Sea surface target identification method applying time-frequency domain multi-scale permutation entropy features

Abstract

The invention relates to a sea surface target recognition method applying time-frequency domain multi-scale permutation entropy features, and belongs to the technical field of radar signal processing and feature recognition. The method comprises the following steps: step 1, caching target echo data; 2, carrying out time-frequency domain analysis on the target echo data; step 3, target time-frequency domain multi-scale permutation entropy feature extraction; and step 4, target identification is carried out by using the time-frequency domain multi-scale permutation entropy features. The objective of the invention is to identify a sea surface target under the condition that target identification is difficult to be effectively performed by traditional target features.

Description

Technical Field

[0001] The invention relates to a sea surface target recognition method using multi-scale arrangement entropy features in time-frequency domains, and belongs to the technical field of radar signal processing and feature recognition. Background Art

[0002] Rapid and accurate identification of sea surface targets helps determine the function and purpose of the target, which is of great significance in both military and civilian fields. The more common sea surface targets are floating targets (small fishing boats, buoys, ice floes, etc.) and ship targets (large cruise ships, cargo ships, cargo roll-on / roll-off ships, etc.), which have very different structures and physical sizes. Currently, high-resolution range image features are mainly used. However, due to the complex and changeable sea surface scenes, different observation angles, different radar working conditions and other conditions, the target features are unstable, and there is a problem that the features between targets are difficult to distinguish due to high similarity. The identification of floating targets becomes particularly difficult, which seriously affects the recognition performance.

[0003] For high-resolution radar, the one-dimensional high-resolution range profile feature is currently used for distinction. This feature is closely related to the observation angle. Especially when the radar beam is irradiated along the side of the ship, the two types of targets have extremely high similarity, which seriously affects the recognition performance. For low-resolution radar, since it is impossible to obtain HRRP features, it becomes more difficult to identify targets without relying on auxiliary information. Therefore, in order to break through the bottleneck of target recognition performance, mining and utilizing more effective feature information is an effective solution. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the above-mentioned prior art and propose a sea surface target recognition method based on multi-scale arrangement entropy features in time-frequency domain, aiming to identify sea surface targets when traditional target features are difficult to effectively perform target recognition.

[0005] The present invention provides a method for identifying sea surface targets using multi-scale permutation entropy features in the time-frequency domain, which is special in that it comprises 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: Time-frequency domain analysis of target echo data The target time domain echo sequence is segmented by sliding window extraction, and each data segment is subjected to time-frequency analysis; Step 3: Extract target multi-scale permutation entropy features in time-frequency domain After the time-frequency domain analysis, coarse-graining processing is performed at different time scales along the Doppler channel, and the permutation entropy value of the corresponding scale is calculated. Then the minimum multi-scale permutation entropy value is selected as the feature, and finally the target feature vector is formed. Step 4: Use multi-scale permutation entropy features in the time-frequency domain for target recognition The support vector machine classification algorithm is used to complete the recognition of two types of targets and test the recognition performance. The feature matrix composed of multiple targets is divided into training set and test set, and the hyperparameters of the model 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.

[0006] Preferably, the specific steps of step 1 are: The radar receives echo data in tracking mode and detects a total of k For each target, select the distance unit where the strongest scattering point of the target is located and cache its pulse echo data sequence , for the k target, which is represented as , L Indicates the coherent pulse train length.

[0007] Preferably, the specific steps of step 2 are: For a target time domain echo sequence detected in step 1) , choose the length M The rectangular window function performs sliding window interception processing on it. as follows: (1); The window function slides forward each time it intercepts W The number of pulses is The length is M subsequence of , which can be expressed as ,in ; After the sliding window is intercepted i Data segment , short-time Fourier transform method is used for time-frequency domain analysis to obtain information in two dimensions of signal time domain and frequency domain, and to achieve the accumulation of target energy. Fourier transform is performed on the intercepted time domain signal each time to obtain the spectrum diagram of the signal in the time period. The time domain signal is intercepted by sliding over the entire time period to obtain the collection of each spectrum. The result is a two-dimensional function of time-frequency. For this two-dimensional matrix, the rows of the matrix represent different Doppler channels, and the columns of the matrix represent different times. The short-time Fourier transform formula is as follows: (2); in, For The time window centered on the rectangle window is used, and the number of points for the segmented Fourier transform is N ,The energy of floating targets swings back and forth between different Doppler channels as ,time changes, while the energy of ship targets changes less over ,time, and there are obvious characteristic differences in the time-frequency images of ,the two types of targets.

