A sea surface slow small target detection method based on three-dimensional feature gridding

Abstract

The application discloses a sea surface slow small target detection method based on three-dimensional feature gridding, first, on the basis of traditional three features, an improved feature extraction method is proposed, combining the idea of whitening, the feature difference of traditional three features is improved; secondly, a three-dimensional feature gridding detector is designed, by uniformly dividing the three-dimensional space grid, combining the closed operation operation, the best clutter distribution fitting envelope is obtained, which is used as the decision threshold of sea clutter and target, the accuracy and real-time of sea target detection is improved. The application solves the problems of low feature distinguishability and poor decision threshold fitting of the existing three feature method, can obtain more accurate decision region, and effectively improves the target detection performance and real-time.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of radar technology, in particular to a sea surface slow small target detection method based on three-dimensional feature gridding, which realizes high-reliable and fast detection of slow small targets in the sea clutter background. BACKGROUND

[0002] As an active microwave target detection device, radar plays an irreplaceable role in the sea area stereoscopic monitoring system. However, when detecting low-altitude slow small targets, the radar works in a low-grazing angle mode, and the small targets take advantage of the natural advantages of sea clutter to make their echoes submerged in strong clutter, greatly increasing the difficulty of radar detection.

[0003] At present, the commonly used sea detection methods can be divided into two categories: energy detection and feature detection. The energy detection method mostly uses the difference between the target and the background clutter energy to achieve the purpose of sea surface target detection, but for low observable targets, the target echo energy is weak, and it is highly coincident with the main clutter area in time domain and frequency domain. It is difficult to achieve stable and robust target detection through a single energy detection method; the feature detection method maps the echo signal into different feature spaces, extracts stable and differentiable features in the feature space, and uses the feature difference to realize high-reliable detection of the target. The feature detection method provides an effective solution for sea target detection, and the core is the extraction and detection of differentiated features. In the aspect of feature extraction, the existing feature extraction mainly includes traditional three features (relative average amplitude RAA, relative Doppler peak height RDPH, and relative vector entropy RVE) obtained in one-dimensional time domain / frequency domain, and time-frequency three features (ridge accumulation, maximum connected region size, and connected region number) obtained in time-frequency two-dimensional domain. Among them, the time-frequency three features can obtain good results, but the one-dimensional time domain echo signal needs to be transformed into time-frequency two-dimensional domain when extracting features, and the calculation complexity limits its application in practice; in the aspect of feature detector, the "convex hull" design method is mainly used, but due to the sparse and incomplete characteristics of the target samples in the echo, and the difficulty of the convex hull to fully fit the distribution characteristics of the clutter in the three-dimensional feature space, a large number of target misses occur. SUMMARY

[0004] In view of the problems of low feature distinguishability and poor decision threshold fitting of the existing three-feature method, the purpose of the present application is to provide a sea surface slow small target detection method based on three-dimensional feature gridding, so as to obtain a more accurate decision region and improve the target detection performance and real-time performance.

[0005] In order to achieve the above-mentioned task, the present application adopts the following technical scheme:

[0006] A sea surface slow small target detection method based on three-dimensional feature gridding, comprising:

[0007] For pulse radar echo signals of known targets, the echo signals are sorted and arranged according to pulses to obtain a two-dimensional echo matrix. In the two-dimensional echo matrix, each row of data represents the sampling point of a pulse, and each column of data represents the sampling point of each range gate. Each pulse is pre-processed to obtain the time spectrum data of each range gate.

[0008] The time spectrum data of each range gate is converted into Doppler spectrum data through fast Fourier transform. Then, each range gate is used as the range gate to be whitened, and the range gates on both sides of it are used as reference range gates. The median and standard deviation of each Doppler frequency point in the range gate to be whitened are calculated. Then, the Doppler frequency point of the range gate to be whitened is subtracted from the corresponding Doppler frequency point median and divided by the standard deviation to obtain the whitened result. The whitened result is then inverse fast Fourier transformed to restore it to time spectrum data.

[0009] The range gates other than the one where the target is located are divided into sliding windows. Each window of the range gate division is used as a sea clutter sample unit, and the units on both sides of each sea clutter sample unit are used as reference units. Three features are extracted from each sea clutter sample unit: RAA feature, RDPH feature, and RVE feature.

[0010] After extracting the three features of all sea clutter sample units, the feature vector of each sea clutter sample unit is constructed using the three extracted features, thereby constructing a training sample set. Then, a three-dimensional feature space is constructed with the three features as coordinate axes, and the feature vector of the sea clutter sample unit is used as the three-dimensional coordinate index to plot the corresponding feature point position of the feature vector of each sea clutter sample unit in the three-dimensional feature space.

