DBCA-CFAR detection method and system based on feature clustering

Abstract

The present application discloses a DBCA-CFAR detection method and system based on feature clustering, which mainly relates to the field of target detection technology and is used to solve the problem that existing solutions are easily interfered with in multi-target or non-uniform clutter environments, resulting in reduced detection performance. The method comprises: extracting the TEM features and amplitude features corresponding to each preset distance unit, and then constructing a two-dimensional feature vector; inputting the two-dimensional feature vector into the DBSCAN clustering algorithm to obtain target interference points and sea clutter points; setting the echo amplitude of the radar echo data of the preset distance unit corresponding to the target interference point to 0; determining the corresponding unit to be detected and the reference unit from the preset distance unit based on a sliding window; calculating the amplitude mean of the non-zero echo amplitudes of the reference unit corresponding to the unit to be detected, multiplying the amplitude mean by a preset normalization factor, and obtaining a threshold value; and determining whether the unit to be detected has a target based on the threshold value and the echo amplitude of the unit to be detected.

Description

Technical Field

[0001] The present application relates to the field of target detection technology, and in particular to a DBCA-CFAR detection method and system based on feature clustering. Background Art

[0002] Constant False-Alarm Rate (CFAR) detection technology is a core approach to radar target detection. Its core goal is to maintain a stable false alarm probability in complex environments. Traditional average-valued CFAR detectors (such as CA-CFAR Cell-Averaging Constant False Alarm Rate) estimate background noise by calculating the average power of reference cells using a sliding window, demonstrating excellent detection performance in uniform clutter environments.

[0003] However, when multiple strong targets exist within the reference window, traditional CA-CFAR can cause the energy of interfering targets to be incorporated into the background estimate, leading to an abnormally high detection threshold and a "masking effect," resulting in missed detection of truly weak targets. Although ordered statistical CFAR (OS-CFAR Ordered Statistic Constant False Alarm Rate) can partially alleviate this problem by sorting and eliminating outliers, its performance still degrades in high-density, multi-target scenarios. Furthermore, inhomogeneous environments such as sea clutter and meteorological clutter are time-varying and spatially correlated, and traditional methods' reliance on local statistical properties can lead to biased background estimation. For example, in clutter-edge regions, CA-CFAR may generate false or missed alarms due to the reference window spanning high and low clutter areas. Existing improved algorithms (such as GO / SO-CFAR) address clutter transition regions by selecting maximum or minimum reference subwindows, but this compromises detection probability. Summary of the Invention

[0004] The present application provides a DBCA-CFAR detection method and system based on feature clustering to solve the problem that existing solutions are susceptible to interference in multi-target or non-uniform clutter environments, resulting in decreased detection performance.

[0005] In a first aspect, the present application provides a DBCA-CFAR detection method based on feature clustering, the method comprising:

[0006] The TEM features and amplitude features corresponding to each preset distance unit are extracted from the radar echo data to construct a two-dimensional feature vector. The neighborhood radius and minimum number of points involved in the DBSCAN clustering algorithm are obtained, and the two-dimensional feature vector is then input into the DBSCAN clustering algorithm to obtain target interference points and sea clutter points. The echo amplitude of the radar echo data of the preset distance unit corresponding to the target interference point is set to 0. Based on a sliding window, the corresponding unit to be detected and the reference unit are determined from the preset distance units. The amplitude mean of the non-zero echo amplitudes of the reference unit corresponding to the unit to be detected is calculated, and the amplitude mean is multiplied by a preset normalization factor to obtain a threshold value. Based on the threshold value and the echo amplitude of the unit to be detected, it is determined whether the unit to be detected has a target.

[0007] In one implementation of the present application, extracting the TEM feature corresponding to each preset distance unit specifically includes:

[0008] By formula:

[0009] , obtain TEM characteristics;

[0010] Where N represents the length of the time domain signal of the radar echo data to be processed, L represents the length of the preset rectangular window, and N+L-1 represents the number of signals with a length of L. represents the time domain entropy value of the signal,

[0011] and ,

[0012] Indicates the signal amplitude i The normalized probability density of the interval, The first signal in the signal sequence representing the radar echo data i interval sample values.