[0008] Preferably, the specific steps of step 3) extracting target time-frequency domain multi-scale permutation entropy features are: First, the data segment The absolute value of each data point in the time-frequency result is , and then calculate the multi-scale permutation entropy value of the time-frequency sequence along the Doppler channel, and take out The k Doppler channel sequence, expressed as In order to obtain more feature information at different scales, the time-frequency series calculated along the Doppler channel is coarse-grained. , the process is as follows: (3); in, is the coarse-grained sequence, s is the coarse-grained scale, is a positive integer. After coarse-graining, the sequence length is Q ,in ; Define the phase space reconstruction matrix Y for: (4); in, m is the embedding dimension, is the number of delayed pulses, K = Q -( m -1), phase space reconstruction matrix Y The dimension of K × m , for the matrix Y Arrange the elements of each row in ascending order to obtain a set of symbol sequences consisting of the index of the position of each element in the row vector , expressed as ,Right now: (5); m There are a total of different symbol sequences mapped into the dimensional phase space m ! species, where ! represents the factorial operator. By counting the frequency of each permutation order in the reconstructed sequence, we can get the probability sequence , then the multi-scale permutation entropy The calculation formula is: (6); The multi-scale permutation entropy is normalized, that is: (7); After normalization, its value range is [0, 1]; According to the above process, we get In scale s The permutation entropy value under the condition is obtained, and the multi-scale permutation entropy value corresponding to all Doppler channels of the time-frequency result of the data segment is obtained. , and select the minimum value As the characteristic value of this data segment, , process each data segment of the current target echo sequence in turn, and obtain the characteristic vector of its multi-scale permutation entropy in the time-frequency domain , expressed as: (8); For all target time domain echo sequences detected in step 1), the above feature extraction method is used to extract the corresponding feature vector for each target.

[0009] Preferably, the specific steps of step 4) using the time-frequency domain multi-scale permutation entropy features to perform target recognition are: The support vector machine classification algorithm is used. This algorithm distinguishes different categories by finding an optimal plane in the feature space. The selection criterion of this hyperplane is to maximize the interval between two categories, that is, the minimum distance from the data point to the hyperplane. The feature matrix composed of multiple targets is divided into a training set and a test set, and the model hyperparameters are optimized. Finally, the feature vector of the test set is input into the model for testing to obtain the target recognition result.

[0010] Compared with the prior art, the sea surface target recognition method based on multi-scale arrangement entropy features in the time-frequency domain described in the technical solution has the following beneficial effects: (1) The method proposed in this patent uses the time-frequency analysis method, comprehensively considers the time domain and frequency domain information, and makes full use of the differences in the energy distribution characteristics of the two types of targets in the time-frequency domain, thus solving the problem that traditional target features are difficult to identify.

[0011] (2) The permutation entropy value is used to measure the mutation degree and randomness of the Doppler channel energy amplitude sequence. The features are selected based on the energy distribution mechanism of the two types of targets to achieve the separability of the features of the two types of samples. The coarse-grained processing at different time scales is used to optimize the multi-scale permutation entropy features and further improve the recognition performance.

[0012] (3) Under the existing data conditions, the average accuracy of the method proposed in this patent for identifying ship targets and floating targets is 97.05% within an observation time of 0.25s, which is 3.08% and 5.65% higher than the two existing methods respectively. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 ] are time-frequency domain images of two types of targets in the embodiment of the present invention. DETAILED DESCRIPTION

[0014] For better understanding and implementation, a specific embodiment is given below to explain in detail a method for recognizing sea surface targets based on multi-scale permutation entropy features in the time-frequency domain of the present invention. The method for recognizing sea surface targets of the invention is as follows: 1) Target echo data cache The radar receives echo data in tracking mode and detects a total of k For each target, select the distance unit where the strongest scattering point of the target is located and cache its pulse echo data sequence , for the k target, which is represented as , L Indicates the coherent pulse train length.

[0015] 2) Time-frequency domain analysis of target echo data For a target time domain echo sequence detected in step 1) , choose the length M The rectangular window function performs sliding window interception processing on it. as follows: (1); The window function slides forward each time it intercepts W The number of pulses is The length is M subsequence of , expressed as ,in .

[0016] After the sliding window is intercepted i Data segment Time-frequency domain analysis is performed, and the main analysis methods include fractional Fourier transform, Wigner-Ville distribution, smoothed pseudo-Wigner-Ville distribution, wavelet transform, etc. This patent uses the short-time Fourier transform method for time-frequency analysis. This method can obtain information in both the time domain and frequency domain dimensions of the signal, and can achieve coherent accumulation of target energy. It has been widely used in the time-frequency analysis of time-varying and non-stationary signals. First, a Fourier transform is performed on the time domain signal of a certain length each time, and the spectrum of the signal in this time period is obtained. Then, the time domain signal is slid over the entire time period, and finally a collection of each spectrum is obtained. The result is a two-dimensional function of time-frequency. For this two-dimensional matrix, the rows of the matrix represent different Doppler channels, and the columns of the matrix represent different times. The short-time Fourier transform formula is as follows: (2); in, For In order to utilize the difference in energy amplitude and avoid the reduction of target main lobe energy, this patent uses a rectangular window, and the number of points when performing Fourier transform in segments is N The time-frequency domain images of the two types of targets are as follows: Figure 1 As shown, Figure 1 (a) is the time-frequency domain image of the ship target. Figure 1 (b) is the time-frequency domain image of the floating target. The energy of the floating target swings back and forth between different Doppler channels as time changes, while the energy of the ship target changes less over time and is relatively stable. There are obvious characteristic differences between the time-frequency images of the two types of targets.

