A method for generating radar target feature templates that incorporate marine environment

Abstract

The invention discloses a radar target feature template generation method integrating an ocean environment, and provides a radar target feature template generation method, which comprises: firstly, collecting multi-source environment and radar target data, and cleaning and processing to form an original material library; then constructing a spatiotemporal feature grid model of the ocean environment, extracting and expressing the features of different ocean environments; using the radar target data to extract the one-dimensional image and modulation spectrum features of the radar target; further using the position and time information of the target, performing spatiotemporal correlation matching between the target feature information and the ocean environment feature network, incorporating the environment feature information into the target feature information, and forming an ocean target basic feature template; finally, performing statistical compilation, verification and deduplication and other feature processing on the basic feature template to form a refined target feature template library.

Description

Technical Field

[0001] The invention is applied to the field of radar target recognition. Technical Background

[0002] Marine environmental information is diverse, involving multi-dimensional and multi-type marine elements such as air (clouds, visibility, weather and wind, temperature, humidity, pressure, etc.), sea surface (wind, waves, currents, tides and temperature, humidity, pressure, precipitation, radiation, typhoons, tornadoes, sea ice, storm surges, etc.), water bodies (temperature, salinity, depth, current, density, pressure, color, turbidity, light and internal waves, jump layers, circulation, sound speed, noise, etc.), seabed (topography, landforms, bottom texture, strata, etc.) and gravity and magnetic fields (gravity, geomagnetism, electromagnetic), which have a huge impact on radar target detection. Affected by the marine environment, the characteristic information of radar targets under different marine environmental conditions may be different. Therefore, it is necessary to collect and compile environmental and target data to form a unified associated data model, extract features based on environmental and radar target data, and further generalize and abstract to obtain target feature templates. Modeling and processing marine environmental data and target data to achieve high-quality storage, processing and utilization of environmental targets are the basis and common problems of marine target information processing.

[0003] At present, for surface or low-altitude targets, using the characteristics of target one-dimensional range image and micro-Doppler to identify targets is an important means of radar target identification. The literature "Chen, VC Micro-Doppler Effect in Radar [M]. Electronic Industry Press, 2013: 92-108" studies the target micro-Doppler characteristics, but the effectiveness and complexity of the target micro-motion characteristics under complex marine environments need to be finely analyzed. "Radar Target Recognition Method Based on Multi-view One-dimensional Image Decision Layer Fusion, CN106597406B, 2016-12-02" discloses a radar target recognition method under multi-view one-dimensional images, but the template and recognition have not yet taken into account the influence of the marine environment, and the extraction of one-dimensional image features under complex marine environments also lacks research and solutions. The one-dimensional range image and micro-Doppler features extracted from the target radar data are further compiled and processed to form template data with different attitude angles, which has become an important part of the target recognition process. The marine environment will cause certain interference to target recognition, and the method for generating radar target feature templates needs to be further improved. At the same time, due to the diversity, variability and complexity of the marine environment, it brings challenges to the fusion of environmental models and target feature models. Therefore, it is a difficult problem in the existing technology to incorporate environmental factors into the process of target feature template generation and target recognition, reduce the impact of different environments on target recognition, and provide data support for improving recognition capabilities. Summary of the invention

[0004] Aiming at the problems brought by the complex ocean environment to the generation of target feature templates represented by target broadband one-dimensional image and narrowband modulation spectrum, the present invention proposes a radar target feature template generation method integrating the ocean environment. By collecting environmental data and target data, compiling and processing them, storing them in a standardized manner, and taking the ocean environment information as an important parameter, an environmental feature model is constructed, and the association process between environmental features and targets is performed to form a target feature template library, which has a reference significance for target recognition.

[0005] The technical solution for achieving the purpose of the present invention includes:

[0006] Step 1: Gather multi-source environment and radar target data, clean and process them to form a raw material library;

[0007] The method comprises: collecting multi-source environment and radar target data to form an original material library; cleaning the multi-source original data in the original material library to form a material library; the cleaning process comprises steps such as quality control, format conversion, and standardization;

[0008] Step 2: Extract features of the radar target to obtain feature information; the feature information includes a one-dimensional image of the target and a modulation spectrum; extract features of the environmental information, construct a spatiotemporal feature network model of the marine environment, and calculate environmental feature values;

[0009] Step 3: Correlation processing of environmental characteristics and target characteristics: Using the location and time information of the target, the target characteristic information is temporally and spatially correlated with the marine environmental characteristic network to obtain the marine environmental characteristic value of the target characteristic information, and the environmental characteristic information is incorporated into the target characteristic information to form marine target characteristic data;

[0010] Step 4: The marine target feature data is processed by statistical compilation, verification and deduplication to obtain feature templates and form a target feature template library.

