Radar sea surface small target detection method based on phase spectrogram feature convex hull, medium and equipment

Abstract

The invention discloses a radar sea surface small target detection method based on a phase spectrogram feature convex hull, a medium and equipment. The method comprises the following steps: calculating a phase difference sequence of an observation signal; defining a normalized histogram entropy of the differential sequence as a first detection statistic; performing undirected weighted quantization graph transformation on the sequence, defining a mean value of each quantization interval as a graph signal, and fusing the Laplace matrix of the graph and the graph signal by using a quadratic form to form a second detection statistic; defining a vertex probability vector of the graph, and calculating an entropy as a third detection statistic; and constructing a convex hull of a clutter sample feature set by using a rapid convex hull algorithm, and performing determinant calculation on a signal to be detected and a vertex of a triangular surface of the convex hull to judge whether the signal is a target or clutter. According to the method, the characteristics of the phase information are used, the graph theory is introduced, and the topological characteristics and the statistical characteristics are utilized, so that the difference between the target sequence and the clutter sequence can be better reflected, and the detection performance and the operation time are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal processing, and in particular relates to a radar sea surface small target detection method, medium and equipment based on phase spectrum feature convex hull. Background Art

[0002] Small target detection in the sea clutter background has become a hot and difficult topic in the field of radar detection. In recent years, with the rise of graph signal processing technology, detection algorithms based on graph domain signal processing have been widely used in radar small target detection and have become increasingly mature. The basic idea is to convert the signal from the traditional time domain, frequency domain or other transform domains to the graph domain through specific rules, and the graph domain features can often provide potential structural features of the signal that traditional statistical signal processing cannot provide, providing a new basis for improving the performance of the detection algorithm. However, the radar echo signal is a complex signal, but the existing radar small target detection algorithms based on graph domain processing mostly use the amplitude sequence of the echo signal, but ignore the fact that the phase still contains certain information, resulting in impaired detection performance; at the same time, most of them only consider a single graph topological feature, such as the maximum eigenvalue of the graph Laplacian matrix. Therefore, the sea surface target detection performance based on these algorithms is relatively one-sided, and the detection effect needs to be improved. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides a radar sea surface small target detection method, medium and device based on phase spectrum feature convex hull.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] In a first aspect, the present invention provides a radar sea surface small target detection method based on phase spectrum feature convex hull, comprising the following steps:

[0006] Step 1: Acquire the echo sequence received by the radar as an observation signal, extract the phase sequence of the observation signal, perform heavy phase difference calculation on the phase sequence, and obtain a corresponding phase difference sequence;

[0007] Step 2: Construct a normalized histogram of the phase difference sequence and calculate the entropy of the normalized histogram as the first detection statistic;

[0008] Step 3: Perform undirected weighted quantization graph transformation on the phase difference sequence, define the mean of each quantization interval as the graph signal, use the quadratic form to fuse the graph Laplace matrix and the graph signal, and use the obtained graph quadratic form feature as the second detection statistic;

[0009] Step 4: Construct the vertex probability vector of the quantized graph and calculate the entropy of the vertex probability vector as the third detection statistic;

[0010] Step 5: Construct the three detection statistics into a three-dimensional feature vector and present it in three-dimensional space. Use the fast convex hull method to calculate the convex hull area where the clutter is located. If the features constituted by the observed signal are outside the convex hull area, it is a target, otherwise it is clutter.

[0011] Optionally, in step 1, the calculation process of the difference sequence is as follows:

[0012] Step 1.1: Calculate the phase sequence p(n) of the observed signal x(n):

[0013] [p(n)=arg[x(n)];

[0014] Step 1.2: Perform N-fold phase difference on the phase sequence p(n) to obtain the N-fold phase difference sequence f N (n):

[0015]

[0016] Where n is the sample number of the signal.

[0017] Optionally, in step 2, the entropy of the normalized histogram is calculated as follows:

[0018] Step 2.1: Phase difference sequence f N (n) Drawing a normalized histogram with B quantities;

[0019] Step 2.2: Calculate the frequency of each bin in the histogram and construct a normalized histogram vector X = (x1, x2, ..., x i , ..., x B ) T ;

[0020] Step 2.3: Calculate the entropy of the normalized histogram vector X:

[0021]

[0022] Where ζ1 represents the first detection statistic.