[0011] Based on the extreme values ​​of the three features of all sea clutter sample units, a cuboid region is constructed in a three-dimensional coordinate system as a segmentation area, the segmentation area is divided into a three-dimensional grid array, and the number of feature points corresponding to the sea clutter sample units falling into each grid is counted to determine the attributes of each grid; a three-dimensional initial binary image with the same dimension as the three-dimensional grid array is established, and pixel points of the three-dimensional initial binary image are assigned values ​​according to the grid attributes;

[0012] Performing a morphological closing operation on the three-dimensional initial binary image, thereby filling and connecting small holes and disconnected parts in the three-dimensional initial binary image and removing the outer boundary noise of the image to obtain an updated binary image; performing outlier removal on the updated binary image to obtain a removed binary image; and adjusting the properties of the grid in the three-dimensional grid array according to the removed binary image;

[0013] In the actual detection stage, after obtaining the new radar echo signal, the pulse data thereof is preprocessed, then white processing is performed according to the range gate, all the range gates after the white processing are windowed, and the range gates in each window are taken as a unit to be detected; three features of each unit to be detected are extracted, a feature vector of the unit to be detected is constructed, the feature vector is taken as a three-dimensional coordinate index and mapped into a feature point to be detected in a three-dimensional feature space, and detection is performed according to the position of the feature point to be detected:

[0014] If the feature point to be detected falls outside the space where the three-dimensional grid array is located or falls into a certain grid of the three-dimensional grid array, but the attribute of the grid is a non-sea clutter grid, then the feature point to be detected is classified as a target; otherwise, it is classified as sea clutter.

[0015] Further, the pre-processing includes down-conversion and pulse compression.

[0016] Further, the median and standard deviation of each Doppler frequency point in the distance gate to be whitened are calculated, then the Doppler frequency point of the distance gate to be whitened is subtracted from the corresponding Doppler frequency point and divided by the standard deviation to obtain the result after whitening, including:

[0017]

[0018] Wherein, x(f) represents the Doppler spectrum data of the distance gate to be whitened, x p (f) represents the Doppler spectrum data of the pth reference distance gate, P represents the number of reference distance gates, median(·) represents the calculation of the median, med and std represent the median sequence and the standard deviation sequence of the Doppler frequency point f respectively; x norm (f) is the Doppler spectrum data of the distance gate after whitening, and the time spectrum data x norm (n) can be restored by performing inverse fast Fourier transform on x norm (n).

[0019] Further, the extraction formulas of the RAA feature, the RDPH feature and the RVE feature are as follows:

[0020] The RAA feature extraction formula is:

[0021]

[0022] Wherein, and respectively represent the mean values of the sea clutter sample unit and the reference unit energy, xnorm(n) and x pnorm (n) are the time spectrum data of the whitened sea clutter sample unit and the corresponding pth reference unit, n is the time sampling point, N is the total number of time sampling points, and P' represents the number of reference units;

[0023]

[0024] where Peak represents the Doppler peak value of each sea clutter sample unit, x norm (f) represents the whitened range-Doppler spectrum data, f represents different Doppler frequency points, T is the pulse repetition time, and the corresponding Doppler channel is denoted as f max (x norm (f));

[0025] At this time, DPH can be represented as:

[0026]

[0027] where Δ = [-δ1, -δ2]∪[δ2, δ1], δ1 and δ2 represent the maximum Doppler bandwidths occupied by sea clutter and targets respectively, and #Δ represents the number of Doppler frequency points remaining after excluding the Doppler frequency points possibly occupied by targets from the frequency point region in the sea clutter sample;

[0028] Finally, the RDPH feature is:

[0029]

[0030] where x pnorm (f) represents the whitened Doppler spectrum reference reference unit;

[0031]

[0032] where, Therefore, the RVE feature is:

[0033]

[0034] Further, the extreme values of the three features based on all sea clutter sample units construct a cuboid region as a segmentation zone in a three-dimensional coordinate system, perform three-dimensional grid array division on the segmentation zone, count the number of feature points corresponding to the sea clutter sample units falling into each grid, and thus determine the attributes of each grid, including:

[0035] First, the maximum and minimum values of the three features of all sea clutter sample units are calculated respectively, and the corresponding endpoint positions in the three coordinate axes are found. A cuboid region surrounded by six planes corresponding to the six endpoints is taken as a segmentation zone, denoted as P ar ; The region is the smallest cuboid space parallel to the coordinate axes containing the feature points corresponding to all sea clutter sample units; the segmentation zone P ar is divided uniformly according to the dimensions of the three coordinates, so as to divide the segmentation zone P arThe three-dimensional grid array is divided, and each grid is set with an initial value num=0 to record the number of feature points of the sea clutter sample unit falling in the grid;

[0036] Then, the feature points of the sea clutter sample unit are traversed one by one, and the num of the grid is incremented by 1 for each feature point falling in the grid.

[0037] Further, a three-dimensional initial binary image with the same dimensions as the three-dimensional grid array is established, and the pixel points of the three-dimensional initial binary image are valued according to the grid attributes, including:

[0038] A three-dimensional initial binary image B with the same dimensions as the three-dimensional grid array is constructed according to the three-dimensional grid array; the grid with a num count greater than 0 is referred to as a sea clutter grid, and the corresponding pixel point in the three-dimensional initial binary image B is recorded as 1, and the rest are non-sea clutter grids, and the corresponding pixel points are recorded as 0.

