Sea surface target identification method applying time-frequency domain waveform entropy features

Abstract

The invention discloses a sea surface target identification method applying time-frequency domain waveform entropy features, and belongs to the technical field of radar signal processing. Comprising the following steps: step 1, caching target echo data; step 2, extracting target time-frequency domain waveform entropy features; and step 3, target identification is carried out by using the time-frequency domain waveform 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 discloses a sea surface target recognition method using time-frequency domain waveform entropy characteristics, belonging to the technical field of radar signal processing. Background Art

[0002] Radar target recognition is a key technology in radar signal processing. Rapid and accurate recognition of sea surface targets helps determine the function and purpose of the target, which is of great significance in both military and civilian fields. Common floating targets (small fishing boats, buoys, ice floes, etc.) and ship targets (large cruise ships, cargo ships, cargo roll-on / roll-off ships, etc.) have very different structures and physical sizes, resulting in different instantaneous amplitude scattering characteristics between floating targets and large ship targets. For high-resolution radars, one-dimensional range profile features are currently used to distinguish between the two. The one-dimensional range profile feature is an important target feature that reflects the distribution of strong scattering points along the distance and is the main basis for identifying ship targets and floating targets. However, the one-dimensional range profile feature is closely related to the observation angle, especially when the radar beam is irradiated along the side of the ship, the ship target and the floating target have similar one-dimensional range profile characteristics, which seriously affects the recognition performance. In addition, for low-resolution radars, since the one-dimensional range profile feature cannot be obtained, it becomes particularly difficult to perform target recognition without relying on auxiliary information.

[0003] Therefore, it is urgent to design a method that can extract more effective feature information and make good use of it 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-frequency domain waveform entropy features, aiming to identify sea surface targets when traditional target features are difficult to effectively perform target recognition.

[0005] The method for identifying sea surface targets using waveform entropy features in the time-frequency domain of the present invention is special in that it comprises 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: Extraction of target waveform entropy features in time-frequency domain

[0009] The target time domain echo sequence is segmented by sliding window extraction, and each data segment is subjected to time-frequency analysis. The waveform entropy values ​​of all time-frequency results are calculated along the Doppler channel, and then the minimum waveform entropy value is selected as the feature to finally form the target feature vector.

[0010] Step 3: Target recognition using waveform entropy features in the time-frequency domain

[0011] The classification algorithm is used to complete the recognition of two types of targets and the recognition performance is tested. The support vector machine method is used to divide the feature matrix composed of multiple targets 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. Finally, the eigenvalues ​​of the test set are input into the model for testing to obtain the target recognition results.

[0012] Preferably, the specific steps of step 1 are: the radar receives echo data in tracking mode, and after pulse compression, clutter suppression, target detection and tracking processing, a total of J targets are detected in the current scene, and the distance unit where the strongest scattering point of the target is located is selected, and its pulse echo data sequence is cached. For the jth target, it is represented as {z j (1),z j (2),...,z j (L)}, L represents the coherent pulse train length.

[0013] Preferably, the specific steps of step 2 are: a target echo sequence {z j (1),z j (2),...,z j (L)}, select a rectangular window function with a length of N to perform sliding window interception processing on it, and the rectangular window function is as follows:

[0014]

[0015] The window function slides forward W pulses each time it intercepts, and we get A subsequence of length N, denoted by x i ={z j [(i-1)W+1],z j [(i-1)W+2],…,z j [(i-1)W+N]}, where

[0016] For the i-th data segment x cut out from the above sliding window i , calculate the data segment and get a two-dimensional matrix, where the rows represent different Doppler channels and the columns represent different times. The specific formula is as follows:

[0017]

[0018] Among them, e is a natural constant, p(t-β) is a time window centered on β, t is a time variable, and f is a Doppler frequency channel;

[0019] Let y(t,f) = |stft(t,f)|, calculate the waveform entropy value of the sequence along the Doppler channel, take out the kth Doppler channel sequence in y(t,f), expressed as y(t,k) = {y(1,k), y(2,k), ..., y(N,k)}, the waveform entropy E k The expression is as follows:

[0020]

[0021] Where, ln represents the logarithmic operator with the natural logarithm e as the base, and the waveform entropy values ​​E1, E2…, E of all Doppler channels are calculated. k ,…E M ;

[0022] For the data segment x that needs to extract features i , the waveform entropy values ​​E1, E2…, E k ,…E M Select the minimum value σ i As the eigenvalue of this data segment, where σ i =min{E1,E2…,E k ,…E M};

[0023] Process each data segment of the current target echo sequence in turn to obtain its time-frequency domain waveform entropy feature vector Φ, which is expressed as:

[0024]

[0025] For all target echo sequences detected in step 1), the above feature extraction method is used to extract the corresponding feature vector for each target.

[0026] Compared with the prior art, the method for identifying sea surface targets using waveform entropy features in the time-frequency domain described in the technical solution has the following beneficial effects:

[0027] (1) The method proposed in the present invention 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, thereby solving the problem that it is difficult to achieve identification by relying on traditional target features.

