Clutter region false alarm suppression method based on multi-feature fusion

Abstract

This invention discloses a clutter false alarm suppression method based on multi-feature fusion, comprising: acquiring radar echo signals and performing pulse compression processing to obtain time-domain echo data; performing clutter suppression and coherent accumulation on the time-domain echo data to obtain a range-Doppler matrix; performing constant false alarm detection on the range-Doppler matrix to obtain several potential target points; extracting the multi-pulse echo signals and unsuppressed frequency domain signals corresponding to the potential target points from the time-domain echo data, extracting the corresponding clutter-suppressed frequency domain signals from the range-Doppler matrix, and constructing multi-dimensional features for each potential target point based on the extracted signals; inputting the multi-dimensional features into a pre-trained classifier, determining whether the potential target point is a target or false alarm clutter based on the output results, and outputting the final target point trace. This invention effectively suppresses clutter false alarms in complex environments and improves radar detection reliability through multi-dimensional feature fusion and secondary decision-making by the classifier.

Description

Technical Field

[0001] This invention belongs to the field of radar signal processing and pattern recognition technology, specifically relating to a method for suppressing false alarms in clutter areas based on multi-feature fusion. Background Technology

[0002] With the widespread application of drones in urban environments, the demand for detecting "low, slow, and small" targets is becoming increasingly urgent. Detecting drone targets in complex urban clutter environments presents challenges such as strong clutter amplitude and abundant residual clutter, leading to a severe deterioration of the detection signal-to-noise ratio and becoming a technical bottleneck in the field of radar detection.

[0003] To achieve clutter suppression and target detection, modern radar systems generally employ a modular, progressive standard processing procedure, including: First, matching filtering of the transmitted wide-bandwidth signal using pulse compression technology to achieve signal focusing in the time domain, obtaining high range resolution and high signal-to-noise ratio; Second, filtering out stationary or slow-moving clutter using clutter suppression mechanisms such as moving target indicator cancellers or FIR filter banks; Next, coherent accumulation of multiple coherent pulses (usually achieved by moving target detection processing), coherently superimposing target energy in the frequency domain to further improve signal-to-noise ratio and velocity resolution; Finally, extracting target traces using a two-dimensional constant false alarm rate (CFAR) detection algorithm on the resulting range-Doppler two-dimensional data plane.

[0004] However, the above-mentioned existing technical solutions still have the following shortcomings in practical applications: First, clutter suppression relies on a fixed model, leading to environmental mismatch. Existing technologies for designing FIR filters for clutter suppression require clutter modeling. Traditional clutter modeling algorithms typically assume that different clutter backgrounds follow a specific prior distribution. However, for large and complex clutter scenarios, it is difficult to accurately describe the entire area using a single distribution model. Even within the same scenario, different times and weather conditions (especially the effects of wind, rain, and snow) can cause changes in the internal motion of clutter, resulting in clutter spectrum broadening or shifting, causing it to follow different probability distributions. Therefore, filters designed using fixed model methods suffer from a severe model mismatch with the actual clutter environment. This leads to a large amount of residual clutter, ultimately manifesting as an excessive number of false alarms detected by radar, severely impacting radar detection and tracking performance.

[0005] Second, the detection and decision-making process relies on a single dimension, making it difficult to distinguish between targets with similar characteristics and clutter. Under the combined effects of strong multipath effects from densely packed urban buildings, non-uniform specular reflection, and dynamic interference sources such as moving vehicles and swaying vegetation, clutter exhibits highly complex characteristics in both the spatial and frequency domains. This results in a high degree of overlap between the residual clutter components after clutter suppression and the characteristics of slow-moving UAV targets in the three-dimensional time-frequency-space feature space. Existing detection methods essentially still rely on the limited feature dimension of "range-Doppler-amplitude," and detectors that rely solely on a single feature plane (such as range-Doppler) cannot utilize richer information to distinguish between targets with similar amplitude and velocity characteristics and clutter, thus falling into a dilemma of high false alarm rates and high missed detections. The aforementioned model mismatch and the problem of a single feature dimension combine to directly lead to an excessive number of false alarms detected by the radar system, accompanied by the risk of missed detections, severely restricting the radar's detection and tracking performance in complex clutter environments. Summary of the Invention

[0006] To address the aforementioned problems in the existing technology, this invention provides a method for suppressing false alarms in clutter regions based on multi-feature fusion.