[0039] Further, the closing operation process includes:

[0040] The specific process of the operation is to set a structure element "se" to first perform "dilation" on the original image B, and then perform "erosion" operation; wherein the structure element "se" is the basic unit of "dilation" and "erosion" operation, and the structure element "se" is made to slide in the initial binary image B, and the "se" origin position is set to 1 and 0 according to the "dilation" and "erosion" definition; here, the structure element "se" can also be selected in the cuboid structure according to the initial binary image B, and the center point of the structure element "se" is taken as the origin position; the updated binary image is denoted as B C , B C The point with a value of 1 in the three-dimensional grid array is recorded as a sea clutter grid, and the point with a value of 0 is set to zero and recorded as a non-sea clutter grid.

[0041] Further, the process of removing outliers is:

[0042] The updated binary image B C with a value of 1 is counted, and the connected region with the largest number of grids in the three-dimensional grid array is referred to as the main connected region, and the remaining connected regions outside the main connected region are set to zero as outliers in the three-dimensional grid array and discarded; the binary image after removing the outliers is denoted as J, and the grid corresponding to the point with a value of 1 in the three-dimensional grid array is set as a sea clutter grid, and the rest are non-sea clutter grids.

[0043] Further, the number of feature points of the sea clutter sample unit corresponding to the grid in the three-dimensional grid array that is set to zero in the morphological closing operation process and the outlier removal operation is counted, and the false alarm rate of the current classifier can be obtained by the number of feature points and the total number of sea clutter sample units.

[0044] A sea surface slow small target detection device, comprising a processor, a memory and a computer program stored in the memory; when the processor executes the computer program, the sea surface slow small target detection method based on three-dimensional feature gridding is realized.

[0045] A computer readable storage medium, the medium has a computer program stored therein; when the computer program is executed by a processor, the sea surface slow small target detection method based on three-dimensional feature gridding is realized.

[0046] Compared with the prior art, the present application has the following technical features:

[0047] 1. The extraction method of the traditional three features is improved, and the discrimination degree of the features in the three-dimensional space is improved; 2. Through the three-dimensional gridding processing of the features, the threshold of the decision maker and the clutter distribution are more fitted, and the detection performance under the complex clutter background is improved. The algorithm is verified by using the public IPIX measured sea clutter data set, compared with the traditional three features + convex hull method, the average detection rate performance of the target is improved by 20.34% under the same false alarm rate condition. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 The sea clutter distribution and the convex hull decision region are shown in the figure.

[0049] Figure 2 The flowchart of the method of the present application is shown in the figure.

[0050] Figure 3 The schematic diagram of the original echo matrix in an embodiment of the present application is shown in the figure.

[0051] Figure 4 The schematic diagram of the segmented region in an embodiment of the present application is shown in the figure.

[0052] Figure 5 The sea clutter grid corresponding to the initial binary image in an embodiment of the present application is shown in the figure.

[0053] Figure 6 The sea clutter grid corresponding to the binary image after the closing operation in an embodiment of the present application is shown in the figure.

[0054] Figure 7 The sea clutter grid corresponding to the binary image after removing the wild points in an embodiment of the present application is shown in the figure.

[0055] Figure 8 The detection rate comparison of the present application and the traditional fast convex hull method under the same false alarm rate is shown in the figure.

[0056] Figure 9 The training time comparison of the present application and the traditional fast convex hull method under the same false alarm rate is shown in the figure.

[0057] Figure 10 The detection time of the present application and the traditional fast convex hull method under the same false alarm rate is compared. DETAILED DESCRIPTION

[0058] In order to further improve the target detection performance under complex random environment, the present application proposes a sea surface slow small target detection method based on three-dimensional feature gridding. Firstly, the method optimizes the extraction of traditional three features (relative average amplitude, relative Doppler peak height, and relative vector entropy) by using the frequency domain whitening method, and improves the difference of sea clutter and target features in three-dimensional space. Then, by using the idea of limit approximation, a three-dimensional feature gridding detector is designed to fit the profile of clutter distribution through the grid, overcoming the "fitting blank area" caused by the "convex hull" method (as shown in Figure 1 ), and improving the detection performance in complex scenes. Figure 1 As can be seen from the figure, there is a large blank area in the convex hull decision region, which does not conform to the sea clutter distribution profile, and this part will affect the subsequent target detection and increase the false alarm rate.

[0059] Referring to the accompanying Figure 2 , the present application provides a sea surface slow small target detection method based on three-dimensional feature gridding, which comprises the following steps:

[0060] Step 1: For the pulse system radar echo signal of a known target, the echo signal is sorted and arranged according to the pulse to obtain a two-dimensional echo matrix; in the two-dimensional echo matrix, each row of data represents a sampling point of a pulse, and each column of data represents a sampling point of each range gate; each pulse (i.e. each row of data) is subjected to pre-processing such as down-conversion and pulse compression to obtain time spectrum data of each range gate data (i.e. each column of data).

