Sea surface target identification method applying time domain approximate entropy features

Abstract

The invention relates to a sea surface target identification method applying time domain approximate entropy features, and belongs to the technical field of radar signal processing and feature identification. Comprising the steps of 1, caching target echo data; 2, performing time domain analysis on echo data extracted from target time domain approximate entropy features; step 3, time domain approximate entropy feature optimization; and 4, target identification is carried out by using the time domain approximate entropy features. The objective of the invention is to identify a sea surface target under the condition that target identification is difficult to effectively perform through traditional target features.

Description

Technical Field

[0001] The invention relates to a sea surface target recognition method using time domain approximate entropy features, belonging to the technical field of radar signal processing and feature recognition. Background Art

[0002] There are many types of sea surface targets. The more common sea surface targets are floating targets (small fishing boats, buoys, floating ice, etc.) and ship targets (large cruise ships, cargo ships, cargo roll-on / roll-off ships, etc.). The detection and identification of these targets is very important in the fields of ocean monitoring, environmental protection and shipping safety. Especially in the case of floating targets in complex sea scenes, when using high-resolution range images, the observation angle, distance and time will affect the recognizability of the target. The lack of stable recognition features makes it difficult to identify floating targets, affecting the recognition performance.

[0003] Therefore, there is an urgent need to explore a more effective and stable feature information to break through the bottleneck of target recognition performance. Summary of the invention

[0004] Based on radar target detection, the present invention proposes a sea surface target recognition method using time domain approximate entropy features, aiming to identify sea surface targets when traditional target features are difficult to effectively perform target recognition.

[0005] A method for identifying sea surface targets using time domain approximate entropy features is characterized in that it includes the following steps:

[0006] Step 1: Target echo data buffer

[0007] The target detection is completed by using the echo data after pulse compression processing. According to the target distance unit information, the target time domain echo sequence is intercepted and cached;

[0008] Step 2: Extract target time domain approximate entropy features and analyze echo data in the time domain

[0009] The target time domain echo sequence is segmented by sliding window extraction, the data in each data segment is reconstructed, the corresponding time series approximate entropy is calculated, and then the obtained entropy value information is integrated to form a target feature matrix.

[0010] Step 3: Time domain approximate entropy feature optimization

[0011] For the two parameters in the feature extraction process, namely the dimension m of data reconstruction and the tolerance coefficient r in the threshold F, the optimal parameter range is determined according to the degree of influence of different parameters on the separability of the two types of targets, and feature optimization is achieved to improve the recognition performance;

[0012] Step 4: Use time domain approximate entropy features for target recognition

[0013] The support vector machine classification algorithm is used to complete the recognition of two types of targets, and the recognition performance is tested. The feature matrix composed of multiple targets is divided into a training set and a test set, and the model hyperparameter optimization is performed, including the penalty factor c and the RBF kernel function parameter gamma, to obtain better recognition results. Finally, the feature vector of the test set is input into the model for testing to obtain the target recognition result.

[0014] Preferably, the specific steps of step 1 are:

[0015] The radar receives echo data in tracking mode. After pulse compression, clutter suppression, target detection and tracking processing, a total of k targets are detected in the current scene. The range unit where the strongest scattering point of the target is located is selected, and its pulse echo data sequence z is cached. k (m), for the kth target, it is represented as {z k (1),z k (2),...,z k (L)}, L represents the coherent pulse train length.

[0016] Preferably, the specific steps of step 2 are:

[0017] For a target time domain echo sequence z detected in step 1) k (m), and segment it. The number of pulses in each segment is N. After segmentation, I subsequences can be obtained, which are expressed as {x1(n), x2(n), ..., x I (n)|1≤n≤N}, where

[0018] For the truncated i-th time domain sequence x i (n) Perform time domain approximate entropy feature extraction. The specific process steps are as follows:

[0019] ① For the truncated i-th time domain sequence x i (n), reconstruct the data with m as the dimension and divide it into k = n-m + 1 sequences, namely X i =[x(i),x(i+1),…,x(i+m-1)];

[0020] ② Calculate the distance d between each sequence and all k sequences ij =max|x i+k (t)-x j+k (t)|, where k = 0, 1, ..., m-1, forming a k × k d ij matrix;

[0021] ③Define the threshold F and count d ij The ratio of the number of values ​​less than F in each row of the matrix to the total number k = n-m+1 is expressed as Then according to k Calculate the logarithmic mean Where F = r·SD, r is the tolerance coefficient, SD is the standard deviation of the sequence;

[0022] ④ Increase the dimension from m to m+1 and repeat the process ① to ③;

[0023] ⑤ Calculate the approximate entropy ApEn(t) = Φ m (t)-Φ m+1 (t).

