reference cells in the left and right complementary reference sliding windows is selected to estimate the background clutter power, and then the detection threshold is calculated.

[0119] ③: When the left complementary reference sliding window is non-uniform and the right complementary reference sliding window is uniform, the cell-averaged constant false alarm rate (CA) algorithm is used, and the adaptive threshold is calculated as follows:

[0120] ;

[0121] The detection threshold is calculated based on the average power of the reference cells within the right complementary reference sliding window, combined with relevant parameters.

[0122] ④: When the left complementary reference sliding window is uniform and the right complementary reference sliding window is non-uniform, the cell average constant false alarm rate (CA) algorithm is used, and the adaptive threshold is calculated as follows:

[0123] ;

[0124] The detection threshold is calculated based on the average power of the reference cells within the left complementary reference sliding window, combined with relevant parameters.

[0125] ⑤: When the left complementary reference sliding window is non-uniform and the right complementary reference sliding window is non-uniform, the minimum selection constant false alarm (SO) algorithm is adopted, and the adaptive threshold is calculated as follows:

[0126] ;

[0127] The minimum sum of the reference cell powers in the left and right complementary reference sliding windows is selected to estimate the background clutter power, thereby calculating the detection threshold.

[0128] Based on the above, the flowchart of the two-dimensional adaptive constant false alarm rate (CFAR) detection algorithm based on gridded complementary reference sliding window elements is as follows: Figure 3 As shown.

[0129] To implement the two-dimensional adaptive constant false alarm rate (CFAR) detection algorithm using a gridded complementary reference sliding window (CFR) cell, the main steps are as follows: The algorithm is applied to a set of 12 sonar images to perform target detection and output the results. An overall picture of the target detection is provided. Since each frame of the sonar image contains points that appear to match the target power, these points are marked. The results are as follows: Figure 4 As shown in the results, the accuracy of the detection results of this invention is improved by 49.4% compared with the traditional adaptive constant false alarm rate (VI-CFAR).

[0130] To determine the effectiveness of this method, based on the evaluation metrics of constant false alarm rate (CFAR) algorithm data, a comparison is presented between the improved form complementary reference sliding window (CFR) two-dimensional adaptive CFAR algorithm and the traditional adaptive CFAR algorithm (VI-CFAR) on different evaluation metrics. The results are as follows:

[0131] This paper examines the characteristics of detection probability (Pd) under different signal-to-noise ratios, and compares the detection thresholds of the improved complementary reference sliding window (CISL) 2D adaptive constant false alarm rate (CFAR) algorithm with the traditional adaptive CFAR algorithm (VI-CFAR) at different distance cells in uniform, multi-target, and clutter edge environments. The former assesses the superiority of different algorithms based on the condition that higher detection probabilities lead to better detector performance; the latter compares the threshold simulation diagrams (2D and 3D simulations) in uniform, multi-target, and clutter edge environments to determine the relative merits of different algorithms.

[0132] To calculate the detection probability (Pd) under different signal-to-noise ratios, the initial conditions are set as follows: the signal matrix dimension is 60. 60, Set the target random position (i,j)( This generates a random signal superimposed with clutter, resulting in an analog signal with a false alarm rate Pfa=10. -5Kvi=4.76, Kmr=1.806, the signal-to-noise ratio of the target under test is 1~35dB cyclic, and Monte Carlo simulation is used for 10... 5 The simulation yielded results for the traditional VI-CFAR adaptive constant false alarm rate algorithm and the two-dimensional adaptive VI-CFAR detection method using a gridded left and right reference sliding window. Simulations were performed in a uniform environment based on the aforementioned preconditions, realizing the relationship curves between the detection probability (Pd) and the signal-to-noise ratio under different conditions, as shown in the figure. Figure 5 and Figure 6 As shown.

[0133] Under uniform environmental conditions, the detection probabilities of the traditional adaptive constant false alarm rate algorithm (VI-CFAR) and the improved constant false alarm rate algorithm (VIGB-CFAR) are roughly the same. However, the detection probability of the improved algorithm is 1% higher than that of the traditional one. The change is not significant because, under uniform conditions, the second-order statistics of the two algorithms are roughly the same and are calculated based on the entire signal matrix. This results in the same mean ratio between the left and right windows, eliminating the need for separate left and right windows.