[0013] In one implementation of the present application, based on a sliding window, determining the corresponding to-be-detected unit and the reference unit from the preset distance units specifically includes:

[0014] The window slides in the order of the preset distance units, with the unit to be detected as the center, and 1 protection unit and a preset number of units are set on both sides. reference units, 2N represents the total number of preset reference units;

[0015] When the number of protection units or reference units corresponding to the unit to be detected is insufficient, it is supplemented by zero padding or mirroring expansion.

[0016] In one implementation of the present application, calculating the mean amplitude of the non-zero echo amplitudes of the reference unit corresponding to the unit to be detected specifically includes:

[0017] Obtain the number P and Q of zero echo amplitudes corresponding to the reference units on the left and right sides of the unit to be detected;

[0018] pass -P, obtain the number of non-zero echo amplitudes on the left side, and then sum and average them to obtain X; where, represents the number of reference units on the left;

[0019] pass -Q, obtain the number of non-zero echo amplitudes on the right side, and then sum and average them to obtain Y; where, represents the number of reference units on the right;

[0020] By taking the average value of X+Y, the amplitude mean of the non-zero echo amplitude of the reference unit corresponding to the unit to be detected is obtained.

[0021] In one implementation of the present application, judging whether a target exists in the unit to be detected by using a threshold value and the echo amplitude of the unit to be detected specifically includes:

[0022] When the echo amplitude of the unit to be detected is greater than the threshold, it is determined that the target exists, otherwise it is clutter.

[0023] In one implementation of the present application, obtaining the neighborhood radius and minimum number of points involved in the DBSCAN clustering algorithm specifically includes:

[0024] Set the Neighborhood Radius to 150 and the Minimum Number of Points to 3.

[0025] In a second aspect, the present application provides a DBCA-CFAR detection system based on feature clustering, the system comprising:

[0026] The two-dimensional construction module is used to extract the TEM characteristics and amplitude characteristics corresponding to each preset distance unit from the radar echo data, and then construct a two-dimensional feature vector; the interference acquisition module is used to obtain the neighborhood radius and minimum number of points involved in the DBSCAN clustering algorithm, and then input the two-dimensional feature vector into the DBSCAN clustering algorithm to obtain target interference points and sea clutter points; the target detection module is used to set the echo amplitude of the radar echo data of the preset distance unit corresponding to the target interference point to 0; based on the sliding window, the corresponding unit to be detected and the reference unit are determined from the preset distance units; the amplitude mean of the non-zero echo amplitude of the reference unit corresponding to the unit to be detected is calculated, and the amplitude mean is multiplied by the preset normalization factor to obtain a threshold value; based on the threshold value and the echo amplitude of the unit to be detected, it is determined whether there is a target in the unit to be detected.

[0027] In one implementation of the present application, the two-dimensional building module includes a TEM feature calculation unit,

[0028] Used by the formula:

[0029] , obtain TEM characteristics;

[0030] Where N represents the length of the time domain signal of the radar echo data to be processed, L represents the length of the preset rectangular window, and N+L-1 represents the number of signals with a length of L. represents the time domain entropy value of the signal,

[0031] and ,

[0032] Indicates the signal amplitude i The normalized probability density of the interval, The first signal in the signal sequence representing the radar echo data i interval sample values.

[0033] In one implementation of the present application, the target detection module includes a window determination unit,

[0034] Used to slide the window in the order of preset distance units, with the unit to be detected as the center, and set 1 protection unit and preset number on both sides reference units, 2N represents the total number of preset reference units;

[0035] When the number of protection units or reference units corresponding to the unit to be detected is insufficient, it is supplemented by zero padding or mirroring expansion.

[0036] In one implementation of the present application, the target detection module includes a target determination unit,

[0037] When the echo amplitude of the unit to be detected is greater than the threshold value, it is determined that the target exists, otherwise it is clutter.

[0038] It can be seen from the above technical solutions that this application has the following advantages:

[0039] This application uses the DBSCAN clustering algorithm to jointly analyze TEM features and amplitude features, which can accurately identify and eliminate strong interfering target points within the reference window (set to amplitude 0). This fundamentally avoids the "target masking effect" caused by the inclusion of interfering target energy in background estimation in traditional CA-CFAR, and significantly improves the probability of weak target detection in high-density multi-target scenarios.

[0040] Multidimensional feature clustering can effectively distinguish sea clutter from real targets. DBSCAN, based on density clustering, can adaptively process clutter edge regions. Compared to the fixed subwindow selection strategy of GO / SO-CFAR, it suppresses false alarms while avoiding sacrificing detection probability.

