Radar radiation source individual identification method based on large-scale streaming data processing

Abstract

The invention relates to a radar radiation source individual identification method based on large-scale streaming data processing, and the method comprises the steps: training a known radiation source intra-pulse characteristic waveform template library, calculating the similarity between waveforms in the template library, and determining the similarity threshold value of each radiation source individual; the Spark Streaming distributed computing platform analyzes and calculates the received binary system intermediate frequency data stream containing the radiation source intra-pulse information to obtain the instantaneous amplitude waveform and the instantaneous frequency waveform in the signal pulse; and taking two characteristics of instantaneous amplitude and instantaneous frequency as individual identification characteristics, performing similarity calculation on a waveform obtained by analysis and a waveform in a template library by using a dynamic time warping algorithm, and comparing a similarity calculation result with a similarity threshold to obtain an individual identification result. According to the method, the individual identification efficiency of the large-scale streaming radar radiation source data can be improved, and the parallel processing capability is enhanced.

Description

Technical Field

[0001] The invention belongs to the technical field of radar radiation source signal recognition. Background Art

[0002] With the development of passive detection technology of electronic signals, individual identification of radar emitter signals has become a research hotspot in the field of passive detection technology of electronic signals. Different from using the conventional characteristics of the five parameters of the signal and the type of intra-pulse characteristics as the classification standard of the emitter, extracting the subtle characteristics of the emitter signal can identify different individuals of the same type of emitter, thereby improving the passive detection and identification capabilities. The subtle characteristics of the signal usually refer to the unintentional modulation characteristics of the signal caused by the subtle differences in the internal components of the radar, which can uniquely identify the individual. The main application of subtle signal characteristics in engineering practice is to extract the unintentional modulation information attached to the waveform of the signal intra-pulse characteristics (including instantaneous amplitude, instantaneous frequency and instantaneous phase, etc.).

[0003] Most existing studies use the method of calculating the similarity of subtle signal features to achieve individual identification of radiation sources. However, the selection of the similarity threshold is overly dependent on human factors such as experts, and the threshold setting is relatively arbitrary. At the same time, most existing studies complete the similarity calculation process on a single computer, without considering the large and high-speed radiation source data flow brought about by the collaborative detection of multiple sensors and multiple types of equipment. This situation puts higher requirements on the single-machine computing and storage capabilities of radiation source individual identification. Furthermore, existing studies rarely discuss how to ensure the uniqueness of individual identification results after similarity calculation. Summary of the invention

[0004] In order to solve the problem of insufficient individual recognition capability of multi-sensor radar emitters in the prior art, the object of the present invention is to provide a radar emitter individual recognition method based on large-scale streaming data processing.

[0005] The technical solution to achieve the purpose of the present invention is:

[0006] First, the known radiation source signal pulse characteristic waveform template library is trained, and the similarity between each pulse characteristic waveform in the template library is calculated to determine the similarity threshold of each radiation source individual, and generate the pulse characteristic waveform threshold sequence. Compared with the threshold determined by engineering experience, the present invention utilizes the classification ability of the known template library; the SparkStreaming distributed computing platform receives the radiation source binary intermediate frequency data stream through the socket, and the data stream is divided into discrete data segments based on the size of the data stream time window. Each data segment is parsed according to the intermediate frequency binary data XML parsing description to obtain the values ​​of each sampling point of the instantaneous amplitude waveform and instantaneous frequency waveform of each signal; the instantaneous amplitude and instantaneous frequency are used as individual identification features, and the dynamic time warping algorithm is used to calculate the similarity of the two waveforms obtained by parsing and the waveforms in the template library; the pulse characteristic waveform threshold sequence is searched, and the similarity calculation result is compared with the threshold of the corresponding radiation source individual in the threshold sequence. If it is greater than or equal to the threshold, it is considered that the two radiation sources are similar. If there are multiple results that meet the threshold condition, the two radiation sources with the largest similarity value are taken as the final individual identification result. The above process uses Map / Reduce operations to achieve parallelism in the calculation process, which meets the requirements of real-time identification of individual radiation sources in the case of large-scale, high-speed radiation source data streams. Due to the uniqueness of individual radiation source identification, for a pair of radiation sources after identification, the radiation source individual with better distinguishability in the characteristic waveform in the template library is selected to update the template library.