[0011] Furthermore, step 2 also includes: first constructing time-space grids and ocean environment element sets of different scales, then normalizing and averaging each element of the ocean environment element set in each time-space grid, using weighted calculation to obtain the grid environment characteristic value, and finally obtaining a multi-layer time-space characteristic grid of the ocean environment.

[0012] Furthermore, in step 4, the target feature data is compiled according to the ocean environment feature value, attitude angle, signal-to-noise ratio, and bandwidth information, so as to finely manage the target feature data and form a target feature template library; by performing relevant calculations on the template data, the template is verified and deduplicated according to the correlation coefficient, and the target feature template product library is refined.

[0013] Ultimately, the generation of radar target feature templates under different marine environmental characteristics can be achieved, the problem of target feature template generation and management in different environments can be solved, and template data support can be provided for radar target recognition.

[0014] The present invention provides a method for generating a radar target feature template that integrates the marine environment. A model and method for expressing marine environment feature information is constructed to obtain a spatiotemporal grid of marine environment features, which can be associated and matched with target feature information, and further statistical compilation, verification and deduplication and other feature processing are performed to incorporate environmental information into the target feature template to form a target feature template based on the marine environment and radar target data, providing data support for target identification and machine learning. The method proposed in the present invention integrates marine environment information into the target feature template, and provides a set of solutions for the generation and management of radar target templates in different environments, which has a wide range of uses. Compared with traditional target feature template products, the present invention has significant advantages:

[0015] 1. Compatible with traditional target template processing;

[0016] 2. Ability to obtain characteristic information of the marine environment;

[0017] 3. Introduce marine environmental information to support refined management and utilization of target feature templates.

[0018] The present invention can be applied to the generation of other characteristic templates of targets to provide data support for further target recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is the target feature template generation block diagram. DETAILED DESCRIPTION

[0020] The present invention is further described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0021] The preferred implementation process of the present invention includes:

[0022] Step 1: Gather multi-source environment and radar target data to form a raw material library. Clean the multi-source raw data, mainly including quality control, format conversion, standardization and other steps to form a material library:

[0023] 1.1: Quality control mainly eliminates abnormal and erroneous data and performs data verification;

[0024] 1.2: Format conversion mainly converts data of different units and data types into a unified standard format;

[0025] 1.3: Standardization On the one hand, a standard environment and target database tables are established, and on the other hand, the processing process is standardized to support effective data processing in different environments.

[0026] Step 2: Extract features of radar targets and ocean environment information:

[0027] 2.1: Broadband feature extraction:

[0028] 2.1.1: According to the scattering point model, the echo of the target within the range unit i is: N i is the number of target scattering points within the distance, σ in is the echo emphasis of the nth scattering point in the range unit i, r i is the distance between the scattering point and the radar. The one-dimensional range image can be expressed as: X = [x(1)|,|x(2)|...|x(n)|] T ;

[0029] 2.1.2: Calculate the radial length: f = (pEnd - pStart) * s, where pEnd is the position of the last range unit in the threshold range unit, pStart is the position of the first range unit in the threshold range unit, s is the radar range resolution, and f is the radial length;

[0030] 2.1.3: Target length estimation: Where f is the radial length of the one-dimensional range image of the radar target, θ is the attitude angle, and L is the length of the radar target;

[0031] 2.2: Narrowband target feature extraction:

[0032] 2.2.1: Frequency domain characteristics of ideal JEM: If there is an even number of blades, then q m =1, otherwise, q m =2;

[0033] 2.2.2: The extracted features include signal-to-noise ratio, maximum spectral spacing, minimum spectral spacing, average spectral spacing, modulation period, number of modulation spectra, number of modulation spectra after desymmetry, average spacing after desymmetry, and target modulation spectrum after translation.