[0134] In order to ensure the validity and generality of the experimental results in a multi-target environment, the present invention sets the following centralized conditions for each experiment: (1) adding two interferences to the left complementary window reference unit and the left reference window of the traditional adaptive constant alarm at the set target position; (2) adding one interference to each of the left and right complementary window reference units and the left and right reference windows of the traditional adaptive constant alarm at the set target position; (3) adding two interferences to each of the left and right complementary window reference units and the left and right reference windows of the traditional adaptive constant alarm at the set target position. Through the above three cases, the relationship curves between the signal-to-noise ratio and the detection probability (Pd) under different signal-to-noise ratio conditions are realized as follows. Figures 7-9 As shown.

[0135] As shown in the figure, under different multi-target environments, the detection probabilities of the two algorithms change with the signal-to-noise ratio. The implementation of the detection probabilities of the traditional adaptive constant false alarm rate algorithm (VI-CFAR) and the improved constant false alarm rate algorithm (VIGB-CFAR) results in the improved algorithm having a better detection probability than the traditional adaptive constant false alarm rate algorithm (VI-CFAR). Especially under high signal-to-noise ratio conditions, although the detection probabilities of the two algorithms change, the overall change and result of the latter detection probability is better than that of the former. This indicates that the improved constant false alarm rate algorithm (VIGB-CFAR) has better target detection performance.

[0136] In addition to relying on the aforementioned prerequisites, the clutter interference environment processing also determined the case where the power ratio of strong to weak clutter is 2, and realized the relationship curves between the signal-to-noise ratio and the detection probability (Pd) under different conditions, as shown in the figure. Figure 10As shown in the figure, in a multi-target environment, the detection probabilities of the two algorithms change with the signal-to-noise ratio. The improved detection probability implementation of the traditional adaptive constant false alarm rate algorithm (VI-CFAR) and the improved constant false alarm rate algorithm (VIGB-CFAR) results in a better detection probability than the traditional adaptive constant false alarm rate algorithm (VI-CFAR), especially when the target is at the edge of clutter. The latter's detection probability change and result are better than the former, indicating that the improved constant false alarm rate algorithm (VIGB-CFAR) performs better in target detection under low signal-to-noise ratio conditions.

[0137] To address the detection thresholds of the improved form complementary reference sliding window (CFAR) 2D adaptive constant false alarm rate (CFAR) algorithm and the traditional adaptive CFAR algorithm (VI-CFAR) in uniform, multi-target, and clutter edge environments with different range cells, the initial conditions are set as follows: the signal matrix dimension is 100. 80. The basic target position in both uniform and multi-target environments is set to (60, 30). The protection unit is set with a long side of 2 and a short side of 2. The reference window has a long side of 4 and a short side of 3. The low-noise signal power is a Gaussian noise signal with a mean of 5dB and a variance of 1. In the uniform environment, a single target is superimposed with the noise to obtain a simulated signal. The target is located at (60, 30) in a clutter edge environment, with a strong-to-weak noise power ratio of 3. To visually observe the detection results, the target peak signal-to-noise ratio is set to 5, and the false alarm rate Pfa = 10. -5 With Kvi=4.76 and Kmr=1.806, a comparison of relevant thresholds was obtained between the traditional VI-CFAR adaptive constant false alarm rate algorithm and the two-dimensional adaptive VI-CFAR detection method with gridded left and right reference sliding windows.

[0138] The comparison of relevant thresholds under uniform conditions yields simulation results of the algorithm thresholds for the signal, the traditional VI-CFAR algorithm, and the two-dimensional adaptive constant false alarm rate (CFAR) detection method with gridded left and right reference sliding windows under uniform conditions. The results include a 3D visualization of the results, a top view, and threshold result plots along the x and y axes corresponding to the signal coordinates. (See figure). Figure 11 , Figure 12 As shown.

[0139] Based on the above results, the following conclusions can be drawn: (1) Based on this situation, it can be concluded that the detection thresholds of the two are generally similar. (2) For boundary problems, the traditional VI-CFAR algorithm and the improved VI-CFAR algorithm have better boundary processing capabilities. (3) For the top view, it can be seen that the red area is mostly the gridded reference window with a higher threshold value, which can cover the conventional VI-CFAR algorithm.

[0140] For threshold detection in a multi-target environment, to ensure experimental rigor, the original single target was positioned at (60, 30), and two additional targets were added at (65, 30) and (60, 25) respectively, with signal-to-noise ratios of 3 and 8. A comparison of relevant thresholds in a multi-target environment was conducted, and the simulation results of the algorithm thresholds for signal detection, the traditional VI-CFAR algorithm, and the two-dimensional adaptive CFAR detection method with a gridded left and right reference sliding window were obtained. The results include a 3D visualization of the results, a top view, and threshold result plots along the x and y axes corresponding to the signal coordinates. The results are shown in the figure below. Figure 13 , Figure 14 As shown.

