Low and slow target and sea surface target identification method based on one-dimensional image and modulation characteristics

Abstract

The present application relates to a low slow small and sea surface target identification method based on one-dimensional image and modulation characteristics. The method is mainly suitable for low slow small target and sea surface target classification and identification of radar with one-dimensional wideband imaging and high repetition frequency detection. The main process is: first, after the radar stably tracks the target, the target window one-dimensional image imaging and high repetition frequency detection information is obtained; the radar is guided to image the target one-dimensionally and detect the target with high repetition frequency; then the target one-dimensional image characteristics and micro-Doppler modulation characteristics are extracted respectively; finally, the target energy coding matrix is constructed by using the target one-dimensional image characteristics and micro-Doppler modulation characteristics, and the low slow small target and sea surface target classification and identification are carried out. The method provided by the present application can fully exert the advantages of radar one-dimensional image imaging and high repetition frequency detection, make up for the insufficient classification and identification ability of low slow small target and sea surface target, and comprehensively improve the identification ability.

Description

Technical Field

[0001] This invention relates to a target recognition method. Background Technology

[0002] Non-traditional air and sea targets with speeds less than 50 m / s and altitudes below 300 meters (such as slow-moving helicopters, hovering helicopters, and drones) and surface targets (such as small fishing boats) are increasingly becoming important targets for early warning and detection equipment. These non-traditional threat targets generally have small radar cross-sections (RCS) and are often mixed in with various civilian targets, capable of hiding among large ships or even being submerged in sea clutter. Relying solely on target motion information is insufficient for effectively classifying and identifying these targets as low-speed, small, or surface targets.

[0003] Radar target recognition refers to the process of extracting target features and determining target type and attributes based on electromagnetic scattering mechanisms, using radar-detected target and environmental echo signals. Modern radar systems have evolved from traditional narrowband measurement to broadband measurement. By transmitting broadband / ultra-wideband signals, range resolution is improved, allowing for the separation of scattering structures on the target. This facilitates the characterization of the target's shape and contour, leading to the development of high-resolution one-dimensional image-based target recognition technology. Furthermore, modern radar systems have evolved from traditional non-coherent systems to coherent systems. By transmitting coherent pulse train signals, the phase information of the target echo can be better extracted, enabling a more detailed description of the target's motion and attitude. This has led to the development of target recognition technologies based on velocity, micro-Doppler, and other parameters.

[0004] High-resolution one-dimensional image target recognition and micro-Doppler target recognition each have their limitations. High-resolution one-dimensional images can obtain the radial length of the target through the distribution of scattering points, and the actual length can be estimated by combining the target's flight attitude. However, it is greatly affected by the target's attitude angle, and its recognition capability is difficult to guarantee when the attitude angle exceeds 70 degrees. Micro-Doppler contains detailed information such as the geometric components and micro-motions of the target's micro-movements. By detecting whether the target has additional modulation features, it can determine whether it is an aerial target. However, it is greatly affected by the radar's dwell time and repetition rate.

[0005] The complementary advantages of high-resolution one-dimensional imaging and micro-Doppler target recognition are obvious. They can give full play to the advantages of each recognition method, make up for their shortcomings, and comprehensively improve the classification and recognition capabilities of low, slow and small targets and sea surface targets. Summary of the Invention

[0006] The purpose of this invention is to provide a method for classifying and identifying low-speed, small targets and sea surface targets in radar systems.

[0007] The technical solution of the present invention includes:

[0008] First, after the radar has stabilized its tracking of the target, it acquires one-dimensional image imaging and high repetition rate (HRPR) detection guidance information to guide the radar in one-dimensional image imaging and HPR detection of the target. Then, it extracts the target's one-dimensional image features and micro-Doppler modulation features respectively. Finally, it constructs a target energy coding matrix using the target's one-dimensional image features and micro-Doppler modulation features to classify and identify low-speed, small targets and sea surface targets. Specific steps include:

[0009] Step 1: After the radar has stabilized its tracking of the target, acquire the target's azimuth, range, elevation angle, heading, and speed information. Based on the acquired information, calculate the target's motion attitude and radial velocity to form a one-dimensional image and high repetition rate detection guidance information.