[0024] For all target time domain echo sequences detected in step 1), the above feature extraction method is used to extract the corresponding I feature vectors for each target, and these feature vectors are integrated to obtain the feature matrix of the target.

[0025] Preferably, the specific steps of step 3 are:

[0026] The Bhattacharyya distance, which can measure the similarity of two types of samples, is used as an indicator to quantitatively analyze the separability between the ship target sample space S1 and the floating target sample space S2 in the feature space. The appropriate parameter range is selected through analysis and optimization. The Bhattacharyya distance estimation method is as follows:

[0027] Calculate the mean vector μ and covariance matrix C of the two sample spaces

[0028] μ1=mean(S1),C1=cov(S1) (1)

[0029] μ2=mean(S2),C2=cov(S2) (2)

[0030] In the formula, mean and cov represent the mean and covariance operators respectively, and the Bhattacharyya distance d between the two sample spaces is B The estimated value of (S1,S2) is obtained by

[0031]

[0032] In the formula, C = (C1 + C2) / 2, d μ =(μ1-μ2) / chol(C), where det represents the determinant operator and chol represents the Cholesky decomposition operator.

[0033] Compared with the prior art, the method for identifying sea surface targets using time-domain approximate entropy features described in the technical solution has the following beneficial effects:

[0034] (1) The method proposed in the present invention uses the approximate entropy of the target as a feature, measures the similarity within the sequence by reconstructing the time series data, and then studies the similarity difference between two adjacent reconstructed data to obtain more target feature information, thus solving the problem of relying on traditional target features but being difficult to achieve recognition.

[0035] (2) Under the existing data conditions, the average accuracy of the proposed method for identifying ship targets and floating targets is 95.05% in an observation time of 0.5s, which is 40.63% higher than that of the existing methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 : It is a comparison diagram of the Bhattacharyya distance of two types of target samples in different dimensions m;

[0037] Figure 2 : It is a comparison diagram of Bhattacharyya distances of two types of target samples under different tolerance coefficients r;

[0038] Figure 3 : The scatter point envelope diagram of the two-dimensional approximate entropy features of the two types of targets before and after feature optimization. DETAILED DESCRIPTION

[0039] For better understanding and implementation, a specific embodiment is given below to explain in detail a method for sea surface target recognition using time domain approximate entropy features of the present invention. The processing flow of the present invention is described in detail below:

[0040] 1) Target echo data cache

[0041] The radar receives echo data in tracking mode. After pulse compression, clutter suppression, target detection and tracking processing, a total of k targets are detected in the current scene. The range unit where the strongest scattering point of the target is located is selected, and its pulse echo data sequence z is cached. k (m), for the kth target, it is represented as {z k (1),z k (2),…,z k (L)}, L represents the coherent pulse train length.

[0042] 2) Time domain approximate entropy feature extraction of target echo data

[0043] For a target time domain echo sequence z detected in step 1) k (m), and it is segmented and intercepted. The number of pulses in each segment of data is N. After interception, I subsequences can be obtained, which are expressed as {x1(n), x2(n),…, x I (n)|1≤n≤N}, where

[0044] For the truncated i-th time domain sequence xi (n) Perform time domain approximate entropy feature extraction. The specific process steps are as follows:

[0045] ① For the truncated i-th time domain sequence x i (n), reconstruct the data with m as the dimension and divide it into k = n-m + 1 sequences, namely X i =[x(i),x(i+1),…,x(i+m-1)];

[0046] ② Calculate the distance d between each sequence and all k sequences ij =max|x i+k (t)-x j+k (t)|, where k = 0, 1, ..., m-1, forming a k × k d ij matrix;

[0047] ③Define the threshold F and count d ij The ratio of the number of values ​​less than F in each row of the matrix to the total number k = n-m+1 is expressed as Then according to k Calculate the logarithmic mean Where F = r·SD, r is the tolerance coefficient, SD is the standard deviation of the sequence;

[0048] ④ Increase the dimension from m to m+1 and repeat the process ① to ③;

[0049] ⑤ Calculate the approximate entropy ApEn(t) = Φ m (t)-Φ m+1 (t).

[0050] For all target echo sequences detected in step 1), the above feature extraction method is used to extract the corresponding I feature vectors for each target, and these feature vectors are integrated to obtain the feature matrix of the target.

[0051] 3) Time domain approximate entropy feature optimization

[0052] In step 2), the two parameters in the feature extraction process, namely the dimension m of data reconstruction and the tolerance coefficient r in the threshold F, should be selected in combination with their impact on the separability of different targets to achieve feature optimization.