[0034] 2.2 Extract environmental feature information, construct a spatiotemporal feature network of the marine environment, and express the characteristics of the marine environment:

[0035] 2.2.1: Rasterize space and discretize time to build time-space grids of different scales The time and space scales can be selected according to needs, such as 10km for space scale and 1min for time scale;

[0036] 2.2.2: Select the marine environment elements that affect the expression of target features to form a marine environment element set E. For example, if n elements are considered, E i Represents the i-th environmental element, i = 1, 2..n;

[0037] 2.2.3: For any time-space network The values ​​of the environmental elements in are normalized and averaged, and the normalized value of the i-th element is e i ;

[0038] 2.2.4: Establish an environmental characteristic value calculation expression model. This method uses weighted calculation of environmental characteristic values: Thus, the grid environment characteristic value is obtained;

[0039] 2.2.5: Weight α i It can be determined by experts. If the influence of various environmental factors on radar target feature extraction and target recognition is basically similar, the average weight can be selected.

[0040] 2.2.6: Process the environmental data in the material library according to the above steps to obtain a multi-layer spatiotemporal characteristic network of the marine environment.

[0041] Step 3: Correlation processing of environmental features and target features: Using the position and time information of the target, the target feature information is temporally and spatially correlated with the ocean environment feature network to obtain the ocean environment feature value of the target feature information, and the environmental feature information is incorporated into the target feature information to form a radar target basic feature template;

[0042] Step 4: Perform statistical compilation, verification and deduplication on the target basic feature template to form a target feature template library:

[0043] 4.1 Statistical compilation and processing: multiple feature data of the same target are statistically compiled according to the attitude angle, signal-to-noise ratio, bandwidth, environmental characteristic value classification parameters and other conditions. The one-dimensional image feature template information includes: upper and lower limits of viewing angle, average radial length, average estimated target length, average autocorrelation coefficient, average center position, starting position, ending position, bandwidth, sampling rate, etc.; the modulation spectrum feature template information includes: upper and lower limits of viewing angle, maximum spectrum spacing, minimum spectrum spacing, average spectrum spacing, maximum number of modulation spectra after desymmetry, minimum number of modulation spectra after desymmetry, average number of modulation spectra after desymmetry, maximum average spectrum spacing after desymmetry, minimum average spectrum spacing after desymmetry, average average spectrum spacing after desymmetry, etc.;

[0044] 4.2 Verification and deduplication: Perform correlation calculations on the spectrum data of the one-dimensional image and modulation spectrum in the feature template, and remove duplicates based on the correlation, thereby ensuring the simplicity of the target feature template product.

[0045] The present invention can be applied to the fields of radar data analysis and target identification, collects and compiles marine environment and target data, solves the problem of target feature template generation and management under different marine environment backgrounds, and provides a basis for further target identification.

Claims

1. A method for generating a radar target feature template integrating an ocean environment, characterized in that: Step 1: Gather multi-source environment and radar target data to form an original material library; clean the multi-source original data of the original material library to form a material library; the cleaning process includes quality control, format conversion, and standardization; Step 2: Extract features of radar targets to obtain feature information; The characteristic information includes a one-dimensional image of the target and a modulation spectrum; feature extraction is performed on the environmental information, a spatiotemporal characteristic network model of the marine environment is constructed, and environmental characteristic values ​​are calculated; Step 2.1: Rasterize space and discretize time to construct time-space grids of different scales The time and space scales are selected according to needs; Step 2.2: Select the marine environment elements that affect the expression of target features to form a marine environment element set E, E i Represents the i-th environmental element, i = 1, 2..n; Step 2.3: For any time-space grid The values ​​of the environmental elements in are normalized and averaged, and the normalized value of the i-th element is e i ; Step 2.4: Establish an environmental characteristic value calculation expression model and use weighted calculation of environmental characteristic values: Thus, the grid environment characteristic value is obtained, where α i is the weight; Step 2.5: Weight α i It can be determined by experts that if the environmental factors have similar effects on radar target feature extraction and target recognition, the average weight Step 2.6: Process the environmental data in the material library according to the above steps 2.1 to 2.5 to obtain a multi-layer spatiotemporal characteristic network of the marine environment; Step 3: Correlation processing of environmental features and target features: using the target's position and time information, the target feature information is temporally and spatially correlated with the ocean environment feature network to obtain a radar target basic feature template that integrates the ocean environment; Step 4: Perform feature processing on the basic feature template to form a target feature template library; the feature processing includes statistical compilation and verification and deduplication.

2. The method for generating a radar target feature template integrating an ocean environment according to claim 1, characterized in that: The step 4 also includes: statistically compiling the target feature data according to the ocean environment feature values, attitude angles, signal-to-noise ratios, and bandwidth information to form a target feature template library; performing correlation calculations on the template data, verifying and removing duplicates of the templates based on the correlation coefficients, and refining the target feature template product library.

Images