[0007] The technical problem to be solved by this invention is achieved through the following technical solution: This invention provides a method for suppressing false alarms in clutter regions based on multi-feature fusion, comprising: Acquire radar echo signals, perform pulse compression processing on the radar echo signals, and obtain pulse-compressed time-domain echo data; Clutter suppression and coherent accumulation are performed on the time-domain echo data to obtain the range-Doppler matrix; By performing constant false alarm rate (CFAR) detection on the range-Doppler matrix, several potential target points are obtained; For each potential target point, the corresponding multipulse echo signal and the unsuppressed frequency domain signal are extracted from the time-domain echo data, and the corresponding clutter-suppressed frequency domain signal is extracted from the range-Doppler matrix. Based on the multipulse echo signal, the clutter-suppressed frequency domain signal, and the unsuppressed frequency domain signal, multidimensional features of the potential target point are extracted. The multidimensional features include at least one of the following: target energy distribution features, spectral difference features before and after clutter suppression, and energy-phase composite features. The multidimensional features are input into a pre-trained classifier. Based on the output of the classifier, the potential target point is determined to be a target or a false alarm clutter, and the final target point trace is output.

[0008] Optionally, the target energy distribution characteristics are calculated in the following manner: ; in, This represents the frequency domain signal without clutter suppression. Indicates the Doppler cell index. Indicates the index of the Doppler cell where the potential target point is located. Indicates the total number of Doppler elements. express The Middle The amplitude of each Doppler unit, This indicates the percentage of the target energy.

[0009] Optionally, the spectral difference characteristics before and after clutter suppression are calculated in the following manner: ; ; in, Indicates the Doppler cell index. This represents the frequency domain signal after clutter suppression. Normalized information entropy represents the difference in signal energy before and after clutter suppression.

[0010] Optionally, the energy-phase composite feature includes: the product of the target-zero frequency energy ratio and the phase variance after clutter suppression, and the product of the target-zero frequency energy ratio and the phase variance before clutter suppression; The product of the target-zero frequency energy ratio after clutter suppression and the phase variance is calculated in the following manner: ; in, express The amplitude of clutter in the zero-frequency channel. express The amplitude of the Doppler cell containing the potential target. Indicates a multi-pulse echo signal. express The Middle The complex value of the echo of each pulse. express phase, Represents variance. This represents the product of the target-zero frequency energy ratio and the phase variance after clutter suppression. The product of the target-zero frequency energy ratio before clutter suppression and the phase variance is calculated as follows: ; in, express The amplitude of clutter in the zero-frequency channel. express The amplitude of the Doppler cell containing the potential target. This represents the product of the target-zero frequency energy ratio and the phase variance before clutter suppression.

[0011] Optionally, the classifier is trained in the following manner: Multiple potential target points are acquired, and the true category of the potential target points is identified using prior information; Extract the multidimensional features of the potential target points, and use the true category as the true category label corresponding to the multidimensional features to generate a training sample set; The classifier is trained using the training sample set to obtain a trained classifier.

[0012] Optionally, the prior information includes at least one of terrain data, target trajectory data, or manually calibrated data.

[0013] Optionally, the classifier is one of a support vector machine, a linear decision analysis classifier, a correlation vector machine, or a K-nearest neighbor classifier.

[0014] Optionally, the step of performing clutter suppression and coherent accumulation on the time-domain echo data to obtain the range-Doppler matrix includes: Based on clutter modeling, an FIR filter bank is designed and applied in parallel to the slow time-dimensional sequence of the time-domain echo data to obtain the range-Doppler matrix.

