A method for identifying target categories based on dictionary matching degree

Abstract

The present invention relates to a method for identifying target categories based on dictionary matching degree, belonging to the field of underwater target recognition and classification. First, target dictionaries of the AUV model, torpedo model, and mine model are respectively constructed. The received echo of the unknown target is sparsely represented in each dictionary to obtain sparse vectors, and then the original signal is reconstructed from the obtained sparse vectors. Using the concept of dictionary matching degree, the unknown signal is matched with the dictionary, and it is determined that if the matching degree with a certain dictionary is high, it belongs to that type of target. The method of the present invention can achieve good classification effects, without the need for a classifier for classification, reducing the influence of the design of the classifier and incomplete feature extraction on target classification, and making up for the deficiencies of the original algorithm.

Description

Technical Field

[0001] The present invention belongs to the field of underwater target recognition and classification, and relates to a method for judging the target type from the recoverable degree of the received echo in each dictionary by using the idea of target dictionaries, sparse reconstruction and dictionary matching degree of different models. Background Art

[0002] Traditional classifiers work well when dealing with data with large differences, but when the data becomes complex, higher requirements are imposed on the feature extraction method, and the target feature expression ability and generalization of the classifier decline. Due to the complexity and instability of underwater acoustic signals, traditional classification methods cannot handle data better and cannot accurately classify targets. In order to improve the efficiency and accuracy of target recognition and classification using target echo signals under low signal-to-noise ratio, this invention is very necessary.

[0003] The invention patent discloses "a target category recognition algorithm based on sparse features". It uses the sparse representation of the target echo signal in the dictionary domain as the target feature and adopts a vector machine classification algorithm for target recognition and classification. This method uses an SVM classifier for classification. The SVM algorithm is difficult to implement for large-scale training samples, is sensitive to missing data, and is sensitive to the selection of parameters and kernel functions. Therefore, the recognition method is greatly affected by the classifier. Summary of the Invention

[0004] Technical Problems to be Solved

[0005] In order to avoid the deficiencies of the prior art, the present invention provides a target category recognition method based on dictionary matching degree.

[0006] Technical Solution

[0007] A target category recognition method based on dictionary matching degree, which uses the idea of target dictionaries, sparse reconstruction and dictionary matching degree to judge from the recoverable degree of the received echo in each dictionary. If the matching degree is high, it is determined as this type of target. The characteristics are as follows:

[0008] Step 1: Rotate the target with its geometric center as the origin, and divide the rotation angle of the target into N s angles, denoted as For each angle θ i (i = 1, 2,..., N s ) there exists an echo signal x i (i = 1, 2,..., N s ). Construct a dictionary D1 about the angle, which is a matrix of (L*n)×N s The i-th (1 ≤ i ≤ N s ) column of the matrix is

[0009] ψi = [x1(1, θ i ), …, x1(n, θ i ), x2(1, θ i ), …, x2(n, θ i ), …, x L (1, θ i ), …, x L (n, θ i )] T (1 - 1) Then the dictionary of the target model angles That is

[0010]

[0011] Similarly, the angle dictionaries of different target models can be obtained;

[0012] Step 2: Sparsely represent the underwater unknown target echoes in each dictionary to obtain sparse vectors; When solving the sparse representation α of the unknown target echo signal x in different target dictionaries, that is, solving

[0013]

[0014] where γ represents the weight coefficient, which changes with noise, and α is the sparse representation vector; Reconstruct the original signal using the sparse vector α under different target dictionaries respectively;

[0015] Step 3: Reconstruct the unknown target signal through the constructed known model target dictionary, and judge which type of target the unknown target is based on the reconstruction effect; In order to better characterize and compare the reconstruction performance of the target dictionary, the concept of dictionary matching degree is used. Let x be the unknown target signal, be the signal restored by a certain target dictionary, then the matching degree a of the unknown target to this target dictionary is denoted as:

[0016]

[0017] It can be known from the definition of the dictionary matching degree that the value of a is a positive number, and the range is between 0 and 1. If the reconstruction effect of the target dictionary on the signal is good, the value of the matching degree tends to 1. If the reconstruction effect of the target dictionary on the signal is poor, the value of the matching degree tends to 0; Select the one with the larger value, then the dictionary matching degree between the unknown target and this known target is the highest, and it is determined that the unknown target is this type of target.

[0018] A computer system, characterized in that it includes: one or more processors, a computer-readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the above method.

[0019] A computer-readable storage medium, characterized in that it stores computer-executable instructions, which are used to implement the above-mentioned method when executed.

