Target micro-Doppler curve extraction method based on high-order particle filtering

Abstract

The invention discloses a target micro-Doppler curve extraction method based on high-order particle filtering. The method comprises the steps of determining a fact that a target exists in a radar detecting range, decomposing motion of the target relative to the radar to target translation and target inching, and establishing a state model of a target instantaneous Doppler frequency at the time k by means of a high-order particle filtering method; determining an observing vector which is composed of target observation values of N discrete short-time Fourier transform units at the time k, and establishing a target observation model at the time k; calculating a state variable estimated value beta<k+1><~> which corresponds an instantaneous Doppler frequency in target motion at the time k+1, astate variable estimated value g<k+1><~> which corresponds with an instantaneous Doppler frequency in target inching at the time k+1, and an instantaneous Doppler frequency estimated value gamma<k+1><~> of the target at the time k+1, setting the value of k to be 0 to k-1, respectively obtaining the instantaneous Doppler frequency estimated value gamma1<~> of the target at the first time to the instantaneous Doppler frequency estimated value gammaK<~> of the target at the time K, and drawing a curve, wherein the curve is the extracted target micro-Doppler curve.

Description

technical field [0001] The invention belongs to the technical field of radar target recognition, and relates to a method for extracting target micro-Doppler curves based on high-order particle filtering, which is suitable for extracting micro-Doppler curves for sinusoidal modulation. Background technique [0002] The micro-Doppler curve is generally considered to be able to uniquely characterize the motion characteristics of the target, and has the ability to provide reliable information for the identification and classification of radar targets; however, how to effectively extract the micro-Doppler curve of the target is still a problem in practical applications. a challenge. [0003] Generally, the methods for extracting micro-Doppler curves can be roughly divided into two categories. The first category is to establish a sub-signal dictionary based on the micro-Doppler signal model, and match the radar echo of the target with the established dictionary to estimate the corr...

AI Technical Summary

Problems solved by technology

[0002] The micro-Doppler curve is generally considered to be able to uniquely characterize the motion characteristics of the target, and has the ability to provide reliable information for the identification and classification of radar targets; however, how to effectively extract the micro-Doppler curve of the target is still a problem in practical applications. a challenge

[0003] Generally, the methods for extracting micro-Doppler curves can be roughly divided into two categories. The first category is to establish a sub-signal dictionary based on the micro-Doppler signal model, and match the radar echo of the target with the established dictionary to estimate the corresponding model parameters; this type of method requires the prior information of the target micro-Doppler signal model, and when there are many model parameters, it is necessary to construct a high-dimensional sub-signal dictionary, resulting in a large amount of calculation; taking the spin target as an example, even if The spin target translation has been fully compensated, and the sub-signal dictionary is also three-dimensional, because it contains three unknown model parameters; the second type of method is to use the time-frequency analysis method to obtain the time-instantaneous Doppler map, and The model parameters of the target are estimated according to the instantaneous Doppler curve in the time-instantaneous Doppler image. Commonly used time-frequency analysis methods include Short Time Fourier Transform (STFT), Wigner-Ville distribution, and S-method , Time-varying Autoregressive Model (Time-varyingAutoregressive Model, TVAR), etc.; the computational complexity of the commonly used time-frequency analysis method is usually relatively low, but requires a clear time-instantaneous Doppler map, and requires a single distance The number of scattering centers in the unit is as small as possible; the former is usually difficult to guarantee, and the latter can be achieved by increasing the bandwidth of the radar transmitting signal

[0004] The Radon transform is a commonly used method for extracting sinusoids from the time-frequency plane in the low SNR environment, which requires the assumption that the target translation is fully compensated; however, this assumption is usually difficult to achieve, especially for high frequencies For radar, small compensation errors may also lead to very large micro-Doppler modulation; therefore, it is necessary to describe the translation of the target by the third-order or higher-order velocity, but this will lead to the computational complexity of the Radon transform In fact, the problem of extracting micro-Doppler curves from the time-frequency plane under the condition of low SNR is consistent with the problem of tracking before detection (TBD) of weak targets, which is very sensitive to the fluctuation of target echo amplitude. Robust, but it is mainly aimed at the first-order Markov process and cannot be directly extended to the extraction of micro-Doppler curves

Images