[0053] The Bhattacharyya distance (B-distance), which can measure the similarity of two types of samples, is used as an indicator to quantitatively analyze the separability between the ship target sample space S1 and the floating target sample space S2 in the feature space, and the appropriate parameter range is selected through analysis and optimization. The Bhattacharyya distance estimation method is as follows:

[0054] Calculate the mean vector μ and covariance matrix C of the two sample spaces

[0055] μ1=mean(S1),C1=cov(S1) (1)

[0056] μ2=mean(S2),C2=cov(S2) (2)

[0057] In the formula, mean and cov represent the mean and covariance operators respectively, and the Bhattacharyya distance d between the two sample spaces is B The estimated value of (S1,S2) is obtained by

[0058]

[0059] In the formula, C = (C1 + C2) / 2, d μ =(μ1-μ2) / chol(C) where det represents the determinant operator and chol represents the Cholesky decomposition operator.

[0060] The following uses measured data to study the effects of these two parameters on the separability of the two types of targets in order to determine the appropriate parameter range.

[0061] (1) Reconstruct the dimension m of the data space

[0062] The Bhattacharyya distance comparison of two types of target samples in different dimensions m is obtained through measured data. Figure 1 As shown in the figure, the larger the Bhattacharyya distance, the better the feature separability. Obviously, when the dimension m is selected as 2, the feature separability is generally good. As m increases, the feature separability becomes worse.

[0063] (2) Tolerance coefficient r in threshold F

[0064] The Bhattacharyya distance comparison of two types of target samples under different tolerance coefficients r is obtained through measured data. Figure 2 Obviously, when the tolerance coefficient r is selected as 0.2, the feature separability is generally good, and the smaller the tolerance coefficient r is, the worse the feature separability is generally.

[0065] After feature optimization, the appropriate parameter range is selected to increase the Bhattacharyya distance between the two types of samples. The extracted two-dimensional features are compared before and after optimization. Figure 3 As shown, Figure 3 (a) is the scatter envelope diagram of the two-dimensional approximate entropy features of the two types of targets before feature optimization. Figure 3 (b) is the scatter envelope diagram of the two-dimensional approximate entropy feature after feature optimization. After optimization, the overlap of the two types of target features is significantly reduced, and the feature separability is improved.

[0066] 4) Use time domain approximate entropy features for target recognition

[0067] In step 3), the feature vectors of all targets are obtained, and the classification algorithms in machine learning and deep learning are used to complete the recognition of two types of targets, and the recognition performance is tested. The classification algorithms include decision tree, support vector machine, naive Bayes, convex hull algorithm, and convolutional neural network.

[0068] Three sets of measured data were used for feature extraction, and their basic information is shown in Table 1. The echo data were segmented, with 1024 pulses in each segment, corresponding to a duration of about 0.5 s. A total of 128 feature vectors could be extracted from each target, and a total of 256 feature values ​​related to the two types of targets could be obtained from each set of data to form a feature matrix.

[0069] Table 1 Introduction to measured data

[0070]

[0071] The present invention adopts the recognition method of support vector machine. First, each feature vector label value is assigned during recognition, for example, the ship target is 1 and the floating target is 2, and the feature matrix is ​​divided into a training set and a test set. Then the training set is used to build the SVM classifier model, and the hyperparameters in the SVM are optimized by the cyclic iteration method, including the penalty factor c and the RBF kernel function parameter gamma. The range of the two parameters is from the minimum value 2 -8 Start, maximum value 2 8 End, the loop step is exponentially increased by one, and a five-fold cross-validation is performed on the model under each set of parameters. The parameters with the best recognition effect are selected to obtain the classifier model. Finally, the test set data is input into the model for recognition. The following four situations will appear in the test results:

[0072] ① True Positive (TP): The model correctly predicts a sample that is actually a ship target as a ship target;

[0073] ②False Negative (FN): The model mistakenly predicts samples that are actually ship targets as floating targets;

[0074] ③False Positive (FP): The model mistakenly predicts a sample that is actually a floating target as a ship target;

[0075] ④True Negative (TN): The model correctly predicts a sample that is actually a floating target as a floating target.

[0076] The recognition results are statistically analyzed and the results are shown in Table 2. For comparison purposes, the recognition results using time domain multi-waveform entropy as a feature are given in Table 3. It can be seen that the average recognition accuracy that the existing methods can achieve is 54.42%, while the average accuracy of the method proposed in the present invention is 95.05%. Under the existing data conditions, the recognition accuracy is improved by 40.63% compared with the existing methods.