[0015] Optionally, the constant false alarm rate (CFAR) detection performed on the range-Doppler matrix yields several potential target points, including: For each unit to be detected in the distance-Doppler matrix, calculate the adaptive detection threshold corresponding to that unit; The signal amplitude of the unit to be detected is compared with the adaptive detection threshold. If the signal amplitude of the unit to be detected exceeds the adaptive detection threshold, the unit is identified as a potential target point.

[0016] This invention proposes a clutter region false alarm suppression method based on multi-feature fusion. Addressing the problem of numerous false alarms detected by existing technologies due to model adaptation and limited feature dimensions, this method further extracts multi-dimensional features from the time-domain and frequency-domain data corresponding to potential target points after constant false alarm detection. Specifically, this invention combines clutter suppression filters with expert knowledge information, designing and introducing target energy distribution characteristics, spectral differences before and after clutter suppression, and energy-phase composite features. These features characterize the essential differences between targets and clutter from multiple dimensions, including energy concentration, filter response differences, and phase stability, effectively amplifying the subtle differences between real targets and clutter false alarms. Based on the constructed multi-dimensional feature space, this invention inputs the extracted multi-dimensional features into a pre-trained classifier for secondary decision-making. Utilizing the classifier's pattern recognition capabilities, it learns and leverages the distribution patterns of targets and clutter in the high-dimensional feature space, thereby more accurately eliminating false alarms. Through the combination of multi-dimensional feature fusion and secondary decision-making by the classifier, this invention achieves effective suppression of clutter false alarms in complex environments, significantly improving the reliability of radar target detection.

[0017] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0018] Figure 1 This is a flowchart of a clutter region false alarm suppression method based on multi-feature fusion provided in an embodiment of the present invention; Figure 2 This is a flowchart of the training process of the clutter region false alarm suppression method based on multi-feature fusion provided in the embodiments of the present invention; Figure 3 This is a feature scattering map of the target and clutter in the clutter region false alarm suppression method based on multi-feature fusion provided in this embodiment of the invention; Figure 4 This is a detection result diagram of target detection using a traditional constant false alarm rate (CFAR) detection algorithm provided in an embodiment of the present invention; Figure 5 This is a detection result diagram of the target detection using the clutter region false alarm suppression method based on multi-feature fusion provided in the embodiment of the present invention. Detailed Implementation

[0019] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0020] To address the problem of existing methods detecting a large number of false alarms and having missed detections, this invention provides a method for suppressing false alarms in clutter regions based on multi-feature fusion. See also... Figure 1 The method includes the following steps: S10. Acquire radar echo signals, perform pulse compression processing on radar echo signals, and obtain pulse-compressed time-domain echo data.

[0021] Specifically, the radar system transmits a burst of pulse signals into the target airspace, while the radar receiving antenna captures the echo signals reflected from the UAV target and the environment. Within one coherent processing interval, assuming the radar transmits... Each pulse is sampled in the fast time dimension (distance dimension). If there are 10 points, the original echo matrix can be expressed as: , ; in, It is a length of The row vector represents the row vector of the first row. The pulse echo is sampled at different distance units. Each row in the array corresponds to a pulse repetition cycle (slow time), and each column corresponds to a specific distance unit (fast time).

[0022] Here, to achieve high range resolution and improve the signal-to-noise ratio, pulse compression is performed on each pulse echo. Pulse compression is essentially matched filtering, which involves convolving the received and transmitted signals with their conjugate time inversions. For linear frequency modulated (LFM) signals, pulse compression can compress wide pulse energy into narrow pulses, concentrating the energy within the range cell where the target is located. Specifically, for each pulse echo... Perform pulse compression processing to obtain the pulse-compressed echo signal. All Arrange the pulse compression results of each pulse in rows to obtain the pulse-compressed time-domain echo matrix. : .

[0023] Understandably, the dimensions of this time-domain echo matrix are the same as those of the original echo matrix, and it remains the same. .

[0024] S20. Perform clutter suppression and coherent accumulation on the time-domain echo data to obtain the range-Doppler matrix.