[0061] Feature extraction is a pre-step of detector design, and a corresponding detector can be trained only after a training sample set is obtained by extracting sea clutter features from the sea clutter radar echo of a known target. In this step, the two-dimensional echo matrix of the received echo signal of the pulse system radar is established and each pulse data is pre-processed for feature extraction; wherein the range gate of the target is known; the two-dimensional echo matrix constructed by an embodiment of the present application is shown in Figure 3 .

[0062] Step 2: The time spectrum data of each range gate is first transformed into Doppler spectrum data by fast Fourier transform, then each range gate is taken as a to-be-whitened range gate, the range gates on both sides are taken as reference range gates, the median and improved standard deviation of each Doppler frequency point in the to-be-whitened range gate are calculated, then the Doppler frequency points of the to-be-whitened range gate are subtracted from the corresponding Doppler frequency points and divided by the improved standard deviation to obtain the whitened result, and the whitened result is inverse fast Fourier transformed to restore the time spectrum data.

[0063] The application adopts the whitening processing idea, and performs whitening processing before extracting three features of sea clutter, so as to improve the difference between the target and the sea clutter features.

[0064] The whitening operation is performed in units of range gates. The range gate data after pre-processing is time spectrum data, denoted as x(n), wherein n is a time sampling point. Fast Fourier transform is performed on x(n) to obtain Doppler spectrum data, denoted as x(f), wherein f is a Doppler frequency point. The Doppler spectrum of sea clutter has the same distribution between range gates, so the range gates on both sides of the distance gate to be whitened are used as references to calculate the mean value and standard deviation of each Doppler frequency point. Then, the Doppler frequency point of the distance gate to be whitened is divided by the standard deviation after being subtracted by the mean value of the corresponding frequency point to obtain the whitened result, and inverse fast Fourier transform is performed on the whitened result to restore the time spectrum data. In order to avoid errors caused by individual maximum or minimum values, the mean value in the mean value and standard deviation calculation can be further replaced by the median value. The calculation formula of whitening is as follows:

[0065]

[0066] Wherein, x(f) represents the Doppler spectrum data of the distance gate to be whitened, x p (f) represents the Doppler spectrum data of the pth reference distance gate, and P represents the number of reference distance gates, which is set according to actual needs; median(·) represents the calculation of the median value, that is, the sequence composed of the median value of each Doppler frequency point f in the P reference distance gates; med and std represent the median value sequence and the improved standard deviation sequence of the Doppler frequency point f respectively; x norm (f) is the Doppler spectrum data of the whitened distance gate, and inverse fast Fourier transform is performed on x norm (f) to restore the time spectrum data x norm (n).

[0067] After all the range gates are whitened in the frequency domain according to the above operation, feature extraction can be performed.

[0068] Step 3, the remaining distance gates except the distance gate where the target is located are divided into windows, each window of the distance gate division is a sea clutter sample unit, and the units on both sides of each sea clutter sample unit are its reference units; RAA feature, RDPH feature and RVE feature are extracted from each sea clutter sample unit.

[0069] The scheme performs sliding window processing on all distance gates before feature extraction; the sliding window processing is only performed on the remaining distance gates except the distance gate where the target is located, i.e., the sea clutter distance gate, so as to obtain a sea clutter sample unit; the number of windows divided by each distance gate is set according to actual requirements, and specifically depends on the number of pulses actually contained by each distance gate; finally, three features of each sample unit are extracted.

[0070] (1) Extract the relative average amplitude (RAA) feature.

[0071] The RAA feature is extracted from the time domain, and mainly reflects the difference in energy between the target and the sea clutter. It is known from the analysis of the time domain graph that the distance gate containing the target echo has stronger energy, and therefore the RAA calculation formula is as follows:

[0072]

[0073] wherein, and respectively represent the mean values of the sea clutter sample unit and the reference unit energy, xnorm(n) and x pnorm (n) are the time spectrum data of the whitened sea clutter sample unit and the corresponding pth reference unit, n is a time sampling point, N is the total number of time samples, and P' represents the number of reference units.

[0074] (2) Extract the relative Doppler peak height (RDPH) feature.

[0075] The RDPH feature is extracted from the Doppler frequency domain, and mainly reflects the difference in Doppler amplitude spectrum (DAS) energy between the target and the sea clutter. Through the analysis of the frequency domain graph, it can be known that the target energy on the DAS is generally higher, and there is a sharp peak, while the sea clutter peak is generally lower than the target, and presents a flat peak. Therefore, first, the Doppler peak height (DPH) in each sea clutter sample unit is calculated, and the calculation formula is as follows:

[0076]

[0077] wherein, Peak represents the Doppler peak value of each sea clutter sample unit, x norm (f) represents the whitened distance gate Doppler spectrum data, f represents different Doppler frequency points, T is the pulse repetition time (PRT), and the corresponding Doppler channel is denoted as f max (x norm (f)).

[0078] At this time, the DPH can be expressed as:

[0079]

[0080] where Δ = [-δ1, -δ2]∪[δ2, δ1], δ1and δ2represent the Doppler maximum bandwidths occupied by the sea clutter and the target respectively, and #Δ represents the number of Doppler bins remaining after excluding the Doppler bins possibly occupied by the target from the frequency bins in the sea clutter sample.