[0017] 3) Target time-frequency domain multi-scale permutation entropy feature extraction First, the data segment The absolute value of each data point in the time-frequency result is , and then calculate the multi-scale permutation entropy value of the time-frequency sequence along the Doppler channel, and take out The k Doppler channel sequence, expressed as In order to obtain more feature information at different scales, the time-frequency series calculated along the Doppler channel is coarse-grained. , the process is as follows: (3); in, is the coarse-grained sequence, s is the coarse-grained scale, is a positive integer. After coarse-graining, the sequence length is Q ,in .

[0018] Define the phase space reconstruction matrix Y for: (4); in, m is the embedding dimension, is the number of delayed pulses, K = Q -( m -1), phase space reconstruction matrix Y The dimension of K × m . For the matrix Y By arranging the elements of each row in ascending order, we can get a set of symbol sequences consisting of the index of the position of each element in the row vector. , expressed as ,Right now: (5); m There are a total of different symbol sequences mapped into the dimensional phase space m ! species, where ! represents the factorial operator. By counting the frequency of each permutation order in the reconstructed sequence, we can get the probability sequence , then the multi-scale permutation entropy The calculation formula is: (6); The multi-scale permutation entropy is normalized, that is: (7); After normalization, its value range is [0, 1].

[0019] According to the above process, we can get In scale s The permutation entropy value under this condition can be used to calculate the multi-scale permutation entropy value corresponding to all Doppler channels of the time-frequency result of the data segment. , and select the minimum value As the characteristic value of this data segment, . Process each data segment of the current target time domain echo sequence in turn to obtain the feature vector of its multi-scale permutation entropy in the time-frequency domain , expressed as: (8); For all target time domain echo sequences detected in step 1), the above feature extraction method can be used to extract the corresponding feature vector for each target. For example, two targets are detected in the echo data, and the number of data segments obtained after the sliding window interception of each target echo data is M , we can get the multi-scale permutation entropy feature vector of the target in time-frequency domain , , the number of columns of the vector is M .

[0020] 4) Use multi-scale permutation entropy features in the time-frequency domain for target recognition In step 3), the feature vectors of all targets were obtained, and the classification algorithms in machine learning and deep learning were used to complete the recognition of two types of targets, and the recognition performance was tested. The classification algorithms include decision tree, support vector machine, naive Bayes, convex hull algorithm, and convolutional neural network.

[0021] Three sets of measured data were used for feature extraction, and their basic information is shown in Table 1. The echo data was segmented, with 4096 pulses in each segment, and 512 pulses sliding forward each time, corresponding to a duration of about 0.25 s. A total of 249 eigenvalues ​​can be extracted from each target, and a total of 498 eigenvalues ​​related to the two types of targets can be obtained from each set of data to form a feature matrix.

[0022] Table 1 Introduction to measured data ; 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 RBF kernel function parameters gamma , both parameters range from a minimum of 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: ① True Positive (TP): The model correctly predicts a sample that is actually a ship target as a ship target; ② False Negative (FN): The model mistakenly predicts a sample that is actually a ship target as a floating target; ③ False Positive (FP): The model mistakenly predicts a sample that is actually a floating target as a ship target; ④ True Negative (TN): The model correctly predicts a sample that is actually a floating target as a floating target.

[0023] Table 2 Statistics of target recognition results using multi-scale permutation entropy features in time-frequency domain ; Table 3 Statistics of target recognition results using waveform entropy features in the time-frequency domain ; Table 4 Statistics of target recognition results using entropy features proposed in reference [2] ; The recognition results are statistically analyzed, and the results are shown in Table 2. For comparison purposes, Table 3 shows the recognition results of multi-scale permutation entropy in the time-frequency domain as a feature, and Table 4 shows the recognition results of the entropy feature proposed in Reference [2] (Reference [2]: Zhao Yue, Chen Zhichun, Jiu Bo, et al. A method for extracting narrowband radar aircraft target classification features based on time-frequency analysis [J]. Journal of Electronics and Information Technology. 2017, 39(9): 2225-2231). It can be seen that the average recognition accuracy that the existing methods can achieve is 93.97% (using a single feature) and 91.4% (using four features), while the average recognition accuracy of the present invention using a single feature is 97.05%. Under the existing data conditions, the recognition accuracy is improved by 3.08% and 5.65% compared with the existing methods.

