CN-122020371-A - Fuse accurate height measurement method based on parameterized model and deep learning

CN122020371ACN 122020371 ACN122020371 ACN 122020371ACN-122020371-A

Abstract

The invention belongs to the technical field of radio fuses, and particularly relates to a fuse accurate height measurement method based on a parameterized model and deep learning, which comprises the steps of determining electromagnetic scattering parameter combinations in the process of establishing a ground echo model, and sampling in a multidimensional prediction parameter space; the method comprises the steps of constructing a controllable electromagnetic scattering parameterized ground echo simulator based on an improved Ulaby model, carrying out echo modeling on N groups of parameter combinations by utilizing the simulator to obtain an original echo signal, extracting original echo signal characteristics to be combined into a high-dimensional characteristic vector to serve as input of a deep learning model, constructing a deep full-connected network, training the network model by taking the characteristic vector as input and an electromagnetic scattering parameter combination estimated value as a label, collecting actual fuze echo data, acquiring the electromagnetic scattering parameter combination estimated value by utilizing the trained network model, optimizing, carrying out error correction on height information in the optimized electromagnetic scattering parameters, and realizing accurate fuze height measurement.

Inventors

PAN XI
ZHANG YAO

Assignees

北京理工大学

Dates

Publication Date: 20260512
Application Date: 20260120

Claims (10)

1. A fuze accurate height measurement method based on parameterized model and deep learning is characterized by comprising the following specific steps: The method comprises the steps of firstly determining that electromagnetic scattering parameter combinations in the ground echo model building process comprise Ulaby model fitting parameters, incidence angles and heights, sampling in a multidimensional prediction parameter space to obtain N groups of electromagnetic scattering parameter combinations, and secondly building a controllable electromagnetic scattering parameterized ground echo simulator based on an improved Ulaby model, wherein the ground echo simulator is formed by introducing correction factors Finally, carrying out echo modeling on N groups of parameter combinations by using the simulator to obtain an original echo signal, extracting the characteristics of the original echo signal and combining the characteristics of the original echo signal into a high-dimensional characteristic vector to be used as the input of a deep learning model; Training a network model, namely constructing a deep full-connection network, and training the network model by taking a feature vector as input and an electromagnetic scattering parameter combination estimated value as a label; The method comprises the steps of collecting actual fuze echo data, extracting high-dimensional feature vectors, inputting the high-dimensional feature vectors into a trained network model, outputting an electromagnetic scattering parameter combination estimated value to be optimized, inverting the fuze echo data to optimize the electromagnetic scattering parameter combination estimated value to be optimized, and correcting errors of the height information in the optimized electromagnetic scattering parameters to achieve fuze accurate height measurement.
2. The method for precisely measuring the height of the fuze based on the parameterized model and the deep learning according to claim 1, wherein the inversion by utilizing the fuze echo data is used for optimizing the electromagnetic scattering parameter estimation value to be optimized, and the optimization objective function is as follows: Wherein: represents the jth eigenvalue of the measured echo signal, The j-th characteristic value is calculated for the electromagnetic scattering parameter to be optimized; taking q as a step length, carrying out electromagnetic scattering parameter value in a set [ Qmin, qmax ] interval to serve as an optimized parameter interval, and calculating an electromagnetic scattering parameter estimated value corresponding to a minimum objective function; Repeating the steps until the objective function meets the set requirement, and obtaining the inverted electromagnetic scattering parameter estimated value.
3. The method for precisely measuring the height of the fuze based on the parameterized model and the deep learning according to claim 2, wherein the error correction is performed on the height information in the optimized electromagnetic scattering parameters, specifically: Acquiring real echo data of the ground with a known height, and obtaining an inverted electromagnetic scattering parameter estimated value by using the trained DNN; taking inverted electromagnetic scattering parameters as input, taking errors between real heights and measured heights as output, and training a linear regression correction model based on a gradient lifting tree; and carrying out error correction on the height information in the optimized electromagnetic scattering parameters by using the linear regression model.
4. The method for precisely measuring the height of the fuze based on the parameterized model and the deep learning according to claim 2, wherein the ground echo simulator model is as follows: Where N U is the total number of scattering units in the ground echo region, P t is the power of the transmitted signal, G i is the antenna gain of the ith scattering unit, σ i is the backscatter coefficient of the scattering unit, λ is the operating wavelength, R i (t) is the distance from the ith scattering unit to the detector at the current time, Is the time delay of the target echo signal relative to the transmit signal, Is the phase shift caused by the target reflection, f 0 is the carrier frequency, Is the frequency modulation slope.
5. The method for accurately measuring the height of the fuze based on the parameterized model and the deep learning according to claim 4, wherein the method is characterized in that the frequency domains of the echo signals under different landform conditions are fitted by using measured data to obtain corresponding experience correction factors 。
6. The fuze accurate height measurement method based on parameterized model and deep learning is characterized in that when original echo signal features are extracted, echo signals of several continuous periods are integrated into a time domain matrix, multi-period echo signals are processed through data statistics analysis to extract the mean value of the time domain features of each period echo signal, fourier transformation is conducted on the multi-period echo signals to extract the mean value of the frequency spectrum features, short-time Fourier transformation is used to obtain combined time-frequency distribution of each period signal, the mean value of the time spectrum features is extracted, the energy feature mean value of each period echo is calculated, and the extracted features are spliced in series to form a high-dimensional echo signal feature vector.
7. The method for fuze accurate height measurement based on parameterized model and deep learning according to claim 6, wherein the feature vector is normalized to map the value in the feature vector between (0, 1).
8. The method for precisely measuring the height of the fuze based on the parameterized model and the deep learning according to claim 7, wherein the obtained data set is randomly divided into a training set, a verification set and a test set according to a set proportion, and the network model is trained, verified and tested.
9. The method for accurately measuring the height of the fuze based on the parameterized model and the deep learning according to claim 1 is characterized in that the neural network model adopts a mean square error as a loss function to enable the model to converge, and a gradient descent algorithm is utilized to update the weight and the biased gradient according to the loss function.
10. The method for precisely measuring the height of the fuze based on the parameterized model and the deep learning according to claim 1, wherein N groups of parameter combinations are obtained by utilizing a full coverage data set of Latin hypercube sampled high-dimensional parameter space.