[0010] Step 2: Based on the guidance information, select the one-dimensional imaging and high repetition rate detection parameters to perform one-dimensional imaging and high repetition rate detection on the target;

[0011] Step 3: Guide the radar to perform one-dimensional image imaging of the target, obtain a high-resolution one-dimensional image of the target, calculate the imaging quality evaluation index of the high-resolution one-dimensional image of the target, wherein the imaging quality uses the signal-to-clutter ratio of the strongest scattering point and the mean noise as the evaluation index, and use the imaging quality evaluation index to adjust the imaging and obtain a high-quality high-resolution one-dimensional image of the target.

[0012] Step 4: Use fine edge extraction to identify target type. Set the scattering point extraction threshold for the left and right regions of the strongest scattering point of the high-resolution one-dimensional image according to the target's approaching and moving posture. Obtain the final target edge scattering points through edge detection, calculate the target radial length, and further calculate the target length in combination with the target posture to determine the target's large, medium and small type.

[0013] Step 5: Guide the radar to perform high repetition rate detection on the target, obtain the target frequency domain spectrum and the background noise frequency domain spectrum, calculate the target high repetition rate detection quality evaluation index, wherein the detection quality is evaluated by the signal-to-noise ratio of the maximum value of the target spectrum to the maximum value of the background noise spectrum, and use the detection quality evaluation index to adjust the high repetition rate detection to obtain a high-quality target frequency domain spectrum.

[0014] Step 6: Use the normal distribution kurtosis method to detect the micro-motion modulation features of the target frequency domain spectrum. Based on the zero-frequency width of the background noise spectrum, remove the zero frequency from the target spectrum, calculate the kurtosis of the target spectrum after removing the zero frequency, and determine whether there are micro-motion modulation features in the target frequency domain spectrum.

[0015] Step 7: Construct the target energy coding matrix by combining the detection results of target size type and micro-Doppler modulation features;

[0016] Step 8: Use the target energy encoding matrix to classify and identify low-speed, small targets and sea surface targets.

[0017] The invention and its engineering implementation have high application value in the field of classification and identification of low-speed, small targets and sea surface targets. It enables effective classification of low-speed, small targets and sea surface targets in radar systems, improving the accuracy of target air and sea attribute identification. The method of identifying target types using fine edge extraction effectively solves the problems of weak scattering points being easily submerged by sea clutter, difficulty in identifying them through wake currents, and difficulty in estimating ship lengths at different attitudes, improving the accuracy of ship length estimation and enhancing the ability to distinguish between large, medium, and small targets. The method of detecting target frequency domain spectral micro-motion modulation features using the normal distribution kurtosis method can accurately and effectively determine whether a target possesses micro-Doppler modulation features. This method features good adaptability and high extraction probability.

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0019] Figure 1 This is a data processing flowchart of the present invention.

[0020] Figure 2 This is a schematic diagram of the high repetition rate detection echo of a low-speed, small target.

[0021] Figure 3 This is a schematic diagram of the low-slow-small-target micro-Doppler modulation characteristics of the present invention. Detailed Implementation

[0022] This invention relates to a method for identifying low-speed, small, and sea-surface targets based on one-dimensional images and modulation features. See the attached flowchart for the flowchart. Figure 1 The steps of the preferred embodiment are as follows:

[0023] Step 1: After the radar has stabilized its tracking of the target, acquire the target's azimuth, range, elevation angle, heading, and speed information. Based on the acquired information, calculate the target's motion attitude and radial velocity to form a one-dimensional image and high-repetition-rate detection guidance information, specifically:

[0024] 1. Radar target attitude calculation

[0025]

[0026]

[0027] Where θ is the radar target attitude angle, α is the radar target heading, and β is the radar target azimuth.

[0028] 2. Based on the calculated target attitude and target velocity, calculate the target radial velocity;

[0029] 3. Based on the target's current azimuth, distance, elevation angle, speed, and other information, calculate the target's azimuth, distance, and elevation angle for the next scanning cycle, which will serve as guidance information for one-dimensional image imaging and high repetition rate detection.

[0030] Step 2: Based on the guidance information, select the one-dimensional imaging and high repetition rate detection parameters to perform one-dimensional imaging and high repetition rate detection on the target, specifically as follows:

[0031] 1. Based on the calculated target distance, select one-dimensional image imaging parameters to avoid the target falling outside the maximum unambiguous detection range;

[0032] 2. Based on the target's radial velocity and the calculated target distance, select high repetition rate detection parameters to avoid the target falling into the blind range and blind velocity of the detection.