[0041] Only the amplitude mean statistics are calculated for the clustered "non-interference points," reducing inefficient computation. Compared to the sorting operations (O(nlogn) complexity) of traditional OS-CFAR and the feature training overhead of machine learning CFAR, this method maintains accuracy while reducing real-time overhead.

[0042] In addition, this application jointly models TEM features (time-frequency energy modulation characteristics) and amplitude features for the first time, breaking through the limitation of traditional CFAR relying only on amplitude information, and providing a new dimension of judgment basis for target detection in complex electromagnetic environments.

[0043] In summary, the traditional CA-CFAR method relies on the statistical characteristics of the local reference window to estimate the background power, but is susceptible to interference in multi-target or non-uniform clutter environments, resulting in a decrease in detection performance. The DBCA-CFAR detector based on feature clustering in this application improves its robustness in complex scenarios by introducing data clustering analysis. This application first extracts features from the radar echo data, including amplitude and TEM; then, the data is divided into target class and background clutter class through the DBSCAN clustering algorithm; finally, based on the clustering results, pure background samples are selected to estimate the noise power, thereby optimizing the detection threshold. This solves the problem of increased false alarm rate caused by interference included in the reference window in multi-target scenarios in traditional methods, while reducing the impact of clutter non-uniformity. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for the description. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 This is a flow chart of a DBCA-CFAR detection method based on feature clustering provided in an embodiment of the present application.

[0046] Figure 2 This is a comparison diagram of the mean numerical distribution of time domain entropy values ​​of the target and clutter provided in the embodiment of the present application.

[0047] Figure 3 This is a comparison diagram of the histogram distribution of the mean time domain entropy values ​​of the target and clutter provided in the embodiment of the present application.

[0048] Figure 4 This is a schematic diagram of the #DATA1 clustering results provided in an embodiment of the present application.

[0049] Figure 5 This is a graph of the #DATA1 data before screening provided in an embodiment of the present application.

[0050] Figure 6 This is a data graph after #DATA1 screening provided in an embodiment of the present application.

[0051] Figure 7 This is a graph of the test results of the #DATA1 data before screening provided in an embodiment of the present application.

[0052] Figure 8 This is a graph of the data test results after #DATA1 screening provided in an embodiment of the present application.

[0053] Figure 9 This is a schematic diagram of the #DATA2 clustering results provided in an embodiment of the present application.

[0054] Figure 10 This is a graph of the #DATA2 data before screening provided in an embodiment of the present application.

[0055] Figure 11 This is a data graph after #DATA2 screening provided in an embodiment of the present application.

[0056] Figure 12 This is a graph of the test results of the data after #DATA2 screening provided in an embodiment of the present application.

[0057] Figure 13 This is a schematic diagram of the internal structure of a DBCA-CFAR detection system based on feature clustering provided in an embodiment of the present application. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0059] It should be understood by those skilled in the art that the embodiments described below are merely preferred embodiments of the present disclosure and do not imply that the present disclosure can only be implemented through these preferred embodiments. These preferred embodiments are merely intended to explain the technical principles of the present disclosure and are not intended to limit the scope of protection of the present disclosure. Based on the preferred embodiments provided by the present disclosure, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of the present disclosure.

[0060] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0061] The technical solutions proposed in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0062] The embodiment provides a DBCA-CFAR detection method based on feature clustering, such as Figure 1 As shown, the method provided in the embodiment of the present application mainly includes the following steps:

[0063] Step 110: extract the TEM feature and amplitude feature corresponding to each preset distance unit from the radar echo data, and then construct a two-dimensional feature vector.

[0064] In some embodiments, extracting the TEM feature corresponding to each preset distance unit specifically includes:

[0065] By formula:

[0066] , obtain TEM characteristics;

[0067] Where N represents the length of the time domain signal of the radar echo data to be processed, L represents the length of the preset rectangular window, and N+L-1 represents the number of signals with a length of L. represents the time domain entropy value of the signal,

[0068] and ,

[0069] Indicates the signal amplitude i The normalized probability density of the interval, The first signal in the signal sequence representing the radar echo data i interval sample values.

[0070] It should be noted that the TEM feature reflects the temporal variation characteristics of the signal, while the amplitude feature reflects the strength of the signal. The combination of the two can more comprehensively describe the characteristics of the target and sea clutter.