[0020] Beneficial effects

[0021] A target category recognition method based on dictionary matching degree provided by the present invention abandons the disadvantage that information loss may be caused by extracting features in the past, is not affected by the classifier, constructs target dictionaries of different models respectively, and uses the ideas of target dictionary, sparse reconstruction and dictionary matching degree to judge from the recoverable degree of the target echo signal received by the sonar array in each dictionary. If the matching degree is high, it is determined as this type of target. The invention can achieve good classification effects, does not require a classifier for classification, reduces the influence of the design of the classifier and incomplete feature extraction on target classification, and makes up for the deficiencies of the original algorithm.

[0022] The target recognition and classification method provided by the present invention can establish dictionaries of different models for different environments and different interferences, so as to meet the requirement of reconstructing target signals and calculating dictionary matching degrees in various environments, thereby achieving the purpose of recognition and classification, and is no longer affected by the selection of classifier parameters. Verified by experiments, this method has achieved very good effects. Under the same signal-to-noise ratio conditions, the target recognition and classification accuracy is higher than that of using a classifier for recognition and classification. Brief description of the drawings

[0023] The drawings are only for the purpose of showing specific embodiments, and are not considered as a limitation of the present invention. Throughout the drawings, the same reference signs denote the same components.

[0024] Figure 1 It is a flow chart of the invention.

[0025] Figure 2 They are the target dictionaries and coherence characteristic diagrams of the AUV model, torpedo model and mine model established in the invention: (a) Target dictionary of AUV model; (b) Target dictionary of torpedo model; (c) Target dictionary of mine model; (d) Three-dimensional coherence characteristic diagram of the target dictionary of AUV model; (e) Three-dimensional coherence characteristic diagram of the target dictionary of torpedo model; (f) Three-dimensional coherence characteristic diagram of the target dictionary of mine model; (g) Contour map of coherence characteristics of the target dictionary of AUV model; (h) Contour map of coherence characteristics of the target dictionary of torpedo model; (i) Contour map of coherence characteristics of the target dictionary of mine model; (j) Side view of coherence characteristics of the target dictionary of AUV model; (k) Side view of coherence characteristics of the target dictionary of torpedo model; (l) Side view of coherence characteristics of the target dictionary of mine model.

[0026] Figure 3Simulation results of the matching degree of unknown signals based on different model dictionaries at different signal-to-noise ratios in this invention.

[0027] Figure 4 Recognition rate of target categories based on the matching degree of different transmitted signal dictionaries in this invention. Detailed implementation manners

[0028] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0029] The basic idea of the present invention is to first construct target dictionaries for the AUV model, torpedo model, and mine model respectively, sparsely represent the received unknown target echo in each dictionary to obtain a sparse vector, then reconstruct the original signal from the obtained sparse vector, and use the concept of dictionary matching degree to match the unknown signal with the dictionary. If the matching degree with a certain dictionary is high, it is the corresponding type of target.

[0030] Step 1: First, construct the AUV model target dictionary D1, the torpedo model target dictionary D2, and the mine model target dictionary D3.

[0031] Rotate the AUV model with its geometric center as the origin, rotate 360° in steps of 1°, and each rotation angle direction corresponds to an element in the dictionary, that is, construct the dictionary D1 about the angle, which is an (L*n)×N s matrix, and the i-th (1≤i≤N s ) column of the matrix is

[0032] ψ i =[x1(1,θ i ),…,x1(n,θ i ),x2(1,θ i ),…,x2(n,θ i ),…,x L (1,θ i ),…,x L (n,θ i )] T (1-1) Then the dictionary of the AUV model angle That is

[0033]

[0034] Similarly, the torpedo model target dictionary D2 and the mine model target dictionary D3 are as follows

[0035]

[0036]

[0037] Step 2: Sparsely represent the underwater unknown target echo in the AUV model target dictionary D1, the torpedo model target dictionary D2, and the mine model target dictionary D3 respectively, that is, three sparse vectors α, β, and γ can be obtained. When solving the sparse representations α, β, and γ of the unknown target echo signal x in different target dictionaries, that is, solving

[0038]

[0039] where γ represents the weight coefficient, which changes with noise, and α is the sparse representation vector.

[0040] l1 norm is a convex function and this function is non-differentiable. Therefore, it can obtain a very sparse and globally optimal solution. Then the unknown target signal can be reconstructed through the sparse vectors α, β, and γ, that is

[0041] Step 3: Calculate the matching degree of the target dictionary. Denote the matching degree between the unknown signal and the AUV model target dictionary D1 as a, the matching degree with the torpedo model target dictionary D2 as b, and the matching degree with the mine model target dictionary D3 as c. Select the one with the largest value among a, b, and c. Then the unknown target has the highest dictionary matching degree with this known target, and it is determined that the unknown target is this type of target.