[0141] Based on the above results, the following conclusions can be drawn: (1) Based on the previous detection probability, the threshold value of the gridded method is higher than that of the conventional algorithm in terms of the threshold of the target boundary on the x-axis and y-axis. (2) In terms of the stability of the threshold around the target, the gridded threshold is more stable than the conventional one. For boundary problems, the traditional VI-CFAR algorithm and the improved VI-CFAR algorithm have better boundary processing capabilities. (3) For the top view, it can be seen that the red area is mostly the gridded reference window with a higher threshold value, which can cover the conventional VI-CFAR algorithm.

[0142] In clutter edge environments, noise power values ​​transition from low to high. For constant false alarm rate (CFAR) algorithms, the rationality and feasibility of the algorithm are evaluated by statistically analyzing the rate of change of the threshold in areas with significant clutter power. At clutter edges, the faster the threshold changes, the less likely false alarms will occur when a target is present at the edge. A comparison of relevant thresholds in clutter edge environments is presented, resulting in simulation results of the algorithm thresholds for signal detection in multi-target environments, the traditional VI-CFAR algorithm, and the two-dimensional adaptive CFAR detection method with a gridded left and right reference sliding window. The results include a 3D visualization of the results, a top view, and threshold results along the x and y axes corresponding to the signal coordinates. (See figure). Figure 15 , Figure 16 As shown.

[0143] Based on the results in the figure below, the following conclusions can be drawn: (1) Based on the previous detection probability, the threshold value of the gridded method is higher than that of the conventional algorithm in terms of the threshold of the target boundary on the x-axis and y-axis. (2) In terms of the stability of the threshold around the target, the gridded threshold is more stable than the conventional one. The reason is that the threshold value of the gridded method changes faster than that of the conventional method when the threshold obtained from the noise signal changes from low power to high power. This proves that the gridded threshold will rise rapidly and sensitively, and it is not easy to cause false alarms. (3) For boundary problems, the traditional VI-CFAR algorithm and the improved VI-CFAR algorithm have better boundary processing capabilities. (4) For the top view, it can be seen that the red area at the edge of the clutter is mostly the gridded reference window with a higher threshold value, which can cover the conventional VI-CFAR algorithm.

[0144] Therefore, the present invention employs the above-mentioned two-dimensional adaptive constant false alarm rate (CFAR) detection method using a gridded complementary reference sliding window. In uniform environments, multi-target environments, and clutter edge environments, the two-dimensional adaptive CFAR detection method using a gridded complementary reference sliding window (VIGB-CFAR) shows significant improvement over the traditional adaptive CFAR algorithm (VI-CFAR) in terms of threshold variation. Under different signal-to-noise ratio conditions, it is not easily affected by background clutter near the detection unit, and the adaptive decision threshold will not cause false alarms or missed detections due to noise interference. Furthermore, the detection probability (Pd) is significantly improved compared to previous algorithms.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A two-dimensional adaptive constant false alarm rate (CFAR) detection method using a gridded complementary reference sliding window, characterized in that, Includes the following steps: S1. Construct a gridded complementary reference sliding window under the detection matrix and assign reference elements, including guard elements, left complementary window reference elements, right complementary window reference elements and detection elements (CUT); S2. Initialize the key parameters of the detection matrix and convert the original image signal matrix into a signal power matrix using a power conversion algorithm; S3. Based on the window topology of complementary reference sliding window, the coordinate characteristics of the left complementary window reference cell and the right complementary window reference cell are identified and assigned values. After traversing and identifying the assigned values, the mean value is calculated to extract background clutter. S4. Calculate the second-order statistics VI and mean ratio MR of the left complementary window reference cell and the right complementary window reference cell. Calculate the detection threshold using the constant false alarm rate algorithm selected by the three threshold factor calculation formulas and the judgment rules. In S3, the coordinate positions of the left complementary reference unit in the detection matrix satisfy the property that the horizontal and vertical coordinates are either both odd or both even, that is: ； The coordinate positions of the right complementary reference unit in the detection matrix satisfy the mathematical relationship that the horizontal and vertical coordinates are mutually exclusive, i.e.: ； in, This indicates that the coordinate value is odd. This indicates that the coordinate value is even. The x-coordinate represents the reference cell of the left complementary window. The vertical coordinate representing the reference cell of the left complementary window. The x-coordinate represents the reference cell of the right complementary window. The ordinate of the right complementary window reference cell; S3 employs a traversal processing method to perform cell-level analysis of the detection matrix, including the following steps: S31. Based on the preset window topology, perform coordinate feature discrimination on each unit: If the cell position coordinates satisfy the spatial distribution characteristics of the left complementary window reference cell, then the power value of that cell is assigned as the left complementary window reference matrix. The corresponding element is incremented, and the left window reference cell counter is incremented; otherwise, the right complementary window reference matrix is ​​executed. The corresponding assignment and counting operations; S32. After completing the traversal, refer to the left complementary window matrix. And right complementary window reference matrix The unassigned regions are initialized with zero values ​​to eliminate numerical errors caused by undefined variables in subsequent statistical calculations, and finally the matrix values ​​of the left complementary window reference cell and the right complementary window reference cell are obtained. S33. Based on the signals in the power matrix, sum the left complementary window reference cell matrix, the right complementary window reference cell matrix, and the corresponding detected values, and record the result as follows: Then divide by the statistical number of left complementary window reference units and right complementary window reference units. The mean values ​​of the left complementary window reference cell matrix and the right complementary window reference cell matrix are obtained. .