[0071] Step 120: Obtain the neighborhood radius and minimum number of points involved in the DBSCAN clustering algorithm, and then input the two-dimensional feature vector into the DBSCAN clustering algorithm to obtain target interference points and sea clutter points; set the echo amplitude of the radar echo data of the preset distance unit corresponding to the target interference point to 0.

[0072] It should be noted that in the mean-value constant false alarm detection, the interference target and sea spikes far exceed the power level of sea clutter. If they are not removed from the reference unit, the reference window will be seriously affected when averaging the power level of background clutter. Therefore, accurately distinguishing interference from sea clutter is a key issue. Figure 2 As shown in the figure, it can be found that the TEM eigenvalues ​​of sea clutter are mainly concentrated between 0.01 and 0.15, and the target TEM characteristic sequence fluctuates around 0.05. The TEM eigenvalues ​​of sea clutter units are all greater than the target TEM eigenvalues. Figure 3 As shown in the figure, the histogram distribution shows the ability of the time domain entropy mean to distinguish between the target and the clutter. It can be seen that the TEM characteristics of the target signal and the TEM characteristics of the sea clutter signal are somewhat distinguishable.

[0073] The DBSCAN clustering algorithm requires setting two key parameters: the neighborhood radius (Eps) and the minimum number of points (MinPts). Through multiple tests on measured data, we determined that when the neighborhood radius (Eps) is set to 150 and the minimum number of points (MinPts) is set to 3, the clustering results can accurately distinguish suspected targets from interference and clutter.

[0074] The two-dimensional feature vector constructed in step 110 is input into the DBSCAN algorithm for clustering. Based on the density relationship between the data points, the data points are divided into different clusters, and noise points are identified. During the clustering process, dense and adjacent data points are assigned to the same cluster, while isolated data points are marked as noise points. Through clustering, suspected target interference points and sea clutter points are initially distinguished.

[0075] In the clustering results, the amplitudes of the points marked as outliers are set to 0. These outliers may be caused by interference, noise or other factors. Setting their amplitudes to zero can effectively reduce the impact on the constant false alarm detection threshold.

[0076] Step 130: Based on the sliding window, determine the corresponding to-be-detected unit and the reference unit from the preset distance units.

[0077] This step can be specifically as follows:

[0078] The window slides in the order of the preset distance units, with the unit to be detected as the center, and 1 protection unit and a preset number of units are set on both sides. reference units, 2N represents the total number of preset reference units;

[0079] When the number of protection units or reference units corresponding to the unit to be detected is insufficient, it is supplemented by zero padding or mirroring expansion.

[0080] Step 140: Count the mean amplitude of the non-zero echo amplitudes of the reference unit corresponding to the unit to be detected, multiply the mean amplitude by a preset normalization factor to obtain a threshold value; and determine whether there is a target in the unit to be detected based on the threshold value and the echo amplitude of the unit to be detected.

[0081] In some embodiments, calculating the mean amplitude of the non-zero echo amplitudes of the reference unit corresponding to the unit to be detected specifically includes:

[0082] Obtain the number P and Q of zero echo amplitudes corresponding to the reference units on the left and right sides of the unit to be detected;

[0083] pass -P, obtain the number of non-zero echo amplitudes on the left side, and then sum and average them to obtain X; where, represents the number of reference units on the left;

[0084] pass -Q, obtain the number of non-zero echo amplitudes on the right side, and then sum and average them to obtain Y; where, represents the number of reference units on the right;

[0085] By taking the average value of X+Y, the amplitude mean of the non-zero echo amplitude of the reference unit corresponding to the unit to be detected is obtained.

[0086] In some embodiments, judging whether a target exists in a unit to be detected by using a threshold value and an echo amplitude of the unit to be detected specifically includes:

[0087] When the echo amplitude of the unit to be detected is greater than the threshold, it is determined that the target exists, otherwise it is clutter.

[0088] As described above, this embodiment integrates the DBSCAN clustering algorithm with the CA-CFAR algorithm. This integration is primarily reflected in the DBSCAN algorithm's ability to effectively handle data point density distribution, which is highly compatible with the diverse sea clutter characteristics and target distribution in complex ocean environments. Given the diverse noise interference sources in ocean environments, DBSCAN's ability to identify noise points enables the CA-CFAR algorithm to remove interfering noise from radar echoes, thereby mitigating its adverse effects on the algorithm's detection results and improving detection reliability.