[0007] Furthermore, the determination of the individual similarity threshold of the radiation source utilizes the similarity between the individual characteristic waveforms of known radiation sources in the knowledge base to determine whether the characteristic waveform to be identified is an individual in the knowledge base. Its similarity threshold must be greater than or equal to the maximum value of the calculated similarities between other characteristic waveforms in the knowledge base and the individual. The more characteristic waveforms in the knowledge base, the higher the confidence of the similarity threshold.

[0008] Furthermore, the individual similarity calculation process of the radiation sources is performed on a distributed computing platform.

[0009] Furthermore, the maximum solution after calculating the individual similarity of the radiation source and the template in the template library are optimally selected, and the template library is updated using the radiation source characteristic waveform with the smallest sum of similarities with other waveforms in the template library.

[0010] The present invention can improve the efficiency of individual identification of large-scale streaming radar radiation source data and enhance the parallel processing capability. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 Flowchart of an embodiment of the present invention.

[0012] Figure 2 Map / Reduce operations. DETAILED DESCRIPTION

[0013] The technical solution of the present invention is further explained in conjunction with the accompanying drawings. Figure 1 shown.

[0014] The radiation source individual identification process first trains the known radiation source template library, calculates the similarity between the template waveforms in the template library, determines the similarity threshold of each template waveform, and finally generates the waveform threshold sequence Q. Let T = {t 1 , t 2 , t 3 ...t i} is the waveform set in the radiation source template library, and any signal waveform w and t to be identified j The similarity calculated by the dynamic time warping algorithm (DynamicTimeWarping) is defined as:

[0015] Sim=αSim amp +βSim freq (1)

[0016] Among them, Sim amp =DTW(w amp , t j ) is the similarity after calculating the instantaneous amplitude characteristic waveform using the dynamic time warping algorithm, Sim freq =DTW(w freq , t j ) is the similarity of the instantaneous frequency waveform calculated using the dynamic time warping algorithm, α, β are the weighted coefficients of the instantaneous amplitude and instantaneous frequency, and α+β=1. The instantaneous frequency is better than the instantaneous amplitude as an individual identification feature, so the value of β needs to be greater than α. For any waveform t in the set T j , and the threshold value of similarity judgment is but Expression 2 needs to be satisfied:

[0017]

[0018] Where k≠j. In order to reduce the amount of calculation and speed up the preprocessing process, all template waveform data are normalized, and the threshold sequence generated is It can be seen from Formula 2 that the threshold determination method in the preprocessing process proposed by the present invention utilizes the classification capability that can be provided by the known waveform information of the template library. Compared with determining a threshold based on experience, this process can scientifically guide the individual classification and identification process of radiation sources.

[0019] The analysis of intermediate frequency binary data utilizes the ability of XML format description files to analyze and describe binary files. The intermediate frequency binary data format consists of four parameters: address, data packet length, amplitude value of each sampling point, frequency value of each sampling point, and time.

[0020] Map / Reduce operations are mainly used for distributed parallel computing. Using Map / Reduce operations to perform radiation source individual identification operations meets the following two requirements: the time window size determines the size of the elastic distributed data set (RDD) allocated to each computing node in the distributed computing, but the size of the elastic distributed data set does not affect the process of allocating binary intermediate frequency data of the radiation source at each node; any two signal data in the intermediate frequency data set are completely independent, and there is no necessary connection when performing dynamic time warping calculations. The dynamic time warping calculation of any signal waveform does not require the use of the data of another signal waveform, so Map / Reduce operations can be introduced for parallel computing. The Map operation in the Map / Reduce operation mainly includes three parts, such as Figure 2 As shown in ①, ② and ③. ① The SparkStreaming distributed computing platform receives the values ​​of each sampling point of the instantaneous amplitude waveform and instantaneous frequency waveform obtained after parsing the established XML format description file, obtains the data format that can be processed by SparkStreaming through data format conversion, and obtains the initialized RDD; ② Before similarity calculation, the elastic distribution data set in the data stream is distributed to each computing node according to the node performance (mainly memory usage). In addition, the similarity threshold sequence Q of each radiation source in the template library is also sent to each node. At the same time, each computing node saves a copy of the waveform information of the radiation source template library. In each node, the data of each sampling point of the instantaneous amplitude and instantaneous frequency are normalized and the template waveform in the template library is used to calculate the similarity using expression (1) to obtain the RDD after the similarity calculation; ③ Search the waveform threshold sequence Q, filter out the waveforms that do not meet the requirements, and output the waveform information with a similarity greater than or equal to the threshold. The filtered RDD is expressed as a three-dimensional vector <w,t j , Sim j > indicates that the Reduce operation in the Map / Reduce operation is as follows Figure 2 As shown in ④, it mainly uses the vector obtained after the Map operation <w,t j , Sim j >Set, set parameter t j As the key of Reduce operation, for the same t j , find out the number of Sim in the set j The vector with the largest value, that is, the vector with the largest value that matches the template waveform t j The most similar radiation source individual is taken as the final individual identification result.