2. The two-dimensional adaptive constant false alarm rate (CFAR) detection method using a gridded complementary reference sliding window according to claim 1, characterized in that, S1 includes the following steps: S11. Set the reference sliding window to a regular window type in rectangular form; S12. Extract the detection unit CUT and the protection unit separately to prevent target energy leakage; S13. The complementary state of the grid window is marked with different colors. Color coding technology is used to distinguish the spatial distribution and signal power characteristics of the protection unit, the left complementary window reference unit, the right complementary window reference unit and the detection unit CUT. The functional classification of each color unit is marked with a color code legend, including the spatial distribution relationship of the left complementary window reference unit, the right complementary window reference unit and the detection unit CUT.

3. The two-dimensional adaptive constant false alarm rate (CFAR) detection method using a gridded complementary reference sliding window according to claim 2, characterized in that, In S2, key parameters include the structural parameters of the detection matrix, the detection performance parameters, and the statistical number of left complementary window reference units. Statistical count of right complementary window reference units ; The structural parameters of the detection matrix include the spatial proportion distribution of the guard cells, the left complementary window reference cells, and the right complementary window reference cells, as well as the topological configuration matrix of the left and right complementary windows; the detection performance parameters include the false alarm rate Pfa, the second-order statistic threshold Kvi, and the mean threshold Kmr.

4. The two-dimensional adaptive constant false alarm rate (CFAR) detection method using a gridded complementary reference sliding window according to claim 3, characterized in that, In S4, each cell of each window is traversed. If the cell belongs to the left complementary window reference cell, the second-order statistic of the left window is calculated using the following formula: ； In the formula, Represents the values ​​of each cell in the left complementary window; If the cell belongs to the right complementary window reference cell, the second-order statistic of the right window is calculated using the following formula: ； In the formula, Represents the values ​​of each cell in the right complementary window; The mean values ​​of the left complementary window reference cell and the right complementary window reference cell are as follows: 。 5. The two-dimensional adaptive constant false alarm rate (CFAR) detection method using a gridded complementary reference sliding window according to claim 4, characterized in that, In S4, there are three forms for calculating the threshold factors of the left complementary reference window and the right complementary reference window: methods that only require the number of reference cells in the left complementary reference window, methods that only require the number of reference cells in the right complementary reference window, and methods that require the total number of reference cells in both windows. The formulas are as follows: ； ； ； Based on the second-order statistics of the left complementary window reference unit and the right complementary window reference unit Compared to the mean, MR will be used in three scenarios: uniform background, multi-target background, and clutter edge background. The MR is compared with the second-order statistical threshold Kvi and the mean threshold Kmr to determine whether the background clutter is uniform and whether the mean ratio MR is the same. Based on the judgment results, different constant false alarm rate (CFAR) algorithms are used to calculate the threshold required for different detection units.

6. The two-dimensional adaptive constant false alarm rate (CFAR) detection method using a gridded complementary reference sliding window according to claim 5, characterized in that, The rules for determining whether background clutter is uniform and whether the mean ratio (MR) is the same are as follows: ； ； 。

Images