[0081] Finally, the RDPH is defined as:

[0082]

[0083] where x pnorm (f) represents the whitened Doppler spectrum reference reference unit.

[0084] (3) Extract the relative vector-entropy (RVE) feature.

[0085] The RVE feature is extracted from the Doppler frequency domain, and mainly reflects the concentration of the energy of the target and the sea clutter. Through the analysis of the frequency domain graph, it can be seen that the energy of the target on the DAS is more concentrated than that of the sea clutter. The vector entropy (VE) of the data can be expressed as:

[0086]

[0087] where, Therefore, the RVE can be expressed as:

[0088]

[0089] Step 4, after extracting the three features of all sea clutter sample units, a feature vector F = [RAA, RDPH, RVE] of each sea clutter sample unit is constructed by the three extracted features, so as to construct a training sample set; then a three-dimensional feature space is constructed with the three features as the coordinate axes, and the feature vector of each sea clutter sample unit is used as the three-dimensional coordinate index to draw the corresponding feature point position of the feature vector of each sea clutter sample unit in the three-dimensional feature space.

[0090] Step 5, based on the extreme values of the three features of all sea clutter sample units, a cuboid region is constructed in a three-dimensional coordinate system as a segmentation area, the segmentation area is divided into a three-dimensional grid array, the number of feature points corresponding to the sea clutter sample units falling into each grid is counted, and thus the attributes of each grid (sea clutter grid or non-sea clutter grid) are determined; a three-dimensional initial binary image with the same dimensions as the three-dimensional grid array is established, and the pixel points of the three-dimensional initial binary image are valued according to the grid attributes.

[0091] After obtaining the three-dimensional feature space distribution of the feature points corresponding to the sea clutter sample units, a detector is designed based thereon; first, three-feature grid processing is performed, and the processing procedure is as follows:

[0092] First, the maximum and minimum values of the three features of all sea clutter sample units are calculated respectively, and the corresponding end point positions in the three coordinate axes are found, and a cuboid region surrounded by six planes corresponding to the six end points is taken as a segmentation area, denoted as P ar , as shown in Figure 4 ; the area is the smallest cuboid space parallel to the coordinate axes containing all feature points corresponding to the sea clutter sample units; the segmentation area P ar is uniformly segmented according to the dimensions of the three coordinates (for example, the interval determined by the maximum and minimum values of the RAA feature is uniformly segmented, and the number of partitions can be selected as 200 according to experience), so as to divide the segmentation area P ar into a three-dimensional grid array, wherein each grid is provided with an initial value num=0 for recording the number of feature points corresponding to the sea clutter sample units falling into the grid.

[0093] Then, the feature points corresponding to the sea clutter sample units are traversed one by one, and the num of the grid is counted +1 for each feature point falling into the grid.

[0094] Finally, a three-dimensional initial binary image B with the same dimension size as the three-dimensional grid array is constructed; the grid with num count greater than 0 is called a sea clutter grid, and the pixel point corresponding to the sea clutter grid in the three-dimensional initial binary image B is recorded as 1, and the rest is a non-sea clutter grid, and the corresponding pixel point is recorded as 0; the judgment formula is as follows:

[0095]

[0096] Wherein, x, y, z represent the index values of the three dimensions of the three-dimensional initial binary image B, and the value ranges are x=1, 2,…X, y=1, 2,…Y, z=1, 2,…Z, X, Y, Z are the number of grids in three dimensions respectively; the initial binary image B is as shown in Figure 5 , wherein the marked grid is a sea clutter grid, and there are feature points corresponding to the sea clutter sample units inside, and the non-sea clutter grid is not drawn. From Figure 5It can be seen that the grid is not continuous at this time, and there are many gaps and discrete grids.

[0097] Step 6, the morphological closing operation processing is performed on the three-dimensional initial binary image, so that the small holes and disconnected parts in the three-dimensional initial binary image are filled and connected, and the image boundary noise is removed, to obtain an updated binary image; the wild points are removed from the updated binary image to obtain a removed binary image; and the properties of the grids in the three-dimensional grid array are adjusted according to the removed binary image.

[0098] (1) Morphological closing operation processing.

[0099] The distribution of the feature points corresponding to the sea clutter sample unit in the three-dimensional space has a non-uniform characteristic. At the edge where the probability density is low, the distance between the sea clutter features is large, resulting in many gaps and discrete points in the initial binary image, which seriously affects the judgment. To solve this problem, the morphological closing operation is used to process the initial binary image B.

[0100] The specific process of the operation is to set a structure element "se" to perform "expansion" on the original image B, and then perform "erosion" operation; wherein the structure element "se" is the basic unit of "expansion" and "erosion" operation, and the structure element "se" is slid in the initial binary image B, and the "expansion" and "erosion" definitions are performed at the original position of the center point of "se" to set 1 and 0. Here, the structure element "se" can also be selected in the cuboid style according to the cuboid structure of the initial binary image B, and the center point of the structure element "se" is taken as the original position.