[0028] (2) The waveform entropy is used to quantitatively measure the stationarity of the Doppler channel energy amplitude sequence. The physical meaning of the waveform entropy is combined with the mechanism of energy distribution of the two types of targets to select features, thus achieving the feature separability of the two types of samples.

[0029] (3) Under the existing data conditions, the average accuracy of the proposed method for identifying ship targets and floating targets is 93.98% in an observation time of 0.25s, which is 39.56% higher than that of the existing methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 : Time-frequency domain images of two types of targets;

[0031] Figure 2 : It is the waveform entropy value diagram of all Doppler channels corresponding to two types of targets;

[0032] Figure 3 : Sequence diagram of three types of Doppler channels. DETAILED DESCRIPTION

[0033] For better understanding and implementation, a specific embodiment is given below to explain in detail a method for recognizing sea surface targets using waveform entropy features in the time-frequency domain. The processing flow of the present invention is described in detail below:

[0034] 1) Target echo data cache

[0035] The radar receives echo data in tracking mode. After pulse compression, clutter suppression, target detection and tracking processing, a total of J targets are detected in the current scene. The distance unit where the strongest scattering point of the target is located is selected and its pulse echo data sequence is cached. For the jth target, it is expressed as {z j (1),z j (2),...,z j (L)}, L represents the coherent pulse train length.

[0036] 2) Target time-frequency domain waveform entropy feature extraction

[0037] For a target echo sequence z detected in step 1) j (m), select a rectangular window function with a length of N to perform sliding window interception processing on it. The rectangular window function is as follows

[0038]

[0039] The window function slides forward W pulses each time it intercepts, and we get A subsequence of length N, denoted by x i ={z j [(i-1)W+1],z j [(i-1)W+2],…,z j [(i-1)W+N]}, where

[0040] For the i-th data segment x cut out from the above sliding window iIn order to extract feature information, time-frequency analysis is first performed on it. The methods include fractional Fourier transform (FRFT), Wigner-Ville distribution (WVD), smoothed pseudo Wigner-Ville distribution (SPWVD), and wavelet transform (WT). For example, short-time Fourier transform has been widely used in time-frequency analysis of time-varying and non-stationary signals. This method can obtain information in both the time domain and the frequency domain of the signal, and can achieve coherent accumulation of target energy. The processing method of short-time Fourier transform is to intercept a certain length of time domain signal each time, and further perform Fourier transform on the intercepted time domain signal to obtain the spectrum diagram of this time period. By sliding the intercepted time domain signal over the entire time period, the set of each spectrum segment can be obtained. Therefore, short-time Fourier transform is a two-dimensional function of time and frequency. When calculating the data segment, a two-dimensional matrix can be obtained, where the rows represent different Doppler channels and the columns represent different times. The specific formula is as follows

[0041]

[0042] Wherein, e is a natural constant, p(t-β) is a time window centered on β, and the present invention uses a rectangular window to utilize the difference in energy amplitude without reducing the target main lobe energy, and the number of points for segmented Fourier transform is set to M. The time-frequency domain images of the two types of targets are shown in Figure 1 As shown, Figure 1 (a) for ship targets, Figure 1 (b) is a floating target. It can be seen that the energy of the floating target swings back and forth between different Doppler channels as time changes, and there is an obvious characteristic difference from the image of the ship target.

[0043] In order to calculate the waveform entropy value in the future, we need to take the absolute value of each data point, that is, let y(t,f) = |stft(t,f)|. Next, we calculate the waveform entropy value of the sequence along the Doppler channel. For example, we take the kth Doppler channel sequence in y(t,f), expressed as y(t,k) = {y(1,k), y(2,k), ..., y(N,k)}. The waveform entropy represents the stability of the signal, so the waveform entropy E k The expression is as follows

[0044]

[0045] Where ln represents the logarithmic operator with the natural logarithm e as the base. Thus, the waveform entropy values ​​E1, E2…, E of all Doppler channels can be calculated. k ,…E M . Figure 2 The waveform entropy values ​​of the two types of targets corresponding to all Doppler channels are shown. In this example, the energy of the ship target is concentrated in the 254th to 256th Doppler channels, and the energy of the floating target is concentrated in the 195th to 274th Doppler channels. It can be seen that when the Doppler channel contains target energy, the waveform entropy value will become smaller, and the minimum entropy value of the floating target is smaller than the minimum entropy value of the ship target.

[0046] The three typical Doppler channel sequences, including those without target energy, those containing ship target energy, and those containing floating target energy, are shown, such as Figure 3 As shown. The Doppler channel that does not contain target energy is mostly noise energy, with a small amplitude and little change over time. The sequence is very stable, and the corresponding waveform entropy value is the largest. In this example, N is 4096, and the theoretical maximum waveform entropy value is 8.32. The waveform entropy value of the Doppler channel that does not contain target energy is close to it; in the Doppler channel that contains ship target energy, the energy amplitude is large, and the distribution has a certain degree of fluctuation, so the waveform entropy value is small; in the Doppler channel that contains floating target energy, the target energy is sometimes absent, which is reflected in the sequence image. There will be many "spikes", which means that the probability distribution is well aggregated when calculating the waveform entropy, so the entropy value is the smallest.

