Method and device for improving current statistical model based on neural network, storage medium and computer equipment

Abstract

The invention provides a current statistical model improvement method based on a neural network. The method comprises the steps: establishing an improved current statistical model based on a neural network, wherein the neural network serves as a feedback network to be connected with a traditional statistical model, so as to obtain the maximum acceleration value conforming to the target maneuvering condition; selecting a training set and a label set, and preprocessing the training set and the label set according to a set normalization standardization rule; setting training parameters, and training the neural network by using the training parameters to enable the neural network to finally achieve convergence; inputting a to-be-processed target estimation state and measurement into the improved model to obtain a state estimation value of the target at the current moment. According to the method, by means of the nonlinear expression ability of the neural network, the adaptive adjustment ability of a traditional current statistical model is enhanced, the back-propagation neural network serves as a feedback network to adaptively adjust the maximum acceleration, mapping among the target state, measurement and the maximum acceleration is established, and the tracking error during maneuvering of the target is greatly reduced.

Description

technical field [0001] The invention relates to the technical field of radar data processing, in particular to an improved current statistical model method, device, storage medium and equipment based on a neural network. Background technique [0002] In recent years, target tracking is the core key technology of radar data processing. Through track initiation, track prediction, and filtering algorithm, the real-time state estimation of the target can be carried out through the measurement information collected by the radar, and the target's trajectory and Motion parameters, so as to realize the tracking of the target and complete the display of the radar terminal. Among them, track prediction and filtering algorithms belong to the part of target tracking, which are very important for estimating the target state with high precision. The existing track prediction method is after the start of the track, while the filtering algorithm is after the track prediction, both of which...

AI Technical Summary

Problems solved by technology

[0003] The publication number is a radar target tracking method provided by CN111157983A, which adjusts the parameters of the model according to the change of target maneuvering characteristics, realizes real-time adaptive target tracking update, realizes high-precision target tracking, and improves radar recognition accuracy, but The target tracking model of this method is mainly a uniform velocity model, a uniform acceleration model, a current statistical model, a turning model, and a continuous turning model, and several models are expressed in the form of Bayesian filtering. In practical applications, due to the unknown maneuverability of the target, the tracking effect of the existing algorithm is good and bad, and it is difficult to achieve stable and good tracking effect under any maneuvering situation of the target.

At the same time, the current statistical model method still has the problem that the parameters cannot be adjusted adaptively. When the target suddenly maneuvers, the tracking error will increase significantly.

Manual adjustment is required, and the adjusted tracking effect is difficult to achieve real-time optimality. Therefore, the existing current statistical model methods have problems such as simple model, low complexity, poor adaptability, and lack of learning ability, which are difficult to solve fundamentally as a whole. High precision tracking problem

Images