[0089] Based on the above description, as a specific embodiment:

[0090] To further verify the performance of DBCA-CFAR detection, the measured data used was X-band radar sea clutter data published in the 2022 Journal of Radar. The radar operates in the 9.3-9.5 GHz frequency range, with an HH antenna polarization. During the test, the antenna was in staring mode. In this example, 20221113180037_stare_HH.mat was selected as #DATA1 and 20221114020008_stare_HH.mat was selected as #DATA2 to verify the algorithm's performance. Both data consist of 950 range bins and 131,072 pulse bins. #DATA1 records target 1 at range bins 502-514, target 2 at range bins 667-677, and sea state level 4. #DATA2 records targets 1, 2, and 3 at range bins 502-514, 613-624, and 667-677, respectively, with sea state level 3.

[0091] DBCA-CFAR is used to process echo data #DATA1, setting the reference unit N to 64, the protection units to 8 on each side, and the constant false alarm rate Pfa to 10-4. First, #DATA1 is preprocessed to extract the time domain entropy mean and amplitude characteristics of each distance unit, construct a two-dimensional feature vector, set the neighborhood radius (Eps) to 150, the minimum number of points (MinPts) to 3, and clustered using the DBSCAN clustering algorithm. The results are shown below. Figure 4 As shown. Figure 4 As can be seen, the scattered points are divided into two categories: clutter and targets. Red scattered points represent interfering target clusters, while blue scattered points represent clutter clusters. This is because target signals exhibit localized high-density clustering (red areas) in the time-domain entropy-amplitude joint feature space, while clutter forms a low-density discrete point set (blue areas) due to its random distribution. Its core advantage lies in the absence of prior assumptions (such as clutter distribution model and target number) and its ability to dynamically adapt to complex environments, overcoming the limitations of traditional CFAR, which relies on manual experience-based parameter tuning. The range bins corresponding to the selected interfering target features are then removed. Figure 5 and Figure 6 By comparison, it can be seen that the range cell amplitudes where the target and a small number of high-power points are located are eliminated to a certain extent. Figure 7 and Figure 8 It can be seen that the DBCA-CFAR detection threshold, obtained by estimating the local background noise power using only samples within the clutter cluster (blue dots), successfully detects two targets, while CA-CFAR misses the target on the right. This is because DBCA-CFAR avoids the "threshold lift" caused by the target signal mixing into the reference unit.

[0092] Same as the above steps, the clustering results of #DATA2 are as follows Figure 9As shown in Figure 2, since the target signal amplitude is significantly higher than the clutter background, TEM is around [0.09:0.14], and the two are significantly separable in the joint feature space. Figure 10 and Figure 11 By comparing the data before and after filtering, the high-amplitude interference points are effectively filtered out, and the amplitude distribution of the retained clutter units is more uniform. Finally, the DBCA-CFAR dynamic detection threshold is generated based on the local background power calculated based on the pure clutter, as shown in Figure 12 As shown in the figure, the algorithm successfully detected three targets, while traditional CA-CFAR detected Target 2 as two targets, resulting in a false detection. This may be because the larger size of Target 2 causes the radar scattering intensity of the target to vary at different range units, resulting in a certain attenuation of the target amplitude. The presence of Target 2 and Target 3 significantly increases the CA-CFAR threshold, resulting in the detection of Target 2 but the fragmentation of Target 2.

[0093] The improved algorithm's performance in complex maritime environments is primarily due to the effective data processing of the DBSCAN clustering algorithm. DBSCAN clustering accurately identifies the density distribution areas of sea clutter and targets, providing the CA-CFAR algorithm with more accurate background noise estimation and adaptive detection threshold adjustment. However, the improved algorithm also has certain limitations. For example, in extremely harsh sea conditions, the highly nonlinear and complex sea clutter can still affect its performance. Furthermore, the algorithm's computational complexity is increased compared to traditional CA-CFAR algorithms. In scenarios with extremely high real-time requirements, further optimization of the algorithm implementation may be necessary to increase processing speed.