[0033] Step 3: Guide the radar to perform one-dimensional image imaging of the target, acquire a high-resolution one-dimensional image of the target, and calculate the imaging quality evaluation index of the high-resolution one-dimensional image of the target. The imaging quality is evaluated using the signal-to-clutter ratio of the strongest scattering point to the mean noise value. The imaging quality evaluation index is used for imaging adjustment to acquire a high-quality high-resolution one-dimensional image of the target. Specifically:

[0034] 1. Estimate the number of range cells of the target in the high-resolution one-dimensional image. Based on the relationship between the resolution of the high-resolution one-dimensional image and the radial length of the target, the number of range cells occupied by the target in the high-resolution one-dimensional image can be estimated as follows:

[0035]

[0036] Where L is the radial length of the target, set according to the maximum length of existing ships on the sea surface, and s is the high-resolution one-dimensional image range resolution;

[0037] 2. Calculate the global maximum value max(X) of the high-resolution one-dimensional image X. Take the distance cell where the global maximum value is located as the strongest scattering point of the high-resolution one-dimensional image. Extend M / 2 distance cells to the left and right of the strongest scattering point as the target one-dimensional image region. Take the remaining distance cells as the noise one-dimensional image region Y. Calculate the mean value mean(Y) of the noise one-dimensional image region.

[0038] 3. Calculate the imaging quality evaluation index of this high-resolution one-dimensional image, using the signal-to-clutter ratio of the strongest scattering point to the mean noise as the evaluation index. The imaging quality index of this high-resolution one-dimensional image is P1:

[0039]

[0040] Where X is the high-resolution one-dimensional image, max(X) is the strongest scattering point of the high-resolution one-dimensional image, Y is the noisy one-dimensional image region, and mean(Y) is the mean of the noisy one-dimensional image region.

[0041] 4. Adjust the radar in one direction, then calculate the imaging quality index P2 at this moment. If P1>P2, adjust the radar in the opposite direction until the imaging quality index deteriorates, and the state at the previous moment has the best imaging quality of the target high-resolution one-dimensional image; if P1<P2, continue adjusting the radar in this direction until the imaging quality index deteriorates, and the state at the previous moment has the best imaging quality of the target high-resolution one-dimensional image.

[0042] Step 4: Use the fine edge extraction method to identify the target type, set the scattering point extraction thresholds for the left and right regions of the strongest scattering point of the high-resolution one-dimensional image according to the motion posture of the target approaching or moving away, obtain the final target edge scattering points through edge detection, calculate the radial length of the target, further calculate the target length in combination with the target posture, and distinguish the large, medium and small types of the target, which is preferably as follows:

[0043] 1. Calculate the mean and variance of the noise one-dimensional image region, and set the basic edge detection threshold according to the linear sum form of the mean and variance

[0044] Thr=mean(Y)+3*std(Y)

[0045] Wherein Thr is the basic edge detection threshold, Y is the noise one-dimensional image region, mean(Y) is the mean of the noise one-dimensional image region, and std(Y) is the variance of the noise one-dimensional image region;

[0046] 2. Judge the approaching or moving away state of the target according to the attitude angle θ. The judgment criterion is that when θ is greater than 90 degrees, it is in a moving away state, when θ is less than 90 degrees, it is in an approaching state, and when θ is equal to 90 degrees, it is in a tangential motion state;

[0047] 3. Set the scattering point extraction thresholds for the left and right regions of the strongest scattering point of the high-resolution one-dimensional image according to the motion posture of the target approaching or moving away

[0048]

[0049]

[0050] Wherein ThrLeft is the edge extraction threshold for the left region of the strongest scattering point, ThrRight is the edge extraction threshold for the right region of the strongest scattering point, and k is a constant coefficient greater than zero;

[0051] 4. According to the scattering point extraction thresholds for the left and right regions of the strongest scattering point of the high-resolution one-dimensional image, detect the range cells exceeding the detection threshold in the left and right regions of the strongest scattering point

[0052] RangeLeft={L1,L2,…,L W}

[0053] RangeRight={R1,R2,…,R E}

[0054] Where RangeLeft is the set of over-detection threshold distance units in the left region, RangeRight is the set of over-detection threshold distance units in the right region, W is the number of over-detection threshold distance units in the left region, and E is the number of over-detection threshold distance units in the right region. When (L j -L i When L > 30*s, j The corresponding distance unit is used as the target left edge detection distance unit, when (R h -R g When )>30*s, R j The corresponding distance unit is used as the distance unit for detecting the right edge of the target, where 1 <i<j<W,1<h<g<E;

[0055] 5. Calculate the target radial length

[0056] JXLength=(R j -L j )*s

[0057] Where s is the high-resolution one-dimensional image distance resolution;

[0058] 6. Calculate the target length

[0059]

[0060] Where θ is the attitude angle;

[0061] 7. Finally, identify the target type based on the target length.