[0025] In this embodiment of the invention, clutter suppression and coherent accumulation of time-domain echo data to obtain the range-Doppler matrix includes: designing an FIR filter bank based on clutter modeling, and applying the FIR filter bank in parallel to the slow-time-dimensional sequence of the time-domain echo data to obtain the range-Doppler matrix. This step is specifically implemented in the following manner: First, clutter is modeled based on the prior clutter environment characteristics, and a set of finite impulse response filters with different center frequencies are designed to form an FIR filter bank. Each filter corresponds to a specific Doppler channel, and its frequency response is designed to be low gain or notch in the clutter frequency band (such as near zero frequency) to achieve clutter suppression, and high gain in the Doppler frequency band where the target may appear to achieve signal matching.

[0026] Secondly, the FIR filter bank is applied in parallel to the time-domain echo matrix. The slow-time sequence is used to obtain the range-Doppler matrix. Specifically, for each range unit, the length is... The multi-pulse sequence is processed simultaneously using all filters; the output of each filter is actually a weighted sum of the signals within the Doppler channel, equivalent to completing the coherent accumulation of the Doppler channel. Through this parallel processing, not only are stationary or slow clutters filtered out, but also the signals dispersed within the Doppler channel are filtered out. The target energy in each pulse is coherently superimposed onto the corresponding Doppler channel, thus simultaneously achieving clutter suppression and coherent accumulation in a single processing step. In this step, the filtering results of all range cells are arranged according to the range dimension and the Doppler dimension, ultimately outputting a two-dimensional range-Doppler matrix. : ; in, Indicates the index of the distance cell. The distance-Doppler matrix Each column in the matrix corresponds to the Doppler spectrum of a range unit, and each row corresponds to a specific Doppler channel (i.e., a velocity channel). This range-Doppler matrix... It fully characterizes the scattering intensity of different velocity components at each range cell within the detection airspace, providing a data foundation for subsequent constant false alarm rate (CFAR) detection.

[0027] S30. Perform constant false alarm detection on the range-Doppler matrix to obtain several potential target points.

[0028] This step aims to improve the distance-Doppler matrix generated in step S20. Constant false alarm rate (CFAR) detection is performed to initially screen potential target points from complex clutter backgrounds, which will then be used for subsequent feature extraction and classification. Here, CFAR detection is performed on the range-Doppler matrix to obtain several potential target points, specifically including: S301. For each cell to be detected in the distance-Doppler matrix, calculate the adaptive detection threshold corresponding to that cell.

[0029] Specifically, regarding the range-Doppler matrix For each unit to be detected, a two-dimensional sliding window structure is designed, which consists of three layers: the unit to be detected is located at the center, a ring of protective units is arranged around it, and the outermost layer is a reference unit. The function of the protective units is to prevent the target energy from leaking into the reference unit and affecting the accurate estimation of clutter power, while the reference unit is used to estimate the local clutter power level around the unit to be detected.

[0030] For each unit to be detected, the average signal amplitude of all reference units within the sliding window is calculated. This average value is the estimate of the local clutter power. Then, a preset threshold factor (determined by the desired false alarm probability) is multiplied by this average value to obtain the adaptive detection threshold corresponding to the unit to be detected.

[0031] S302. Compare the signal amplitude of the unit to be detected with the adaptive detection threshold. If the signal amplitude of the unit to be detected exceeds the adaptive detection threshold, then the unit to be detected is identified as a potential target point.

[0032] Specifically, the amplitude of the unit to be detected is compared with the adaptive detection threshold: if the amplitude of the unit to be detected exceeds the adaptive detection threshold, the unit to be detected is identified as a potential target point, and its distance cell index is recorded. And Doppler unit index If the signal amplitude does not exceed the threshold, it is judged as background noise or clutter and is not recorded.