[0047] Based on the above described two types of target waveform entropy characteristics in the time-frequency domain, for the data segment x that needs to extract features i , the waveform entropy values ​​E1, E2…, E k ,…E M Select the minimum value σ i As the eigenvalue of this data segment, where σ i =min{E1,E2…,E k ,…E M}.

[0048] Process each data segment of the current target echo sequence in turn to obtain its time-frequency domain waveform entropy feature vector Φ, which is expressed as

[0049]

[0050] For all target 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, if 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 A, then the time-frequency domain waveform entropy feature vectors Φ1 and Φ2 of the target can be obtained, and the number of columns of the vectors is A.

[0051] 3) Target recognition using waveform entropy features in the time-frequency domain

[0052] In step 2), 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.

[0053] Three sets of measured data were used for feature extraction, and their basic information is shown in Table 1. The echo data was intercepted by sliding window, 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.

[0054] Table 1 Introduction to measured data

[0055]

[0056]

[0057] During recognition, first, each feature value is assigned a label value, for example, a ship target is 1, a floating target is 2, and the feature matrix is ​​divided into a training set and a test set. The recognition process is explained using the support vector machine method as an example. The SVM classifier model is constructed using the training set, and the hyperparameters in the SVM are optimized using a loop iteration method, including the penalty factor c and the RBF kernel function parameter gamma. The range of both 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. Then the test set data is input into the model for recognition. The following four situations will appear in the test results:

[0058] ① True Positive (TP): The model correctly predicts a sample that is actually a ship target as a ship target;

[0059] ②False Negative (FN): The model mistakenly predicts samples that are actually ship targets as floating targets;

[0060] ③False Positive (FP): The model mistakenly predicts a sample that is actually a floating target as a ship target;

[0061] ④True Negative (TN): The model correctly predicts a sample that is actually a floating target as a floating target.

[0062] The recognition results are statistically analyzed and the results are shown in Table 2. For the sake of comparison, the recognition results using the time domain 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 93.98%. Under the existing data conditions, the recognition accuracy is improved by 39.56% compared with the existing methods.

[0063] Table 2 Statistics of target recognition results of waveform entropy features in time-frequency domain

[0064]

[0065] Table 3 Statistics of target recognition results based on waveform entropy features in the time domain

[0066]

[0067]

[0068] 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 waveform 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: Extraction of target waveform entropy features in time-frequency domain The target time domain echo sequence is segmented by sliding window extraction, and each data segment is subjected to time-frequency analysis. The waveform entropy values ​​of all time-frequency results are calculated along the Doppler channel, and then the minimum waveform entropy value is selected as the feature to finally form the target feature vector. Step 3: Target recognition using waveform entropy features in the time-frequency domain The classification algorithm is used to complete the recognition of two types of targets and the recognition performance is tested. The support vector machine method is used to divide the feature matrix composed of multiple targets 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. Finally, the eigenvalues ​​of the test set are input into the model for testing to obtain the target recognition results.

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

3. A method for identifying sea surface targets using time-frequency domain waveform entropy features according to claim 2, characterized in that The specific steps of step 2 are: j (1),z j (2),...,z j (L)}, select a rectangular window function with a length of N to perform sliding window interception processing on it, and the rectangular window function is as follows: The window function slides forward W pulses each time it intercepts, and we get A subsequence of length N, denoted by x i ={z j [(i-1)W+1],z j [(i-1)W+2],…,z j [(i-1)W+N]}, where For the i-th data segment x cut out from the above sliding window i , calculate the data segment and get a two-dimensional matrix, where the rows represent different Doppler channels and the columns represent different times. The specific formula is as follows: Among them, e is a natural constant, p(t-β) is a time window centered on β, t is a time variable, and f is a Doppler frequency channel; Let y(t,f) = |stft(t,f)|, calculate the waveform entropy value of the sequence along the Doppler channel, take out the kth Doppler channel sequence in y(t,f), expressed as y(t,k) = {y(1,k), y(2,k), ..., y(N,k)}, the waveform entropy E k The expression is as follows: Where, ln represents the logarithmic operator with the natural logarithm e as the base, and the waveform entropy values ​​E1, E2…, E of all Doppler channels are calculated. k ,…E M ; For the data segment x that needs to extract features i , the waveform entropy values ​​E1, E2…, E k ,…E M Select the minimum value σ i As the eigenvalue of this data segment, σ i =min{E1,E2…,E k ,…E M }; Process each data segment of the current target echo sequence in turn to obtain its time-frequency domain waveform entropy feature vector Φ, which is expressed as: For all target echo sequences detected in step 1), the above feature extraction method is used to extract the corresponding feature vector for each target.

Images