[0062]

[0063] The criteria for judging a target are as follows: if the target length is less than 50 meters, it is considered a small target; if the target length is between 50 meters and 200 meters, it is considered a medium target; and if the target length is greater than 200 meters, it is considered a large target.

[0064] Step 5: Guide the radar to perform high repetition rate (HRPR) detection on the target, acquire the target frequency domain spectrum and the background noise frequency domain spectrum, and calculate the target HRPR detection quality evaluation index. The detection quality is evaluated using the signal-to-noise ratio (SNR) of the maximum target spectrum value to the maximum background noise spectrum value. The HRPR detection is then adjusted using this evaluation index to obtain a high-quality target frequency domain spectrum. Preferably:

[0065] 1. Calculate the maximum value of the target frequency domain spectrum Fre, max(Fre), and the maximum value of the background noise spectrum NoiseFre, max(NoiseFre);

[0066] 2. Calculate the high-repetition-frequency detection quality evaluation index, where the signal-to-noise ratio of the maximum value of the target spectrum and the maximum value of the background noise spectrum is adopted as the evaluation index, and the obtained high-repetition-frequency detection quality evaluation index at this time is P3:

[0067]

[0068] where Fre is the frequency-domain spectrum of the target, max(Fre) is the maximum value of the target spectrum, NoiseFre is the frequency-domain spectrum of the background noise, and max(NoiseFre) is the maximum value of the background noise spectrum;

[0069] 3. Adjust the radar in one direction, then calculate the high-repetition-frequency detection quality index P4 at this moment. If P3>P4, adjust the radar in the opposite direction until the high-repetition-frequency detection quality index deteriorates, and the state at the previous moment has the best high-repetition-frequency detection quality; if P3<P4, continue to adjust the radar in this direction until the high-repetition-frequency detection quality index deteriorates, and the state at the previous moment has the best high-repetition-frequency detection quality.

[0070] Step 6: Detect the micro-motion modulation feature of the target frequency-domain spectrum by using the normal distribution kurtosis method, perform zero-frequency elimination on the target spectrum according to the zero-frequency width of the background noise spectrum, calculate the kurtosis of the target spectrum after zero-frequency elimination, and judge whether there is a micro-motion modulation feature in the target frequency-domain spectrum, which is preferably as follows:

[0071] 1. Calculate the mean and variance of the background noise spectrum, and set the detection threshold of the background noise spectrum according to the form of linear sum of the mean and the variance

[0072] ThrNoisefre=mean(NoiseFre)+c*std(NoiseFre)

[0073] where ThrNoiseFre is the detection threshold for the zero-frequency width of the background noise spectrum, NoiseFre is the background noise spectrum, mean(NoiseFre) is the mean of the background noise spectrum, std(NoiseFre) is the variance of the background noise spectrum, and c is a constant coefficient greater than zero;

[0074] 2. Determine the region exceeding the threshold as the zero-frequency region of the background noise spectrum;

[0075] 3. Perform zero-setting processing on the corresponding position and region in the target spectrum according to the width and position of the zero-frequency region of the background noise spectrum;

[0076] 4. Calculate the kurtosis of the target spectrum after zero-frequency elimination

[0077]

[0078] Where TargetFre is the target spectrum after removing zero frequencies, μ is the mean of TargetFre, σ is the standard deviation of TargetFre, and E(TargetFre-μ) is the standard deviation of TargetFre. 4 It is the fourth-order center distance of TargetFre;

[0079] 5. Finally, the presence of micro-modulation features is determined based on the degree to which the kurtosis of the target spectrum deviates from the normal distribution after removing the zero frequency. If Kurtosis-3>0, it is determined that micro-modulation features exist; otherwise, it is determined that micro-modulation features do not exist.

