Investigation On State Information Fusion

Information fusion, also called multi-sensor information fusion, is a multi-level, multi-layer process of information from different sensors, in order to get more abundant, accurate, reliable information. During the recent 20 years, we have obtained many achievements from the development of sensor technology and improvement of multi-sensor systems development. The information fusion, especially state information fusion, applied widely in air traffic control, national defense, seaport supervision, rebot vision, remote sensing, airport forecast and artificial traffic.In the research of information fusion, people discovered many processes have the multiscale characteristic or multiscale effects, traditional single scale's analysis can't express the abundant time information in these signals, along with request of the information accuracy raised increasingly, this problem also correspond of increasingly outstanding. Therefore, there is an utmost urgency to the related research of the wavelet information fusion. In recent years, wavelet transformation led to provide an emollient tool for the sake of resolving this problem.The main contents of this dissertation are the research and application of information fusion algorithm. The textual research mainly involves wavelet transformation theories, multiscale signal filter and multi information fusion based on multiscale. The main contributions of this dissertation summarized as follows:(1) Based on the optimal information fusion rule in the mean of linear minimum variance, a new kind of improved weighted fusion rule is proposed. The new algorithm revised the error caused by the matrix power information fusion rule, raised the fusion accuracy and lower time complications in the meantime. We prove the validity of the new rule in both mobile and no-mobile example.(2) We carried on a research in multiscale information fusion in wavelet domain. Firstly analyzed the multiscale transformation filters, and then proposed the multi-scale information unify linear fusion in wavelet domain and multi-scale information amended fusion in wavelet domain. The two kinds of filter methods represented the thinking of multiscale analyze, broke the limit of single scale, introduce temporal fusion into the wavelet area fusion, having powerful applied value.(3) Developed the information fusion software. The information fusion software package RadSys v3.0 is developed by SJTU Intelligent Information Process&Control Lab for infrared and radar data fusion system. That software provided the demo of this dissertation's algorithms can provided various valuations of the single target, team targets. Along with research going deep, the function of this system will also expand further.