[0042] Suppose the unknown signal is the AUV model echo. Using the three constructed target dictionaries, test the signal reconstruction ability of the three dictionaries at different signal-to-noise ratios. The targets are selected as the AUV model, the torpedo model, and the mine model. The transmitted signal uses a bionic signal. Test the matching degree between the unknown signal and the three known dictionaries at -3dB, 0dB, 3dB, and 6dB respectively. The results are shown in Table 1 and Figure 3 as follows.

[0043] Table 1 presents the results of this invention Figure 3 in tabular form, making the results clearer

[0044]

[0045] From Table 1 and Figure 3It can be seen that as the signal-to-noise ratio (SNR) increases, the matching degree between the unknown signal and the target dictionary becomes higher. The AUV model dictionary increases from 0.48 at SNR = -3 dB to 0.71 at 6 dB, the torpedo model dictionary increases from 0.37 at SNR = -3 dB to 0.52 at 6 dB, and the mine model dictionary increases from 0.19 at SNR = -3 dB to 0.35 at 6 dB. This shows that the higher the SNR, the better the signal reconstruction effect. By comparing the matching degrees of the three different targets at the same SNR, it can be seen that at the same SNR, the matching degree of the AUV model dictionary is higher than that of the other two dictionaries. At this time, the unknown target is determined to be an AUV target, indicating that the algorithm can correctly classify the target and verifies its effectiveness.

[0046] Step 4: To analyze the anti-noise performance and stability of the algorithm, a target recognition study based on the matching degree of different transmitted signal dictionaries was carried out, and the recognition rate was tested at different SNRs. The target recognition results based on different transmitted signals at different SNRs are shown in Table 2 and Figure 4 as follows.

[0047] Table 2 presents the results of this invention Figure 4 in tabular form.

[0048]

[0049] From Table 2 and Figure 4 it can be seen that when the transmitted signal is an LFM signal of 60 kHz - 65 kHz, as the SNR increases, the recognition rate rises and reaches 73.89% at SNR = -3 dB. By comparing the two LFM signals, as the bandwidth increases, the recognition rate reaches 77.69% at SNR = -3 dB. The recognition rate of the first harmonic of the bionic signal reaches 89.63% at SNR = -3 dB, and the recognition rate of the second harmonic of the bionic signal reaches 91.94% at -3 dB. The recognition effect of the bionic signal is the best, reaching 97.22% at SNR = -3 dB. Therefore, the feasibility of the multi-target category recognition method based on the dictionary matching degree is verified, and when the transmitted signal is a bionic signal, the target category recognition effect based on the dictionary matching degree is the best, with the recognition rate reaching 97.22% at SNR = -3 dB.

[0050] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present invention.

Claims

1. A method for identifying a target category based on dictionary matching degree, which utilizes the ideas of target dictionary, sparse reconstruction, and dictionary matching degree to judge from the recoverable degree of the received echo in each dictionary. If the matching degree is high, it is determined as this type of target; it is characterized in that The steps are as follows: Step 1: Rotate the target with its geometric center as the origin, and divide the rotation angle of the target into N s angles, denoted as each angle θ i (i = 1, 2, …, N s ). There exists an echo signal x i (i = 1, 2, …, N s ). Construct a dictionary D1 about the angle, which is a matrix of (L*n)×N s . The i-th (1 ≤ i ≤ N s ) column of the matrix is ψ i = [x1(1, θ i ), …, x1(n, θ i ), x2(1, θ i ), …, x2(n, θ i ), …, x L (1, θ i ), …, x L (n, θ i )] T (1 - 1) Dictionary of the target model angle That is Similarly, the angle dictionaries of different target models can be obtained; Step 2: Sparsely represent the underwater unknown target echo in each dictionary to obtain sparse vectors; When solving the sparse representation α of the unknown target echo signal x in different target dictionaries, that is, solving where γ represents the weight coefficient, which changes with noise, and α is the sparse representation vector; Reconstruct the original signal from the sparse vector α under different target dictionaries respectively; Step 3: Reconstruct the unknown target signal through the constructed known model target dictionary, and judge which type of target the unknown target is according to the reconstruction effect; in order to better characterize and compare the reconstruction performance of the target dictionary, the concept of dictionary matching degree is used. Let x be the unknown target signal, and y be the signal restored by a certain target dictionary. Then the matching degree a of the unknown target to this target dictionary is denoted as: According to the definition of dictionary matching degree, the value of a is a positive number, and the range is between 0 and 1. If the target dictionary has a good signal reconstruction effect, the value of the matching degree tends to 1. If the target dictionary has a poor signal reconstruction effect, the value of the matching degree tends to 0; Select the one with a larger value, then the dictionary matching degree between the unknown target and the known target is the highest, and the unknown target is determined to be this type of target.

2. A computer system, characterized in that Including: One or more processors, a computer-readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in claim 1.

3. A computer-readable storage medium, characterized in that Stored with computer-executable instructions, the instructions are used to implement the method described in claim 1 when executed.

Images