Research On Several Key Technologies Of Multi-Sensor Data Fusion

With the development of science and technology, particularly microelectronics technology, integrated circuits technology, computer science, modern signal processing techniques and sensor technology, multi-sensor data fusion has become a new area of research and academic direction. Multi-sensor data fusion technology is widely applied in the military field and civilian areas. Data fusion technology is developed from the military field. The general concept and model of data fusion in the military field is given by Joint Directors of Laboratories Data Fusion Subpanel (JDLDFS).This thesis lists some of the key technologies of data fusion in the framework of JDL model. We have done in-depth studies on several issues including sensor management, feature level data fusion algorithm and feature analysis, automatic evaluation of data fusion system and situation assessment. The main contributions are:1. A sensor management approach based on Discrete Particle Swarm Optimizer (DPSO) is presented. This approach presents the objective function of sensor management according to certain optimization discipline and select optimal sensor suite by DPSO. The discipline adopted here is maximizing the information gain and minimizing the process time. In order to measure the information gain, we propose a method based on features of objects which utilizes fuzzy K-L discrimination gain to measure the information gain.2. Present a new feature level data fusion method which is able to combine fuzzy conflict evdidences effectively. Study methods of fuzzy information fusion including the method based on fuzy integral and fuzzy logic. Realize the feature level data fusion method based on fuzy integral to deal with the problem of classification in the situation of small example scale. In order to resolve the problem of fuzzy logic that reasoning method is too simple to take full advantage of useful information, apply D-S theory to fuzzy reasoning. We analyze the flaw of D-S theory in combining conflict evidences and present a new method to combine fuzzy conflict evidences.3. Present an automatic method based on uncertainty to evaluate performance of...