[0021] Due to the uniqueness of individual identification, it is necessary to update the known radiation source template library and rebuild the template waveform. After the Map / Reduce operation, the solution with the highest similarity is obtained. Due to a series of factors such as errors, the similarity cannot be 100%. Therefore, the maximum solution of the similarity calculation and the template in the known template library need to be processed to obtain an optimal result as the template for individual identification of this type of radiation source. Assume that the signal waveform w and t j After being similarly calculated by the dynamic time warping algorithm, w replaces t in the template library. j Expression 3 needs to be satisfied:

[0022]

[0023] Where n≠j, m≠j. From formula 3, we can see that w replaces t j It is necessary to satisfy that the sum of the similarities between w and other template waveforms in the template library is less than the original template t j , that is, w is greater than t j There is better differentiation.

Claims

1. A radar emitter individual identification method based on large-scale streaming data processing, characterized in that: Firstly, the template library of the intra-pulse characteristic waveforms of known radiation sources is trained. By calculating the similarity between the intra-pulse characteristic waveforms in the template library, the similarity threshold of each radiation source individual is determined, and the threshold sequence of the intra-pulse characteristic waveforms is generated. The Spark Streaming distributed computing platform receives the binary intermediate frequency data stream of the radiation source through the socket. The data stream is divided into discrete data segments based on the size of the data stream time window. Each data segment is parsed according to the XML parsing description of the intermediate frequency data to obtain the values ​​of each sampling point of the instantaneous amplitude waveform and the instantaneous frequency waveform in the signal pulse. The instantaneous amplitude and instantaneous frequency are used as identification features, and the dynamic time warping algorithm is used to calculate the similarity between the two waveforms obtained by parsing and the waveforms in the template library. Search the characteristic waveform threshold sequence, compare the similarity calculation result with the threshold of the corresponding radiation source individual in the threshold sequence, and if it is greater than or equal to the threshold, the two radiation sources are considered similar. If there are multiple results that meet the threshold condition, the two radiation sources with the largest similarity value are taken as the final individual identification result. The above process uses Map / Reduce operations to achieve the parallelism of the calculation process; for a pair of radiation sources after individual identification, select the radiation source individual with better distinguishability in the characteristic waveform in the template library to update the template library.

2. According to claim 1, a radar emitter individual identification method based on large-scale streaming data processing is characterized in that: The determination of the individual similarity threshold of the radiation source utilizes the similarity between the individual characteristic waveforms of known radiation sources in the knowledge base to determine whether the characteristic waveform to be identified is an individual in the knowledge base. Its similarity threshold must be greater than or equal to the maximum value of the calculated similarities between other characteristic waveforms in the knowledge base and the individual. The more characteristic waveforms in the knowledge base, the higher the confidence of the similarity threshold.

3. The method for identifying individual radar emitters based on large-scale streaming data processing according to claim 1 is characterized in that: The radiation source individual similarity calculation process is performed on a distributed computing platform.

4. The method for identifying radar emitters based on large-scale streaming data processing according to claim 1 is characterized in that: The maximum solution after calculating the individual similarity of the radiation source and the template in the template library are optimally selected, and the template library is updated using the radiation source characteristic waveform with the smallest sum of similarities with other waveforms in the template library.

Images