The Maneuvering Detection Based On Wavelet Transformation

In recent 20 years, people have conducted the massive research to the maneuvering target track, and obtained some beneficial results in establishing system model, maneuvering detection and adapted algorithms using the Kalman filtering algorithm. The difficulty of maneuvering target track lies in determining whether and when started maneuvering, according to which it transforms the system model promptly and accurately, and then realizes super track to maneuvering and the non-maneuvering target. But in practical, it is difficult to choose appropriate model matching the target travel situation, tracks filter is specially sensitive to the noise and the disturbance, especially in the strong noise, relatively small maneuvering nearly submerges in the noise, the target model is prone to transform wrongly by triggering of the noise, and finally causes unbalance between the simulated target and the target travel situation. If the model is not correct, the tracking system may diverge and cause the wrong track.This article firstly introduces in detail several traditional maneuvering detection algorithms as follows: Adjustable White Noise Model, the Input Estimation and the Variable Dimension algorithm, and analyzes their weaknesses: they will cause two kinds of Probability of error—False alarm probability and Police probability of leakage regardless of the examination threshold selected; besides that, these traditional algorithms do not have the ability of anti-Interference, and possibly causes the maneuvering detection unsuccessful in low SNR situation.In order to solve the above problems, Hong firstly applied multi-resolution technology in the target tracking based on the wavelet transformation, introduced the multi-resolution target tracking method, and realized the transformation from single resolution measurement to multi-resolution Measurement using the wavelet theory. When the primary data which contains the noise is decomposed to the lower resolution level, the noise is greatly reduced because of the wavelet transformation low pass filter ability, so the target maneuvering state turns out discernible, thus it can detect the maneuvering promptly and accurately. Although the maneuvering target multi-resolution track method demonstrates the unique performance in the noise, according to target tracking, the operand is doubled and it application in real-time situation will be seriously restricted.In order to overcome the question, this paper makes the improvement to the multi-resolution target tracking algorithm which is proposed by Hong, that is, substitutes original multi-resolution method by two wavelet transformation fast algorithms: MALLAT algorithm and Lifting wavelet transformation algorithm, thus effectively speed up the running of this algorithm and can be comparatively utilize in the real-time target tracking. Simulations in MATLAB indicate that, the new algorithm realizes the maneuvering detection well.