[0080] Step 7: Combining the target's size type and micro-Doppler modulation feature detection results, construct the target energy encoding matrix, specifically as follows:

[0081] 1. The target energy encoding matrix is ​​constructed using 4 bits of 0 and 1 encoding. The first 3 bits represent the target size type: large target is encoded as [1 0 0], medium target is encoded as [0 1 0], and small target is encoded as [0 0 1]. The last bit represents the micro-Doppler modulation feature: the presence of micro-Doppler modulation feature is encoded as [1], and the absence of micro-Doppler modulation feature is encoded as [0].

[0082] 2. The constructed target energy encoding matrix includes six types: [1 0 0 0], [1 0 0 1], [0 1 0 0], [0 1 0 1], [0 0 1 0], and [0 0 1 1].

[0083] Step 8: Classify and identify low-speed, small targets and sea surface targets using the target energy encoding matrix, specifically as follows:

[0084] The target energy coding matrices are constructed for classification and identification. The judgment criterion is that when the target energy coding matrix is ​​[0 0 1 1], it is judged as a low, slow and small target, and the rest are judged as sea surface targets.

Claims

1. A method for recognizing low-speed, small, and sea-surface targets based on one-dimensional images and modulation features, characterized in that: Step 1: After the radar has stabilized its tracking of the target, acquire the target's azimuth, range, elevation angle, heading, and speed information. Based on the acquired information, calculate the target's motion attitude and radial velocity to form a one-dimensional image and high repetition rate detection guidance information. Step 2: Based on the guidance information, select the one-dimensional imaging and high repetition rate detection parameters to perform one-dimensional imaging and high repetition rate detection on the target; Step 3: Guide the radar to perform one-dimensional image imaging of the target, obtain a high-resolution one-dimensional image of the target, calculate the imaging quality evaluation index of the high-resolution one-dimensional image of the target, wherein the imaging quality uses the signal-to-clutter ratio of the strongest scattering point and the mean noise as the evaluation index, and use the imaging quality evaluation index to adjust the imaging and obtain a high-quality high-resolution one-dimensional image of the target. Step 4: Use fine edge extraction to identify target type. Set the scattering point extraction threshold for the left and right regions of the strongest scattering point of the high-resolution one-dimensional image according to the target's approaching and moving posture. Obtain the final target edge scattering points through edge detection, calculate the target radial length, and further calculate the target length in combination with the target posture to determine the target's large, medium and small type. Step 5: Guide the radar to perform high repetition rate detection on the target, obtain the target frequency domain spectrum and the background noise frequency domain spectrum, calculate the target high repetition rate detection quality evaluation index, wherein the detection quality is evaluated by the signal-to-noise ratio of the maximum value of the target spectrum to the maximum value of the background noise spectrum, and use the detection quality evaluation index to adjust the high repetition rate detection to obtain a high-quality target frequency domain spectrum. Step 6: Use the normal distribution kurtosis method to detect the micro-motion modulation features of the target frequency domain spectrum. Based on the zero-frequency width of the background noise spectrum, remove the zero frequency from the target spectrum, calculate the kurtosis of the target spectrum after removing the zero frequency, and determine whether there are micro-motion modulation features in the target frequency domain spectrum. Step 7: Construct the target energy coding matrix by combining the detection results of target size type and micro-Doppler modulation features; Step 8: Use the target energy encoding matrix to classify and identify low-speed, small targets and sea surface targets.