Description

Fuse accurate height measurement method based on parameterized model and deep learning Technical Field The invention belongs to the technical field of radio fuses, and particularly relates to a fuse accurate height measurement method based on a parameterized model and deep learning. Background Electromagnetic scattering characteristics of complex landforms have significant influence on radio fuze echo signals, and acquisition of typical ground parameters is a key to achieving higher distance accuracy in complex surface environments. The existing empirical model has weak generalization capability and poor interpretability, and parameter values of the existing empirical model often need to be matched with different environmental conditions, fuze characteristics and signal processing methods. In order to solve the problems of reduced distance precision and poor scene adaptability of a radio fuse under the condition of complex landforms, a fuse accurate height measurement method based on a parameterized model and deep learning is provided. The existing classical ground scattering coefficient empirical models, such as GIT models, urapine models and the like, mostly acquire actual measurement data of various ground features aiming at common wave bands of the radar, and obtain corresponding empirical formulas through a curve fitting method, so that the complete ground features are difficult to cover. And the adaptability of the empirical model to the non-uniform topography is insufficient, and the parameter assumption has deviation from the real scene. Disclosure of Invention In view of the above, the invention provides a fuze accurate height measurement method based on a parameterized model and deep learning, which can realize fuze accurate height measurement. The technical scheme for realizing the invention is as follows: In a first aspect, the invention relates to a fuze accurate height measurement method based on parameterized model and deep learning, which comprises the following specific processes: The method comprises the steps of firstly determining that electromagnetic scattering parameter combinations in the ground echo model building process comprise Ulaby model fitting parameters, incidence angles and heights, sampling in a multidimensional prediction parameter space to obtain N groups of electromagnetic scattering parameter combinations, and secondly building a controllable electromagnetic scattering parameterized ground echo simulator based on an improved Ulaby model, wherein the ground echo simulator is formed by introducing correction factors Finally, carrying out echo modeling on N groups of parameter combinations by using the simulator to obtain an original echo signal, extracting the characteristics of the original echo signal and combining the characteristics of the original echo signal into a high-dimensional characteristic vector to be used as the input of a deep learning model; Training a network model, namely constructing a deep full-connection network, and training the network model by taking a feature vector as input and an electromagnetic scattering parameter combination estimated value as a label; The method comprises the steps of collecting actual fuze echo data, extracting high-dimensional feature vectors, inputting the high-dimensional feature vectors into a trained network model, outputting an electromagnetic scattering parameter combination estimated value to be optimized, inverting the fuze echo data to optimize the electromagnetic scattering parameter combination estimated value to be optimized, and correcting errors of the height information in the optimized electromagnetic scattering parameters to achieve fuze accurate height measurement. Optionally, the inversion performed by using the fuze echo data is used for optimizing the electromagnetic scattering parameter estimation value to be optimized, and the optimization objective function is as follows: Wherein: represents the jth eigenvalue of the measured echo signal, The j-th characteristic value is calculated for the electromagnetic scattering parameter to be optimized; taking q as a step length, carrying out electromagnetic scattering parameter value in a set [ Qmin, qmax ] interval to serve as an optimized parameter interval, and calculating an electromagnetic scattering parameter estimated value corresponding to a minimum objective function; Repeating the steps until the objective function meets the set requirement, and obtaining the inverted electromagnetic scattering parameter estimated value. Optionally, the error correction is performed on the height information in the optimized electromagnetic scattering parameter, which specifically includes: Acquiring real echo data of the ground with a known height, and obtaining an inverted electromagnetic scattering parameter estimated value by using the trained DNN; taking inverted electromagnetic scattering parameters as input, taking errors between real heights and measured h