[0033] S40. For each potential target point, extract the corresponding multipulse echo signal and the unsuppressed frequency domain signal from the time-domain echo data, and extract the corresponding clutter-suppressed frequency domain signal from the range-Doppler matrix; based on the multipulse echo signal, the clutter-suppressed frequency domain signal, and the unsuppressed frequency domain signal, extract the multidimensional features of the potential target point; the multidimensional features include at least one of the target energy distribution features, the spectral difference features before and after clutter suppression, and the energy-phase composite features.

[0034] Specifically, for each potential target point obtained in step S30, its index is determined according to the distance cell it belongs to. And Doppler unit index Perform the following operations: S401, Signal Data Extraction: (1) Time-domain echo data obtained from step S10 Extract the first one from the middle. Multi-pulse echo signal corresponding to each distance unit The multi-pulse echo signal It is a length of The slow time series reflects the amplitude and phase changes between pulses in the distance unit where the target is located.

[0035] (2) For multi-pulse echo signals Performing a direct Fourier transform yields the frequency domain signal without clutter suppression. ; (3) The distance-Doppler matrix obtained from step S20 Extract the first one from the middle. Doppler spectrum corresponding to each distance unit , as the frequency domain signal after clutter suppression.

[0036] S402, Multidimensional Feature Construction: Specifically, based on the above three types of signals: multi-pulse echo signals Frequency domain signals without clutter suppression Frequency domain signal after clutter suppression By incorporating expert knowledge from clutter suppression filter design (such as the zero-frequency channel corresponding to stationary clutter and the phase stability of the target Doppler channel), a multidimensional feature vector is constructed for each potential target point. In this embodiment of the invention, the multidimensional features include at least one of the following: target energy distribution features, spectral difference features before and after clutter suppression, and energy-phase composite features.

[0037] In this embodiment of the invention, the target energy distribution characteristics are: The proportion of potential target energy is calculated as follows: ; in, This represents the frequency domain signal without clutter suppression. Indicates the Doppler cell index. Indicates the index of the Doppler cell where the potential target point is located. Indicates the total number of Doppler elements. express The Middle The amplitude of each Doppler unit, This indicates the percentage of the target energy.

[0038] The spectral differences before and after clutter suppression are the same as those of the frequency domain signal without clutter suppression. Frequency domain signal after clutter suppression The entropy value of the difference is calculated as follows: ; ; in, Indicates the Doppler cell index. This represents the frequency domain signal after clutter suppression. Normalized information entropy represents the difference in signal energy before and after clutter suppression.

[0039] In this embodiment of the invention, the energy-phase composite feature includes: the product of the target-zero frequency energy ratio and the phase variance after clutter suppression, and the product of the target-zero frequency energy ratio and the phase variance before clutter suppression. Wherein, the product of the target-zero frequency energy ratio and the phase variance after clutter suppression is... The ratio of the potential target energy value to the zero-frequency channel clutter energy value, multiplied by the phase variance, is calculated as follows: ; in, express The amplitude of clutter in the zero-frequency channel. express The amplitude of the Doppler cell containing the potential target. Indicates a multi-pulse echo signal. express The Middle The complex value of the echo of each pulse. express phase, Represents variance. This represents the product of the target-zero frequency energy ratio and the phase variance after clutter suppression. The product of the target-zero frequency energy ratio and the phase variance before clutter suppression is: The ratio of the potential target energy value to the zero-frequency channel clutter energy value, multiplied by the phase variance, is calculated as follows: ; in, express The amplitude of clutter in the zero-frequency channel. express The amplitude of the Doppler cell containing the potential target. This represents the product of the target-zero frequency energy ratio and the phase variance before clutter suppression.

[0040] By concatenating the above features in a fixed order, a multidimensional feature vector for each potential target point can be obtained. .

[0041] For example, in this embodiment of the invention, four features are extracted and concatenated in the following form to obtain a multidimensional feature vector. .

[0042] S50. Input the multidimensional features into the pre-trained classifier, determine whether the potential target point is a target or a false alarm clutter based on the output of the classifier, and output the final target point trace.

[0043] This step aims to use a pre-trained classifier to make a decision on the multidimensional features extracted in step S40, so as to distinguish between real targets and false alarm clutter from potential target points and output the final target point trace.