[0023] Optionally, in step 3, the calculation process of the graph quadratic feature is as follows:

[0024] Step 3.1: Phase difference sequence f N (n) Perform maximum-minimum normalization to obtain the normalized sequence Uf N (n);

[0025] Step 3.2: Set the quantization level γ to Uf N(n) Perform equal-interval quantization to obtain a quantized sequence Q(k), and map Q(k) into an undirected weighted graph to obtain the corresponding adjacency matrix A;

[0026] Step 3.3: Sum each row of the adjacency matrix A to obtain the degree d of each vertex α for:

[0027]

[0028] In the formula, w αβ represents the element in the αth row and βth column of the adjacency matrix A;

[0029] The degree matrix D of the graph is defined as:

[0030] D = diag(d1, d2, ..., d α , ..., d γ );

[0031] Where diag(·) is a α A diagonal matrix of vectors;

[0032] The Laplace matrix L of the graph is obtained as:

[0033] L = DA;

[0034] Step 3.4: Define the graph signal u of the quantized graph i :

[0035]

[0036] Where k is Uf N (n) The number of sequences falling into the i-th quantization interval after quantization;

[0037] Then the interval mean vector U={u1,...,u i , ..., u γ} T ;

[0038] Step 3.5: Combine the graph Laplacian matrix L and the interval mean vector U through the quadratic form:

[0039] ζ2=U T LU;

[0040] Where ζ2 represents the second detection statistic.

[0041] Optionally, in step 4, the entropy of the vertex probability vector is calculated as follows:

[0042] Step 4.1: For Q(k), calculate its normalized histogram vector Q = (q1, q2, ..., q i, ..., q γ ) T ;

[0043] Step 4.2: Calculate the vertex probability entropy:

[0044]

[0045] Where ζ3 represents the third detection statistic.

[0046] Optionally, the specific process of step 5 is as follows:

[0047] Step 5.1: The three detection statistics ζ1, ζ2, ζ3 form a feature vector ζ = [ζ1, ζ2, ζ3];

[0048] Step 5.2: Set the false alarm probability to calculate the clutter feature vectors that need to be discarded in the clutter sample data feature set, and use the fast convex hull method to construct the convex hull area as the decision area;

[0049] Step 5.3: Based on the characteristic vector ζ of the signal to be measured and the decision region, determine whether the observed signal is a target or clutter through the determinant.

[0050] 7. The radar sea surface small target detection method based on phase spectrum feature convex hull according to claim 6, characterized in that: in step 5.2, the false alarm probability P is set fa , calculate the number of clutter sample feature vectors that need to be discarded in the clutter sample data feature set: N is the number of characteristic samples of the clutter sample dataset, which is the IPIX radar data provided by McMaster University in Canada;

[0051] The convex hull area constructed using the fast convex hull method is:

[0052]

[0053] Where Ω represents the convex hull area for judgment obtained after discarding some clutter feature vectors, which is composed of L triangular faces. Represents the three vertices of the th triangle in the decision area.

[0054] Optionally, in step 5.3, by the determinant Make a judgment: If It is considered that the eigenvector ζ is outside the decision area and the observed signal is the target; otherwise, it is clutter.

[0055] In a second aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the radar sea surface small target detection method based on the phase spectrum feature convex hull as described in the first aspect.

[0056] In a third aspect, the present invention provides an electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the radar sea surface small target detection method based on the phase spectrum feature convex hull as described in the first aspect is implemented.

[0057] The beneficial effects of the present invention are as follows: the present invention fully considers that phase is one of the important factors reflecting echo information, and converts the phase sequence into a graph, which can contribute to the detection of small targets on the sea surface from another perspective. At the same time, multiple features are extracted from multiple aspects such as the phase itself and the graph transformation, which effectively improves the performance of the existing detection algorithm based on the single graph feature of the amplitude sequence. The present invention not only uses the characteristics of phase information, but also makes full use of topological characteristics and statistical characteristics, which can better reflect the difference between the target sequence and the clutter sequence, and has important value in improving the performance of small target detection. Compared with the existing convex hull algorithm, the present invention takes into account both the detection performance and the operation time of the algorithm, and the algorithm efficiency is significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a flow chart of the radar sea surface small target detection method based on the convex hull of phase spectrum features.

[0059] Figure 2a and Figure 2b Schematic diagrams of the target unit and the clutter unit generating 5-fold phase difference sequences respectively.

[0060] Figure 3a , Figure 3b and Figure 3c They are the probability distribution histograms of the target and clutter of the three detection statistics.

[0061] Figure 4 It is the point set composed of the convex hull decision area formed by clutter in three-dimensional space and the target feature vector.