[0094] In summary, the traditional CA-CFAR method relies on the statistical properties of a local reference window to estimate background power. However, it is susceptible to interference in multi-target or non-uniform clutter environments, resulting in degraded detection performance. The feature-based DBCA-CFAR detection scheme improves robustness in complex scenarios by introducing data clustering analysis. First, features are extracted from radar echo data, including amplitude and TEM. Subsequently, the DBSCAN clustering algorithm is used to partition the data into target and background clutter classes. Finally, based on the clustering results, pure background samples are selected to estimate noise power, thereby optimizing the detection threshold. This method effectively addresses the increased false alarm rate caused by interference within the reference window in multi-target scenarios, while also mitigating the impact of clutter heterogeneity. Validation with simulation and field data demonstrates that the improved algorithm improves detection probability in multi-target environments and at the clutter edge. Furthermore, it can detect targets that CA-CFAR cannot detect using field data from sea observations, effectively addressing target obscuration without sacrificing detection performance, demonstrating the effectiveness of DBCA-CFAR.

[0095] In addition, this application Figure 13 The embodiment of the present application provides a DBCA-CFAR detection system based on feature clustering. Figure 13 As shown, the system provided in the embodiment of the present application mainly includes:

[0096] The two-dimensional construction module 210 is used to extract the TEM feature and amplitude feature corresponding to each preset distance unit from the radar echo data, and then construct a two-dimensional feature vector.

[0097] The two-dimensional construction module 210 includes a TEM feature calculation unit,

[0098] Used by the formula:

[0099] , obtain TEM characteristics;

[0100] Where N represents the length of the time domain signal of the radar echo data to be processed, L represents the length of the preset rectangular window, and N+L-1 represents the number of signals with a length of L. represents the time domain entropy value of the signal,

[0101] and ,

[0102] Indicates the signal amplitude i The normalized probability density of the interval, The first signal in the signal sequence representing the radar echo data i interval sample values.

[0103] The interference acquisition module 220 is used to obtain the neighborhood radius and minimum number of points involved in the DBSCAN clustering algorithm, and then input the two-dimensional feature vector into the DBSCAN clustering algorithm to obtain target interference points and sea clutter points.

[0104] The target detection module 230 is used to set the echo amplitude of the radar echo data of the preset distance unit corresponding to the target interference point to 0; based on the sliding window, determine the corresponding unit to be detected and the reference unit from the preset distance unit; calculate the amplitude mean of the non-zero echo amplitude of the reference unit corresponding to the unit to be detected, multiply the amplitude mean by the preset normalization factor to obtain a threshold value; and determine whether there is a target in the unit to be detected based on the threshold value and the echo amplitude of the unit to be detected.

[0105] The target detection module 230 includes a window determination unit,

[0106] Used to slide the window in the order of preset distance units, with the unit to be detected as the center, and set 1 protection unit and preset number on both sides reference units, 2N represents the total number of preset reference units;

[0107] When the number of protection units or reference units corresponding to the unit to be detected is insufficient, it is supplemented by zero padding or mirroring expansion.

[0108] The target detection module 230 includes a target determination unit,

[0109] When the echo amplitude of the unit to be detected is greater than the threshold value, it is determined that the target exists, otherwise it is clutter.

[0110] In addition, an embodiment of the present application further provides a non-volatile computer storage medium having executable instructions stored thereon. When the executable instructions are executed, the above-mentioned DBCA-CFAR detection method based on feature clustering is implemented.

[0111] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A DBCA-CFAR detection method based on feature clustering, characterized in that: The method comprises: Extract the TEM features and amplitude features corresponding to each preset distance unit from the radar echo data, and then construct a two-dimensional feature vector; Obtain the neighborhood radius and minimum number of points involved in the DBSCAN clustering algorithm, and then input the two-dimensional feature vector into the DBSCAN clustering algorithm to obtain the target interference points and sea clutter points; Set the echo amplitude of the radar echo data of the preset distance unit corresponding to the target interference point to 0; Based on the sliding window, the corresponding unit to be detected and the reference unit are determined from the preset distance units; The average amplitude of the non-zero echo amplitude of the reference unit corresponding to the unit to be detected is calculated, and the average amplitude is multiplied by a preset normalization factor to obtain a threshold value; based on the threshold value and the echo amplitude of the unit to be detected, it is determined whether there is a target in the unit to be detected.