[0101] The "expansion" operation can fill and connect the small holes and disconnected parts in the binary image; and the "erosion" operation can remove the image boundary noise. Through the closing operation, the gaps are filled, the disconnected points are connected, the boundary is smoother and more continuous, and the profile is more consistent with the original sea clutter distribution; the updated binary image is denoted as B C , B C The grid corresponding to the point with a value of 1 in the three-dimensional grid array is denoted as sea clutter grid, and the grid corresponding to the point with a value of 0 is set to zero and denoted as non-sea clutter grid, as shown in Figure 6 . At this time, most of the sea clutter grids are connected as a whole, and only a few sea clutter grids are still far away from the main body. At the same time, the profile of the sea clutter grid main body is more consistent with the sea clutter distribution.

[0102] (2) Remove wild points.

[0103] The binary image B C processed by the morphological closing operation may still have a small number of points with a value of 1 far away from the main body, and the grid corresponding to such points in the three-dimensional grid array is denoted as a discrete grid, as shown inFigure 6 Since such grids are far away from the main clutter distribution, they can be discarded as outliers; the specific operation is to count the connected regions of the binary image B C with value 1, and the connected region with the largest number of grids in the three-dimensional grid array is called the main connected region. The remaining connected regions outside the main connected region are set to zero as outliers in the three-dimensional grid array, and discarded; the binary image after removing the outliers is denoted as J, and the grids corresponding to the points with value 1 in J are set as sea clutter grids in the three-dimensional grid array, and the rest are non-sea clutter grids, as shown in Figure 7 The region is the main connected region of the binary image after the closing operation, and the connected mode is 18-connected. The sea clutter grids scattered outside the main connected region are set to zero as outliers, thereby controlling the false alarm rate.

[0104] Step 7, in the actual detection stage, after the new radar echo signal is obtained, the pulse data is preprocessed, then the whitening processing is performed according to the range gate, and the window division is performed on all the range gates after the whitening processing, and the range gates in each window are taken as a detection unit;

[0105] Three features of each detection unit are extracted, a feature vector of the detection unit is constructed, the feature vector is mapped to a detection feature point in a three-dimensional feature space as a three-dimensional coordinate index, and the detection is performed according to the position of the detection feature point:

[0106] If the detection feature point falls outside the space P ar or falls into a grid of the three-dimensional grid array, but the attribute of the grid is a non-sea clutter grid, then the detection feature point is classified as a target; otherwise, it is classified as sea clutter; the decision formula is as follows:

[0107]

[0108] Where CUT represents the detection feature point.

[0109] Based on the number of feature points of the sea clutter sample unit in the grid set to zero in the three-dimensional grid array in the closing operation processing and outlier removal process, the false alarm rate of the classifier is set.

[0110] The number of feature points of the sea clutter sample unit in the grid set to zero in the three-dimensional grid array in the above morphological closing operation processing and outlier removal operation is counted, and the false alarm rate of the current classifier can be obtained based on the number of feature points and the total number of sea clutter sample units.

[0111] The number of discarded grid points is related to the size of the structure element "se", so the false alarm rate of the detector can be controlled by controlling the volume size of the structure element "se"; the larger the volume of the structure element "se", the lower the false alarm rate; the smaller the volume, the greater the false alarm rate.

[0112] Embodiments:

[0113] In this embodiment, IPIX public real sea clutter data is used, which provides echo data of 14 range gates, 131072 pulse echo signals are collected for each range gate, two polarization channels H and V, in order to expand the training and test data set, a sliding window method is used to construct three feature data, the window length is 1024 pulses, 14 range gate features can be extracted each time the window is divided, and the sliding window step is 128 pulses.

[0114] In this embodiment, the maximum and minimum values of the three features of the sea clutter training sample set are calculated respectively, and are denoted as RAA max , RAA min , RDPH max , RDPH min , RVE max , and RVE min . Each set of maximum and minimum values is used as an end point to frame a segmentation region P ar , as shown in Figure 4 . Each dimension is uniformly divided into 200 parts to obtain a three-dimensional grid array with a size of 200x200x200, and each grid is set with an initial value num=0 for recording the number of sea clutter samples falling in the grid.

[0115] The structure element se is set, which is defined as a three-dimensional array with a size of 60x60x60 (the size is selected as 30% of the length of each dimension according to experience).

[0116] The method proposed in the present application is compared with the fast convex hull algorithm, and the detection rate under the same false alarm rate, the detector training time and the detection time are compared, and the comparison results are shown in Figures 8-10 . As can be seen from Figure 8 , under the same false alarm rate, the detection rate of the method of the present application is higher than that of the fast convex hull method, and the average performance is improved by 20.34%. As can be seen from Figure 9 , under the same false alarm rate, the average time of the detector training of the method of the present application is faster than that of the fast convex hull algorithm. The training time of the fast convex hull algorithm increases exponentially with the increase of the set false alarm rate, so the training time of the 10 groups of data will have a large change. The training time of the detector of the present application is mainly reflected in the fall point statistics, the closed operation and the connected region calculation, and these operation calculations are relatively simple, and the time consumption is mainly related to the number of sea clutter samples, so the overall difference is not large. As can be seen from Figure 10It can be found that the detection time of the method is obviously faster than that of the fast convex hull algorithm, and there is a difference of three orders of magnitude.