[0062] Figure 5 It is a schematic diagram comparing the detection probabilities of the method of the present invention, the traditional three-feature convex hull method, and the phase spectrum convex hull method. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0064] In one embodiment, the present invention proposes a radar sea surface small target detection method based on phase spectrum feature convex hull, the process of which is as follows: Figure 1 As shown, the following steps are included:

[0065] Step 1: Obtain the echo sequence received by the radar as the observation signal, extract the phase sequence of the observation signal, calculate the phase difference of the phase sequence, and obtain the corresponding phase difference sequence.

[0066] In this embodiment, the calculation process of the N-fold difference sequence is as follows:

[0067] Step 1.1: Calculate the phase sequence p(n) of the observed signal x(n):

[0068] p[(n) = arg[x(n)];

[0069] Step 1.2: Perform N-fold phase difference on the phase sequence p(n) to obtain the N-fold phase difference sequence f N (n):

[0070]

[0071] Where n represents the sample number of the signal.

[0072] Figure 2a and Figure 2b They are schematic diagrams of the target unit and the clutter unit generating five-fold phase difference sequences in this embodiment.

[0073] Step 2: Construct a normalized histogram of the phase difference sequence and calculate the entropy of the normalized histogram as the first detection statistic.

[0074] In this embodiment, the calculation process of the entropy of the normalized histogram is as follows:

[0075] Step 2.1: Phase difference sequence f N (n) Drawing a normalized histogram with B quantities;

[0076] Step 2.2: Calculate the frequency of each bin in the histogram and construct a normalized histogram vector X = (x1, x2, ..., x i , ..., x B ) T ;

[0077] Step 2.3: Calculate the entropy of the normalized histogram vector X:

[0078]

[0079] Where ζ1 represents the first detection statistic.

[0080] Step 3: Perform undirected weighted quantization graph transformation on the phase difference sequence, define the mean of each quantization interval as the graph signal, use the quadratic form to fuse the graph Laplace matrix and the graph signal, and obtain the graph quadratic form feature as the second detection statistic.

[0081] In this embodiment, the process of calculating the quadratic feature of the graph is as follows:

[0082] Step 3.1: Phase difference sequence f N (n) Perform maximum-minimum normalization to obtain the normalized sequence Uf N (n);

[0083] Step 3.2: Set the quantization level γ to Uf N (n) Perform equal-interval quantization to obtain a quantized sequence Q(k), and map Q(k) into an undirected weighted graph to obtain the corresponding adjacency matrix A;

[0084] Step 3.3: Sum each row of the adjacency matrix A to obtain the degree d of each vertex α for:

[0085]

[0086] In the formula, w αβ represents the element in the αth row and βth column of the adjacency matrix A;

[0087] The degree matrix D of the graph is defined as:

[0088] D = diag(d1, d2, ..., d α , ..., d γ );

[0089] Where diag(·) is a α A diagonal matrix of vectors;

[0090] The Laplace matrix L of the graph is obtained as:

[0091] L = DA;

[0092] Step 3.4: Define the graph signal of the quantized graph, namely:

[0093]

[0094] Where k is Uf N (n) The number of sequences falling into the i-th quantization interval after quantization;

[0095] Then the interval mean vector U={u1,…,u i ,…,u γ} T ;

[0096] Step 3.5: Combine the Laplace matrix and the interval mean vector by quadratic form, that is:

[0097] ζ2=U T LU;

[0098] Where ζ2 represents the second detection statistic.

[0099] Step 4: Construct the vertex probability vector of the quantized graph and calculate the entropy of the vertex probability vector as the third detection statistic.

[0100] In this embodiment, the entropy calculation process of the vertex probability vector is as follows:

[0101] Step 4.1: For Q(k), calculate its normalized histogram vector Q = (q1, q2, ..., q i , ..., q γ ) T ;

[0102] Step 4.2: Calculate the vertex probability entropy:

[0103]

[0104] Where ζ3 represents the third detection statistic.

[0105] Step 5: Construct the three detection statistics into a three-dimensional feature vector and present it in three-dimensional space. Use the fast convex hull method to calculate the convex hull area where the clutter is located. If the features constituted by the observed signal are outside the convex hull area, it is a target, otherwise it is clutter.

[0106] In this embodiment, the convex hull decision method used is as follows:

[0107] Step 5.1: Figure 3a , Figure 3b and Figure 3c They are respectively the probability distribution histograms of the target and clutter of the three detection statistics in this embodiment, and the three detection statistics ζ1, ζ2, ζ3 constitute a feature vector ζ=[ζ1, ζ2, ζ3].