[0044] Specifically, for each potential target point to be decided, its multidimensional feature vector extracted in step S40 is... The input is fed into a pre-trained classifier, and the category of the potential target point is determined based on the classifier's output.

[0045] In this embodiment, the classifier uses a support vector machine, and its decision process is as follows: First, the multidimensional feature vector Substitute the values ​​into the trained support vector machine decision function and calculate the output value: ; in, For the first The support vectors obtained from training For the corresponding first Individual weight coefficients The number of support vectors obtained during training. This is the output of the support vector machine.

[0046] Next, based on the output of the support vector machine Determine the category of the potential target point: if If so, then the potential target point is determined to be the target point; if If the condition is not met, the potential target point is determined to be a false alarm clutter point. All potential target points identified as target points are recorded, and their corresponding range cell index, Doppler cell index, and signal amplitude information are output to form the final target point trace.

[0047] It should be noted that although this embodiment uses support vector machine as an example, in actual applications the classifier can be one of support vector machine, linear decision analysis classifier, correlation vector machine or K nearest neighbor classifier. Its classification process is similar to that of support vector machine, and will not be described in detail here.

[0048] It is understandable that the classifier needs to be trained before use. In embodiments of the present invention, such as Figure 2 As shown, the classifier is trained in the following way: (a) Acquire multiple potential target points and classify the potential target points using prior information; exemplarily, the prior information includes at least one of terrain data, target trajectory data or manually calibrated data.

[0049] Specifically, during the training phase, radar echo data is first acquired and processed in the same manner as steps S10 to S30 to obtain several potential target points. Then, prior information is used to determine the true category of each potential target point, distinguishing it as a target point or a clutter point.

[0050] (b) Extract the multidimensional features of potential target points, use the true category as the true category label corresponding to the multidimensional features, and generate a training sample set.

[0051] Specifically, following the method in step S40, multidimensional features are extracted for each potential target point, and the labeled true category is used as the ground value category label corresponding to the multidimensional feature, thereby forming a training sample set. In this embodiment of the invention, the extracted multidimensional features include target energy distribution features, spectral difference features before and after clutter suppression, and energy-phase composite features.

[0052] (c) The classifier is trained using the training sample set to obtain the trained classifier.

[0053] Specifically, the classifier is trained in a supervised manner using the constructed training sample set, adjusting the classifier parameters until the model converges, resulting in a trained classifier. This training process can be completed offline before system deployment or updated online based on new samples during system operation to ensure the classifier's adaptability to complex environmental changes.

[0054] Similarly, the testing phase process for this classifier is as follows: Figure 2 As shown, the same method can be used to construct a test sample set and input it into the trained classifier to obtain the target point trace.

[0055] Understandably, through the processing in step S50, a secondary screening of the constant false alarm rate (CFAR) detection results is finally achieved, effectively eliminating false alarm points caused by clutter residue and retaining the true target points, thereby improving the target detection reliability and tracking performance of the radar system in complex clutter environments.

[0056] This invention proposes a clutter region false alarm suppression method based on multi-feature fusion. Addressing the problem of numerous false alarms detected by existing technologies due to model adaptation and limited feature dimensions, this method further extracts multi-dimensional features from the time-domain and frequency-domain data corresponding to potential target points after constant false alarm detection. Specifically, this invention combines clutter suppression filters with expert knowledge information, designing and introducing target energy distribution characteristics, spectral differences before and after clutter suppression, and energy-phase composite features. These features characterize the essential differences between targets and clutter from multiple dimensions, including energy concentration, filter response differences, and phase stability, effectively amplifying the subtle differences between real targets and clutter false alarms. Based on the constructed multi-dimensional feature space, this invention inputs the extracted multi-dimensional features into a pre-trained classifier for secondary decision-making. Utilizing the classifier's pattern recognition capabilities, it learns and leverages the distribution patterns of targets and clutter in the high-dimensional feature space, thereby more accurately eliminating false alarms. Through the combination of multi-dimensional feature fusion and secondary decision-making by the classifier, this invention achieves effective suppression of clutter false alarms in complex environments, significantly improving the reliability of radar target detection.