2. The DBCA-CFAR detection method based on feature clustering according to claim 1, characterized in that: Extract the TEM features corresponding to each preset distance unit, specifically including: By formula: , obtain TEM characteristics; Where N represents the length of the time domain signal of the radar echo data to be processed, L represents the length of the preset rectangular window, and N+L-1 represents the number of signals with a length of L. represents the time domain entropy value of the signal, and , represents the normalized probability density of the i-th interval of the signal amplitude, The first signal in the signal sequence representing the radar echo data i interval sample values.

3. The DBCA-CFAR detection method based on feature clustering according to claim 1, characterized in that: Based on the sliding window, the corresponding to-be-detected unit and the reference unit are determined from the preset distance units, specifically including: The window slides in the order of the preset distance units, with the unit to be detected as the center, and 1 protection unit and a preset number of units are set on both sides. Reference units, 2 Indicates the total number of preset reference units; When the number of protection units or reference units corresponding to the unit to be detected is insufficient, it is supplemented by zero padding or mirroring expansion.

4. The DBCA-CFAR detection method based on feature clustering according to claim 1, characterized in that: Count the mean amplitude of the non-zero echo amplitude of the reference unit corresponding to the unit to be detected, specifically including: Obtain the number P and Q of zero echo amplitudes corresponding to the reference units on the left and right sides of the unit to be detected; pass -P, obtain the number of non-zero echo amplitudes on the left side, and then sum and average them to obtain X; where, represents the number of reference units on the left; pass -Q, obtain the number of non-zero echo amplitudes on the right side, and then sum and average them to obtain Y; where, represents the number of reference units on the right; By taking the average value of X+Y, the amplitude mean of the non-zero echo amplitude of the reference unit corresponding to the unit to be detected is obtained.

5. The DBCA-CFAR detection method based on feature clustering according to claim 1, characterized in that: The threshold value and the echo amplitude of the unit to be detected are used to determine whether there is a target in the unit to be detected. Specifically, the following steps are performed: When the echo amplitude of the unit to be detected is greater than the threshold, it is determined that the target exists, otherwise it is clutter.

6. The DBCA-CFAR detection method based on feature clustering according to claim 1, characterized in that: Get the neighborhood radius and minimum number of points involved in the DBSCAN clustering algorithm, including: Set the Neighborhood Radius to 150 and the Minimum Number of Points to 3.

7. A DBCA-CFAR detection system based on feature clustering, characterized in that: The system comprises: A two-dimensional construction module is used to extract the TEM characteristics and amplitude characteristics corresponding to each preset distance unit from the radar echo data, and then construct a two-dimensional feature vector; The interference acquisition module is used to obtain the neighborhood radius and minimum number of points involved in the DBSCAN clustering algorithm, and then input the two-dimensional feature vector into the DBSCAN clustering algorithm to obtain target interference points and sea clutter points; The target detection module is used to set the echo amplitude of the radar echo data of the preset distance unit corresponding to the target interference point to 0; based on the sliding window, determine the corresponding unit to be detected and the reference unit from the preset distance unit; calculate the amplitude mean of the non-zero echo amplitude of the reference unit corresponding to the unit to be detected, multiply the amplitude mean by a preset normalization factor to obtain a threshold value; and determine whether there is a target in the unit to be detected based on the threshold value and the echo amplitude of the unit to be detected.

8. The DBCA-CFAR detection system based on feature clustering according to claim 7, characterized in that: The two-dimensional building block includes a TEM feature calculation unit, Used by the formula: , obtain TEM characteristics; Where N represents the length of the time domain signal of the radar echo data to be processed, L represents the length of the preset rectangular window, and N+L-1 represents the number of signals with a length of L. represents the time domain entropy value of the signal, and , Indicates the signal amplitude i The normalized probability density of the interval, The first signal in the signal sequence representing the radar echo data i interval sample values.

9. The DBCA-CFAR detection system based on feature clustering according to claim 7, characterized in that: The target detection module includes a window determination unit, Used to slide the window in the order of preset distance units, with the unit to be detected as the center, and set 1 protection unit and preset number on both sides Reference units, 2 Indicates the total number of preset reference units; When the number of protection units or reference units corresponding to the unit to be detected is insufficient, it is supplemented by zero padding or mirroring expansion.

10. The DBCA-CFAR detection system based on feature clustering according to claim 7, characterized in that: The target detection module includes a target determination unit, When the echo amplitude of the unit to be detected is greater than the threshold value, it is determined that the target exists, otherwise it is clutter.

Images