[0108] Step 5.2: Set the false alarm probability to calculate the clutter feature vectors that need to be discarded in the clutter sample data feature set, and use the fast convex hull method to construct the convex hull area as the decision area;

[0109] In three-dimensional space, the convex hull can be viewed as a cube composed of a finite number of plane triangles:

[0110]

[0111] In the formula, CH represents the convex hull of the clutter area, M is the number of triangular planes, are three points arranged clockwise, and are also the three vertices of the mth triangle. triangle(a, b, c) represents a triangle formed by connecting the three points a, b and c in sequence.

[0112] Consider this convex hull as a convex body composed of M triangular pyramids, and the relationship between the four vertices is:

[0113]

[0114] Among them, v c represents the geometric center of the convex body and v c Located inside this convex body.

[0115] Then the volume of the three-dimensional convex body is:

[0116]

[0117] Training detection decision area in three-dimensional space: Based on the IPIX radar data provided by McMaster University in Canada, the feature set of clutter samples can be obtained as follows: The number of feature vectors is N. In order to meet the required misjudgment probability in target detection, some samples (i.e., feature vectors) need to be discarded. The number of discarded samples is P fa Indicates the false alarm probability. Finally, the decision area is obtained:

[0118]

[0119] Among them, Ω represents the convex hull for detection and judgment obtained after the set S is trained and some sample points are discarded, which is composed of L triangular faces. Represents the three vertices of the th triangle of the decision area.

[0120] The sample discarding process is as follows:

[0121] (1) Set a training set of clutter and calculate its convex hull CH(S), where the vertex set of the convex hull is {v1, v2, ..., v r};

[0122] (2) When the set {v1, v2, ..., v r} remove a vertex v j When , calculate the volume change of the convex hull;

[0123] (3) Find the vertex with the largest volume change, delete it, and update the feature set S;

[0124] (4) Return to step 1 for the next iteration, until it iterates K times; finally, the obtained convex hull Ω is used as the decision area.

[0125] Figure 4 It is a point set composed of the convex hull decision area formed by clutter in the three-dimensional space and the target feature vector in this embodiment.

[0126] Step 5.3: By the determinant: Make a judgment. It is considered that the eigenvector ζ is outside the decision area and the observed signal is the target; otherwise, it is clutter.

[0127] Figure 5 It is a comparison of the detection probability of the method of this embodiment and the traditional three-feature convex hull method and phase convex hull method.

[0128] (1) Data preprocessing: In order to make full use of the data, the clutter unit is segmented. Each segment has 512 points and slides 64 points each time (i.e., overlaps 64 points). Therefore, each clutter unit can obtain 2032 data segments.

[0129] (2) Feature calculation: For each preprocessed clutter unit segment, a normalized histogram of the differential phase sequence is constructed with B=128, and the entropy of the normalized histogram is calculated. At the same time, a weighted quantization graph transformation is performed on the differential phase sequence with a quantization level of γ=7, and the entropy of the graph quadratic form and vertex probability vector is calculated.

[0130] As can be seen from the figure, the detection probability of the method of this embodiment is similar to that of the existing method in most data sets, or has a certain improvement, and the operation time is less than that of the existing algorithm, which means that the detection performance of the method of this embodiment is more superior. Therefore, the method of this embodiment can be well applied to the radar sea surface small target detection under the sea clutter background, and the detection probability is greatly improved compared with the previous method.

[0131] In another embodiment, the present invention proposes a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the radar sea surface small target detection method based on the phase spectrum feature convex hull of the aforementioned embodiment.

[0132] In another embodiment, the present invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the radar sea surface small target detection method based on the phase spectrum feature convex hull of the aforementioned embodiment is implemented.

[0133] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0134] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0135] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A radar sea surface small target detection method based on phase spectrum feature convex hull, characterized in that: The steps include: Step 1: Acquire the echo sequence received by the radar as an observation signal, extract the phase sequence of the observation signal, perform heavy phase difference calculation on the phase sequence, and obtain a corresponding phase difference sequence; Step 2: Construct a normalized histogram of the phase difference sequence and calculate the entropy of the normalized histogram as the first detection statistic; Step 3: Perform undirected weighted quantization graph transformation on the phase difference sequence, define the mean of each quantization interval as the graph signal, use the quadratic form to fuse the graph Laplace matrix and the graph signal, and use the obtained graph quadratic form feature as the second detection statistic; Step 4: Construct the vertex probability vector of the quantized graph and calculate the entropy of the vertex probability vector as the third detection statistic; Step 5: Construct the three detection statistics into a three-dimensional feature vector and present it in three-dimensional space. Use the fast convex hull method to calculate the convex hull area where the clutter is located. If the features constituted by the observed signal are outside the convex hull area, it is a target, otherwise it is clutter.