[0057] To further illustrate the technical effects of the present invention, the method of the present invention is verified below using measured data collected by a Ku-band low-altitude sounding radar.

[0058] 1. Experimental setup and radar parameters: This experiment uses a Ku-band low-altitude detection radar, which employs two alternating pulse transmission modes to balance detection range and range resolution. The main parameters of the radar system are shown in Table 1.

[0059] Table 1 Radar System Parameters

[0060] The experimental scenario was selected as a typical suburban environment. The area is surrounded by a large number of residential buildings, construction sites, farmland, forests and roads. The ground clutter types are rich and unevenly distributed, with strong multipath effects and dynamic interference sources (such as moving vehicles). This can fully test the false alarm suppression capability of the method of the present invention in complex clutter environments.

[0061] 2. Feature separability verification: Following steps S10 to S40, the acquired radar echo data is processed, and four features are extracted for each potential target point. (See also...) Figure 3 , Figure 3 Two-dimensional feature scatter plots of two of these features at the target point and clutter point are shown, where the horizontal axis represents the target energy distribution characteristics, and the vertical axis represents the spectral difference characteristics before and after clutter suppression. Figure 3 It can be seen that the extracted features are effective for the target point ( Figure 3 (red dots in the image) and clutter dots ( Figure 3 The blue dots in the image are clearly distinguishable, and the target point and clutter points exhibit separable distribution characteristics in the feature space, which lays a good foundation for the accurate decision of the subsequent classifier.

[0062] 3. Comparison of false alarm suppression effects: To verify the false alarm suppression effect of the method of this invention, this experiment used both the traditional constant false alarm rate (CFAR) detection algorithm and the method of this invention to perform target detection on 40 consecutive radar observation data. The detection results are as follows: Figure 4 and Figure 5 As shown.

[0063] Figure 4 The results are from the traditional constant false alarm rate (CFAR) detection algorithm. Figure 4 The image shows all the detection points accumulated over 40 observation cycles, totaling 14,337. It can be seen that, in addition to the real targets, due to the presence of complex ground clutter and dynamic interference, the detection results contain a large number of false alarms caused by residual clutter. These false alarms are distributed throughout the entire detection area, severely affecting the radar's tracking and identification of real targets.

[0064] Figure 5 The results show the detection outcomes achieved using the method of this invention. Statistical analysis shows that the number of target points processed using this method was reduced to 1214, achieving a false alarm suppression rate of over 90%.

[0065] 4. Experimental conclusions: The above-mentioned measured data processing results demonstrate that the clutter region false alarm suppression method proposed in this invention, based on multi-feature fusion, can effectively utilize the separability of targets and clutter in multi-dimensional features such as energy distribution, spectral differences, and phase stability. Combined with a classifier for secondary decision-making, it significantly eliminates false alarms caused by complex environments while preserving the true target. Compared with existing technologies, this invention significantly improves the reliability and tracking performance of radar systems in complex urban and suburban environments, verifying the technical effectiveness and practical value of this invention.

[0066] It should be noted that the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention.

[0067] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0068] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings and the disclosure in carrying out the claimed invention. In the description of the invention, the word "comprising" does not exclude other components or steps, "a" or "an" does not exclude a plurality, and "a plurality" means two or more, unless otherwise explicitly specified. Furthermore, while different embodiments may describe certain measures, this does not mean that these measures cannot be combined to produce good results.