2. The method for identifying low-speed, small, and sea-surface targets based on one-dimensional images and modulation features according to claim 1, characterized in that, Step 4, which uses fine edge extraction to identify the target type, includes: First, estimate the number of range cells of the target in the high-resolution one-dimensional image. Based on the relationship between the resolution of the high-resolution one-dimensional image and the radial length of the target, estimate the number of range cells occupied by the target in the high-resolution one-dimensional image as follows: ; Where L is the radial length of the target, set according to the maximum length of existing ships on the sea surface, and s is the high-resolution one-dimensional image range resolution; Next, the global maximum value of the high-resolution one-dimensional image is calculated. The distance cell containing the global maximum value is taken as the strongest scattering point of the high-resolution one-dimensional image. Extending to the left and right from the strongest scattering point is then performed. One distance cell is used as the target one-dimensional image region, and the remaining distance cells are used as the noise one-dimensional image region. The mean and variance of the noise one-dimensional image region are calculated, and the basic threshold for edge detection is set according to the similarity of the linear sum of the mean and variance: ; Where Thr is the basic threshold for edge detection, Y is the one-dimensional noise image region, mean(Y) is the mean of the one-dimensional noise image region, and std(Y) is the variance of the one-dimensional noise image region. Based on the target's approaching and receding motion posture, set the scattering point extraction threshold for the left and right regions of the strongest scattering point in the high-resolution one-dimensional image: ; Where ThrLeft is the threshold for extracting the edge of the region to the left of the strongest scattering point, ThrRight is the threshold for extracting the edge of the region to the right of the strongest scattering point, and k is a constant coefficient greater than zero; Thresholds are extracted based on the scattering points in the regions to the left and right of the strongest scattering point in the high-resolution one-dimensional image. Distance cells exceeding the detection threshold in the regions to the left and right of the strongest scattering point are then detected. ; Where RangeLeft is the set of over-detection threshold distance cells in the left region, RangeRight is the set of over-detection threshold distance cells in the right region, W is the number of over-detection threshold distance cells in the left region, and E is the number of over-detection threshold distance cells in the right region. hour, The corresponding distance unit serves as the distance unit for detecting the left edge of the target. and These are the i-th and j-th over-detection threshold distance units in the left region, respectively. , The corresponding distance unit serves as the distance unit for detecting the right edge of the target. and These are the h-th and g-th over-detection threshold distance units in the right region, respectively. ; Calculate the target radial length: ； Then calculate the target length: ； in The attitude angle; Finally, the target type is identified based on its length. The criteria are as follows: if the target length is less than 50 meters, it is considered a small target; if the target length is between 50 meters and 200 meters, it is considered a medium target; and if the target length is greater than 200 meters, it is considered a large target.

3. The method for identifying low-speed, small, and sea-surface targets based on one-dimensional images and modulation features according to claim 1, characterized in that, Step 6, which uses the normal distribution kurtosis method to detect the target frequency domain spectrum micro-motion modulation features, includes: First, calculate the mean and variance of the background noise spectrum, and then set the background noise spectrum detection threshold based on the form of the linear sum of the mean and variance: ； The region that exceeds the threshold is defined as the zero-frequency region of the background noise spectrum, where ThrNoiseFre is the zero-frequency width detection threshold of the background noise spectrum, NoiseFre is the background noise spectrum, mean(NoiseFre) is the mean of the background noise spectrum, std(NoiseFre) is the variance of the background noise spectrum, and c is a constant coefficient greater than zero. Based on the width and location of the zero-frequency region in the background noise spectrum, the corresponding positions and regions in the target spectrum are zeroed out. Calculate the target spectral kurtosis after removing zero frequencies: ； Where TargetFre is the target spectrum after removing zero frequencies. It is the mean of TargetFre. It is the standard deviation of TargetFre. It is the fourth-order center distance of TargetFre; Finally, the presence of micro-modulation features is determined based on the degree to which the kurtosis of the target spectrum deviates from a normal distribution after removing zero frequencies. If the condition is met, it is determined that a micro-modulation feature exists; otherwise, it is determined that a micro-modulation feature does not exist.

4. The method for identifying low-speed, small, and sea-surface targets based on one-dimensional images and modulation features according to claim 1, characterized in that: Step 7, which combines the target size type and micro-Doppler modulation feature detection results, constructs the target energy encoding matrix, including: The target energy encoding matrix is ​​constructed using 4 bits of 0 and 1 encoding. The first 3 bits represent the target size type: large targets are encoded as [1 0 0], medium targets as [0 1 0], and small targets as [0 0 1]. The last bit represents the micro-Doppler modulation feature: the presence of micro-Doppler modulation feature is encoded as [1], and the absence of micro-Doppler modulation feature is encoded as [0]. The constructed target energy encoding matrix includes six types: [1 0 0 0], [1 0 0 1], [0 1 0 0], [0 1 0 1], [0 0 10], and [0 0 1 1].

5. The method for identifying low-speed, small, and sea-surface targets based on one-dimensional images and modulation features according to claim 4, characterized in that, Step 8, which uses the target energy encoding matrix to classify and identify low-speed, small targets and sea surface targets, includes: The target energy coding matrices constructed by the weights are used for classification and identification. The judgment criterion is that when the target energy coding matrix is ​​[0 0 1 1], it is judged as a low, slow and small target, and the rest are judged as sea surface targets.

Images