2. The radar sea surface small target detection method based on phase spectrum feature convex hull according to claim 1, characterized in that: In step 1, the calculation process of the difference sequence is as follows: Step 1.1: Calculate the phase sequence p(n) of the observed signal x(n): [p(n)=arg[x(n)]; Step 1.2: Perform N-fold phase difference on the phase sequence p(n) to obtain the N-fold phase difference sequence f N (n): Where n is the sample number of the signal.

3. The radar sea surface small target detection method based on phase spectrum feature convex hull according to claim 1, characterized in that: In step 2, the entropy of the normalized histogram is calculated as follows: Step 2.1: Phase difference sequence f N (n) Drawing a normalized histogram with B quantities; Step 2.2: Calculate the frequency of each bin in the histogram and construct a normalized histogram vector X = (x1, x2, ..., x i , ..., x B ) T ; Step 2.3: Calculate the entropy of the normalized histogram vector X: Where ζ1 represents the first detection statistic.

4. The radar sea surface small target detection method based on phase spectrum feature convex hull according to claim 1, characterized in that: In step 3, the calculation process of the quadratic feature of the graph is as follows: Step 3.1: Phase difference sequence f N (n) Perform maximum-minimum normalization to obtain the normalized sequence Uf N (n); Step 3.2: Set the quantization level γ to Uf N (n) Perform equal-interval quantization to obtain a quantized sequence Q(k), and map Q(k) into an undirected weighted graph to obtain the corresponding adjacency matrix A; Step 3.3: Sum each row of the adjacency matrix A to obtain the degree d of each vertex α for: In the formula, w αβ represents the element in the αth row and βth column of the adjacency matrix A; The degree matrix D of the graph is defined as: D=diag(d1,d2,...,d α ,...,d γ ); Where diag(·) is a α A diagonal matrix of vectors; The Laplace matrix L of the graph is obtained as: L = DA; Step 3.4: Define the graph signal u of the quantized graph i : Where k is Uf N (n) The number of sequences falling into the i-th quantization interval after quantization; Then the interval mean vector U={u1,...,u i , ..., u γ } T ; Step 3.5: Combine the graph Laplacian matrix L and the interval mean vector U through the quadratic form: ζ2=U T LU; Where ζ2 represents the second detection statistic.

5. The radar sea surface small target detection method based on phase spectrum feature convex hull according to claim 4, characterized in that: In step 4, the entropy of the vertex probability vector is calculated as follows: Step 4.1: For Q(k), calculate its normalized histogram vector Q = (q1, q2, ..., q i , ..., q γ ) T ; Step 4.2: Calculate the vertex probability entropy: Where ζ3 represents the third detection statistic.

6. The radar sea surface small target detection method based on phase spectrum feature convex hull according to claim 1, characterized in that: The specific process of step 5 is as follows: Step 5.1: The three detection statistics ζ1, ζ2, ζ3 form a feature vector ζ = [ζ1, ζ2, ζ3]; Step 5.2: Set the false alarm probability to calculate the clutter feature vectors that need to be discarded in the clutter sample data feature set, and use the fast convex hull method to construct the convex hull area as the decision area; Step 5.3: Based on the characteristic vector ζ of the signal to be measured and the decision region, determine whether the observed signal is a target or clutter through the determinant.

7. The radar sea surface small target detection method based on phase spectrum feature convex hull according to claim 6, characterized in that: In step 5.2, set the false alarm probability P fa , calculate the number of clutter sample feature vectors that need to be discarded in the clutter sample data feature set: N is the number of characteristic samples of the clutter sample dataset, which is the IPIX radar data provided by McMaster University in Canada; The convex hull area constructed using the fast convex hull method is: Where Ω represents the convex hull area for judgment obtained after discarding some clutter feature vectors, which is composed of L triangular faces. Represents the three vertices of the th triangle in the decision area.

8. The radar sea surface small target detection method based on phase spectrum feature convex hull according to claim 7, characterized in that: In step 5.3, through the determinant Make a judgment: If It is considered that the eigenvector ζ is outside the decision area and the observed signal is the target; otherwise, it is clutter.

9. A computer-readable storage medium storing a computer program, characterized in that: The computer program enables the computer to execute the radar sea surface small target detection method based on the phase spectrum feature convex hull as described in any one of claims 1 to 8.

10. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the radar sea surface small target detection method based on the phase spectrum feature convex hull as described in any one of claims 1 to 8 is implemented.

Images