[0069] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for suppressing false alarms in clutter regions based on multi-feature fusion, characterized in that, include: Acquire radar echo signals, perform pulse compression processing on the radar echo signals, and obtain pulse-compressed time-domain echo data; Clutter suppression and coherent accumulation are performed on the time-domain echo data to obtain the range-Doppler matrix; By performing constant false alarm rate (CFAR) detection on the range-Doppler matrix, several potential target points are obtained; For each potential target point, the corresponding multipulse echo signal and the unsuppressed frequency domain signal are extracted from the time-domain echo data, and the corresponding clutter-suppressed frequency domain signal is extracted from the range-Doppler matrix. Based on the multipulse echo signal, the clutter-suppressed frequency domain signal, and the unsuppressed frequency domain signal, multidimensional features of the potential target point are extracted. The multidimensional features include at least one of the following: target energy distribution features, spectral difference features before and after clutter suppression, and energy-phase composite features. The multidimensional features are input into a pre-trained classifier. Based on the output of the classifier, the potential target point is determined to be a target or a false alarm clutter, and the final target point trace is output.

2. The clutter region false alarm suppression method based on multi-feature fusion according to claim 1, characterized in that, The target energy distribution characteristics are calculated in the following manner: ； in, This represents the frequency domain signal without clutter suppression. Indicates the Doppler cell index. Indicates the index of the Doppler cell where the potential target point is located. Indicates the total number of Doppler elements. express The Middle The amplitude of each Doppler unit, This indicates the percentage of the target energy.

3. The clutter region false alarm suppression method based on multi-feature fusion according to claim 2, characterized in that, The spectral difference characteristics before and after clutter suppression are calculated in the following manner: ； ； in, Indicates the Doppler cell index. This represents the frequency domain signal after clutter suppression. Normalized information entropy represents the difference in signal energy before and after clutter suppression.

4. The clutter region false alarm suppression method based on multi-feature fusion according to claim 3, characterized in that, The energy-phase composite feature includes: the product of the target-zero frequency energy ratio and the phase variance after clutter suppression, and the product of the target-zero frequency energy ratio and the phase variance before clutter suppression; The product of the target-zero frequency energy ratio after clutter suppression and the phase variance is calculated in the following manner: ； in, express The amplitude of clutter in the zero-frequency channel. express The amplitude of the Doppler cell containing the potential target. Indicates a multi-pulse echo signal. express The Middle The complex value of the echo of each pulse. express phase, Represents variance. This represents the product of the target-zero frequency energy ratio and the phase variance after clutter suppression. The product of the target-zero frequency energy ratio before clutter suppression and the phase variance is calculated as follows: ； in, express The amplitude of clutter in the zero-frequency channel. express The amplitude of the Doppler cell containing the potential target. This represents the product of the target-zero frequency energy ratio and the phase variance before clutter suppression.

5. The clutter region false alarm suppression method based on multi-feature fusion according to claim 1, characterized in that, The classifier is trained in the following manner: Multiple potential target points are acquired, and the true category of the potential target points is identified using prior information; Extract the multidimensional features of the potential target points, and use the true category as the true category label corresponding to the multidimensional features to generate a training sample set; The classifier is trained using the training sample set to obtain a trained classifier.

6. The clutter region false alarm suppression method based on multi-feature fusion according to claim 5, characterized in that, The prior information includes at least one of terrain data, target trajectory data, or manually calibrated data.

7. The clutter region false alarm suppression method based on multi-feature fusion according to claim 1, characterized in that, The classifier is one of the following: support vector machine, linear decision analysis classifier, correlation vector machine, or K-nearest neighbor classifier.

8. The clutter region false alarm suppression method based on multi-feature fusion according to claim 1, characterized in that, The clutter suppression and coherent accumulation of the time-domain echo data to obtain the range-Doppler matrix includes: Based on clutter modeling, an FIR filter bank is designed and applied in parallel to the slow time-dimensional sequence of the time-domain echo data to obtain the range-Doppler matrix.

9. The clutter region false alarm suppression method based on multi-feature fusion according to claim 1, characterized in that, The constant false alarm rate (CFAR) detection performed on the range-Doppler matrix yields several potential target points, including: For each unit to be detected in the distance-Doppler matrix, calculate the adaptive detection threshold corresponding to that unit; The signal amplitude of the unit to be detected is compared with the adaptive detection threshold. If the signal amplitude of the unit to be detected exceeds the adaptive detection threshold, the unit is identified as a potential target point.

Images