[0117] In summary, by whitening, the discrimination between sea clutter and target features can be increased, and the gap between them in the three-dimensional feature space is larger; with the fine fitting of the three-dimensional grid, the decision region is more consistent with the sea clutter distribution profile, and the detection performance is improved. Through the experiment of the measured data, it is found that the method proposed in the application is superior to the traditional fast convex hull algorithm in detection rate, detector training time and detection time, and has greater practical value.

[0118] The above examples are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for detecting slow small targets on the sea surface based on three-dimensional feature gridding, characterized in that: include: For pulse radar echo signals of known targets, the echo signals are sorted and arranged according to pulses to obtain a two-dimensional echo matrix; In the two-dimensional echo matrix, each row of data represents the sampling point of a pulse, and each column of data represents the sampling point of each range gate; Perform pre-processing on each pulse to obtain the time spectrum data of each range gate data; The time spectrum data of each range gate is converted into Doppler spectrum data through fast Fourier transform. Then, each range gate is used as the range gate to be whitened, and the range gates on both sides of it are used as reference range gates. The median and standard deviation of each Doppler frequency point in the range gate to be whitened are calculated. Then, the Doppler frequency point of the range gate to be whitened is subtracted from the corresponding Doppler frequency point median and divided by the standard deviation to obtain the whitened result. The whitened result is then inverse fast Fourier transformed to restore it to time spectrum data. The range gates other than the one where the target is located are divided into sliding windows. Each window of the range gate division is used as a sea clutter sample unit, and the units on both sides of each sea clutter sample unit are used as reference units. Three features are extracted from each sea clutter sample unit: RAA feature, RDPH feature, and RVE feature. After extracting the three features of all sea clutter sample units, the feature vector of each sea clutter sample unit is constructed using the three extracted features, thereby constructing a training sample set. Then, a three-dimensional feature space is constructed with the three features as coordinate axes, and the feature vector of the sea clutter sample unit is used as the three-dimensional coordinate index to plot the corresponding feature point position of the feature vector of each sea clutter sample unit in the three-dimensional feature space. Based on the extreme values ​​of the three features of all sea clutter sample units, a cuboid region is constructed in a three-dimensional coordinate system as a segmentation area, the segmentation area is divided into a three-dimensional grid array, and the number of feature points corresponding to the sea clutter sample units falling into each grid is counted to determine the attributes of each grid; a three-dimensional initial binary image with the same dimension as the three-dimensional grid array is established, and pixel points of the three-dimensional initial binary image are assigned values ​​according to the grid attributes; Performing a morphological closing operation on the three-dimensional initial binary image, thereby filling and connecting small holes and disconnected parts in the three-dimensional initial binary image and removing the outer boundary noise of the image to obtain an updated binary image; performing outlier removal on the updated binary image to obtain a removed binary image; and adjusting the properties of the grid in the three-dimensional grid array according to the removed binary image; In the actual detection phase, after obtaining a new radar echo signal, its pulse data is pre-processed and then whitened according to the range gate. All the range gates after whitening are divided into windows, and the range gates in each window are used as the detection unit. The three features of each detection unit are extracted to construct the feature vector of the detection unit. The feature vector is used as a three-dimensional coordinate index to map it to the feature point to be detected in the three-dimensional feature space, and detection is performed according to the position of the feature point to be detected: If the feature point to be detected falls outside the space of the three-dimensional grid array or falls within a grid of the three-dimensional grid array, but the attribute of the grid is a non-sea clutter grid, then the feature point to be detected is classified as a target; otherwise it is classified as sea clutter.

2. The method for detecting slow small targets on the sea surface based on three-dimensional feature gridding according to claim 1, characterized in that: The calculation of the median and standard deviation of each Doppler frequency point in the range gate to be whitened; Then, the whitened result is obtained by subtracting the corresponding Doppler frequency median from the range gate to be whitened and dividing it by the standard deviation, including: Where x(f) represents the Doppler spectrum data of the range gate to be whitened, x p (f) represents the Doppler spectrum data of the pth reference range gate, P represents the number of reference range gates, median(·) represents the calculated median, med and std represent the median sequence and standard deviation sequence of Doppler frequency point f, respectively; x norm (f) is the range gate Doppler spectrum data after whitening, for x norm (f) Perform inverse fast Fourier transform to restore the time spectrum data x norm (n).