Claims

1. A method for sea surface target recognition using multi-scale permutation entropy features in the time-frequency domain, characterized in that The following steps are involved: 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: Time-frequency domain analysis of target echo data The target time domain echo sequence is segmented by sliding window extraction, and each data segment is subjected to time-frequency analysis; Step 3: Extract target multi-scale permutation entropy features in time-frequency domain After the time-frequency domain analysis, coarse-graining processing is performed at different time scales along the Doppler channel, and the permutation entropy value of the corresponding scale is calculated. Then the minimum multi-scale permutation entropy value is selected as the feature, and finally the target feature vector is formed. Step 4: Use multi-scale permutation entropy features in the time-frequency domain for target recognition The support vector machine classification algorithm is used to complete the recognition of two types of targets and test the recognition performance. The feature matrix composed of multiple targets is divided into training set and test set, and the hyperparameters of the model 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 multi-scale permutation entropy features in time-frequency domain according to claim 1, characterized in that The specific steps of step 1 are: The radar receives echo data in tracking mode and detects a total of k For each target, select the distance unit where the strongest scattering point of the target is located and cache its pulse echo data sequence , for the k target, which is represented as , L Indicates the coherent pulse train length.

3. A method for identifying sea surface targets using multi-scale permutation entropy features in time-frequency domain according to claim 2, characterized in that The specific steps of step 2 are: For a target time domain echo sequence detected in step 1) , choose the length M The rectangular window function performs sliding window interception processing on it. as follows: (1)； The window function slides forward each time it intercepts W The number of pulses is The length is M subsequence of , expressed as ,in ; After the sliding window is intercepted i Data segment , short-time Fourier transform method is used for time-frequency domain analysis to obtain information in two dimensions of signal time domain and frequency domain, and to achieve the accumulation of target energy. Fourier transform is performed on the intercepted time domain signal each time to obtain the spectrum diagram of the signal in the time period. The time domain signal is intercepted by sliding over the entire time period to obtain the collection of each spectrum. The result is a two-dimensional function of time-frequency. For this two-dimensional matrix, the rows of the matrix represent different Doppler channels, and the columns of the matrix represent different times. The short-time Fourier transform formula is as follows: (2)； in, For The time window centered on the rectangle window is used, and the number of points for the segmented Fourier transform is N ,The energy of floating targets swings back and forth between different Doppler channels as ,time changes, while the energy of ship targets changes less over ,time, and there are obvious characteristic differences in the time-frequency images of ,the two types of targets.

4. A method for identifying sea surface targets using multi-scale permutation entropy features in time-frequency domain according to claim 1, characterized in that The specific steps of step 3) target time-frequency domain multi-scale permutation entropy feature extraction are: First, the data segment The absolute value of each data point in the time-frequency result is , and then calculate the multi-scale permutation entropy of the time-frequency sequence along the Doppler channel, and take out The k Doppler channel sequence, expressed as In order to obtain more feature information at different scales, the time-frequency series calculated along the Doppler channel is coarse-grained. , the process is as follows: (3)； in, is the coarse-grained sequence, s is the coarse-grained scale, is a positive integer. After coarse-graining, the sequence length is Q ,in ; Define the phase space reconstruction matrix Y for: (4)； in, m is the embedding dimension, is the number of delayed pulses, K = Q -( m -1), phase space reconstruction matrix Y The dimension of K × m , for the matrix Y Arrange the elements of each row in ascending order to obtain a set of symbol sequences consisting of the index of the position of each element in the row vector , expressed as ,Right now: (5)； m There are a total of different symbol sequences mapped into the dimensional phase space m ! species, where ! represents the factorial operator. Count the frequency of each permutation order in the reconstructed sequence to get the probability sequence , then the multi-scale permutation entropy The calculation formula is: (6)； The multi-scale permutation entropy is normalized, that is: (7)； After normalization, its value range is [0, 1]; According to the above process, we get In scale s The permutation entropy value under the condition is obtained, and the multi-scale permutation entropy value corresponding to all Doppler channels of the time-frequency result of the data segment is obtained. , and select the minimum value As the characteristic value of this data segment, , process each data segment of the current target time domain echo sequence in turn, and obtain the characteristic vector of its multi-scale permutation entropy in the time-frequency domain , expressed as: (8)； For all target time domain echo sequences detected in step 1, the above feature extraction method is used to extract the corresponding feature vector for each target.

Images