[0077] Table 2 Statistics of target recognition results of time domain approximate entropy features

[0078]

[0079] The above calculation examples of the present invention are only used to explain the calculation model and calculation process of the present invention in detail, and are not intended to limit the implementation methods of the present invention. For ordinary technicians in the relevant field, other different forms of changes or modifications can be made based on the above description. It is impossible to list all the implementation methods here. All obvious changes or modifications derived from the technical solution of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for identifying sea surface targets using time domain approximate entropy features, which is special in that it includes the following steps: Step 1: Target echo data buffer The target detection is completed by using the echo data after pulse compression processing. According to the target distance unit information, the target time domain echo sequence is intercepted and cached; Step 2: Extract target time domain approximate entropy features and analyze echo data in time domain The target time domain echo sequence is segmented by sliding window extraction, the data in each data segment is reconstructed, the corresponding time series approximate entropy is calculated, and then the obtained entropy value information is integrated to form a target feature matrix; Step 3: Time domain approximate entropy feature optimization For the two parameters in the feature extraction process, namely the dimension m of data reconstruction and the tolerance coefficient r in the threshold F, the optimal parameter range is determined according to the degree of influence of different parameters on the separability of the two types of targets, and feature optimization is achieved to improve the recognition performance; Step 4: Use time domain approximate entropy features for target recognition The support vector machine classification algorithm is used to complete the recognition of two types of targets, and the recognition performance is tested. The feature matrix composed of multiple targets is divided into training set and test set, and the model hyperparameters are optimized, including the penalty factor c and the RBF kernel function parameter gamma, to obtain better recognition effect. Finally, the feature vector of the test set is input into the model for testing to obtain the target recognition result.

2. A method for identifying sea surface targets using time domain approximate entropy features according to claim 1, characterized in that The specific steps of step 1 are: The radar receives echo data in tracking mode. After pulse compression, clutter suppression, target detection and tracking processing, a total of k targets are detected in the current scene. The range unit where the strongest scattering point of the target is located is selected, and its pulse echo data sequence z is cached. k (m), for the kth target, it is represented as {z k (1),z k (2),…,z k (L)}, L represents the coherent pulse train length.

3. A method for identifying sea surface targets using time domain approximate entropy features according to claim 1, characterized in that Preferably, the specific steps of step 2 are: For a target time domain echo sequence z detected in step 1) k (m), and segment it. The number of pulses in each segment is N. After segmentation, I subsequences can be obtained, which are expressed as {x1(n), x2(n), ..., x I (n)|1≤n≤N}, where For the truncated i-th time domain sequence x i (n) Perform time domain approximate entropy feature extraction. The specific process steps are as follows: ① For the truncated i-th time domain sequence x i (n), reconstruct the data with m as the dimension and divide it into k = n-m + 1 sequences, namely X i =[x(i),x(i+1),…,x(i+m-1)]; ② Calculate the distance d between each sequence and all k sequences ij =max|x i+k (t)-x j+k (t)|, where k = 0, 1, ..., m-1, forming a k × k d ij matrix; ③Define the threshold F and count d ij The ratio of the number of values ​​less than F in each row of the matrix to the total number k = n-m+1 is expressed as Then according to k Calculate the logarithmic mean Where F = r·SD, r is the tolerance coefficient, SD is the standard deviation of the sequence; ④ Increase the dimension from m to m+1 and repeat the process ① to ③; ⑤ Calculate the approximate entropy ApEn(t) = Φ m (t)-Φ m+1 (t); For all target time domain echo sequences detected in step 1), the above feature extraction method is used to extract the corresponding I feature vectors for each target, and these feature vectors are integrated to obtain the feature matrix of the target.

4. A method for identifying sea surface targets using time domain approximate entropy features according to claim 1, characterized in that The specific steps of step 3 are: The Bhattacharyya distance, which can measure the similarity of two types of samples, is used as an indicator to quantitatively analyze the separability between the ship target sample space S1 and the floating target sample space S2 in the feature space. The appropriate parameter range is selected through analysis and optimization. The Bhattacharyya distance estimation method is as follows: Calculate the mean vector μ and covariance matrix C of the two sample spaces μ1=mean(S1),C1=cov(S1) (1) μ2=mean(S2),C2=cov(S2) (2) In the formula, mean and cov represent the mean and covariance operators respectively, and the Bhattacharyya distance d between the two sample spaces is B The estimated value of (S1,S2) is obtained by the following formula: In the formula, C = (C1 + C2) / 2, d μ =(μ1-μ2) / chol(C), where det represents the determinant operator and chol represents the Cholesky decomposition operator.

Images