3. The method for detecting slow small targets on the sea surface based on three-dimensional feature gridding according to claim 1, characterized in that: The extraction formulas of the RAA feature, RDPH feature and RVE feature are as follows: The RAA feature extraction formula is: in, and represent the mean energy of the sea clutter sample unit and the reference unit, respectively, x norm (n) and x pnorm (n) is the time spectrum data of the whitened sea clutter sample unit and the corresponding p-th reference unit, n is the time sampling point, N is the total number of time samples, and P′ represents the number of reference units; Where Peak represents the Doppler peak value of each sea clutter sample unit, x norm (f) represents the range gate Doppler spectrum data after whitening, f represents different Doppler frequencies, T is the pulse repetition time, and the corresponding Doppler channel is recorded as f max (x norm (f)); At this time, DPH can be expressed as: Where Δ = [-δ1, -δ2] ∪ [δ2, δ1], δ1 and δ2 represent the maximum Doppler bandwidths occupied by sea clutter and target, respectively, and #Δ means the number of Doppler frequency points remaining in the frequency point area of ​​the sea clutter sample after excluding the Doppler frequency points that may be occupied by the target; Finally, RDPH features are: Among them, x pnorm (f) represents the reference unit of Doppler spectrum after whitening; in, Therefore, the RVE characteristics are:

4. The method for detecting slow small targets on the sea surface based on three-dimensional feature gridding according to claim 1, characterized in that: Based on the extreme values ​​of the three characteristics of all sea clutter sample units, a rectangular area is constructed in a three-dimensional coordinate system as a segmentation area, the segmentation area is divided into a three-dimensional grid array, and the number of characteristic points corresponding to the sea clutter sample units falling into each grid is counted to determine the attributes of each grid, including: First, calculate the maximum and minimum values ​​of the three features of all sea clutter sample units and find the corresponding endpoint positions in the three coordinate axes. The rectangular area surrounded by the six planes corresponding to the six endpoints is used as the segmentation area, which is recorded as P ar ; This area is the minimum rectangular space parallel to the coordinate axis containing all the feature points corresponding to the sea clutter sample units; the segmentation area P ar Divide the three coordinate dimensions evenly, so that the segmentation area P ar Divide into a three-dimensional grid array, where each grid is set to an initial value num=0 to record the number of feature points corresponding to the sea clutter sample unit falling into the grid; Then, the feature points corresponding to the sea clutter sample units are traversed one by one. Every time a feature point falls into the grid, the grid num count is increased by 1.

5. The method for detecting slow small targets on the sea surface based on three-dimensional feature gridding according to claim 1, characterized in that: Creating a three-dimensional initial binary image having the same dimensions as the three-dimensional grid array, and assigning values ​​to pixels of the three-dimensional initial binary image according to the grid attributes, including: A three-dimensional initial binary image B with the same dimension size is constructed based on the three-dimensional grid array; the grids with grid num count greater than 0 are called sea clutter grids, and the corresponding pixels in the three-dimensional initial binary image B are recorded as 1; the rest are non-sea clutter grids, and the corresponding pixels are recorded as 0.

6. The method for detecting slow small targets on the sea surface based on three-dimensional feature gridding according to claim 1, characterized in that: The closing operation processing includes: The specific process of this operation is to set the structural element "se" to perform "dilation" and then "erosion" operations on the original image B. Among them, the structural element "se" is the basic unit of the "dilation" and "erosion" operations. The structural element "se" is made to slide in the initial binary image B, and the "dilation" and "erosion" definitions are defined to set 1 and 0 at the origin of "se". Here, the structural element "se" can also select a rectangular parallelepiped style based on the rectangular parallelepiped structure of the initial binary image B, and the center point of the structural element "se" is used as the origin position. The updated binary image is recorded as B C , B C The grid corresponding to the point with a median value of 1 in the three-dimensional grid array is recorded as a sea clutter grid, and the grid corresponding to the point with a median value of 0 is set to zero and recorded as a non-sea clutter grid.

7. The method for detecting slow small targets on the sea surface based on three-dimensional feature gridding according to claim 1, characterized in that: The process of outlier removal is as follows: For the updated binary image B C The connected areas with a value of 1 are counted, and the connected area with the largest number of grids in the corresponding three-dimensional grid array is called the main connected area. The remaining connected areas outside the main connected area are set to zero in the corresponding grids in the three-dimensional grid array as outliers and discarded; The binary image after removing outliers is recorded as J. The corresponding grids in the three-dimensional grid array of the points with a value of 1 in J are set as sea clutter grids, and the rest are non-sea clutter grids.

8. The method for detecting slow small targets on the sea surface based on three-dimensional feature gridding according to claim 1, characterized in that: In the statistical morphological closing operation and the outlier removal operation, the number of feature points corresponding to the sea clutter sample units in the grid of the three-dimensional grid array is set to zero, and the false alarm rate of the current classifier can be obtained by combining the number of feature points with the total number of sea clutter sample units.

9. A device for detecting slow small targets on the sea surface, comprising a processor, a memory, and a computer program stored in the memory; characterized in that: When the processor executes the computer program, it implements the method for detecting slow small targets on the sea surface based on three-dimensional feature gridding according to any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program; wherein: When the computer program is executed by a processor, the method for detecting slow small targets on the sea surface based on three-dimensional feature gridding according to any one of claims 1 to 8 is implemented.

Images