Uncertainty Reasoning And Spatial And Temporal Alignment Technology

Multi-sensor data fusion is a new multi-source information process technique. In multi-sensor fusion system, the research of uncertain information processing and time space registration is a hot point. As a widely adopted uncertain information inference method, Dempster-Shafer evidence theory offers us an available approach of fusing multi source evidence. But it can only be used when each of evidence is independent, in practical applications there are many evidences can not satisfy this condition, so how to improve this theory has become to a focus problem; More over, multi source information fusion needs data of multi platform, so how to unify data of multi platform to the same time space so as to fuse is an important factor influencing the performance of fusion. Therefore the dissertation researches the uncertain information process and time space registration problem of dissimilar sensor fusion, the main work and innovations are as fellow1. Anew dependent evidence process method is proposed. Based on analyzing the current dependent evidence process method and for the Bayesian framework, a new dependent evidence process method based on identifying the independent original evidence and dependent original evidence is proposed. And some conclusions of solving equation group are given.2. A new conflict evidence process approach is proposed. Based on analyzing the current conflict evidence process method, a new method of processing conflict evidence is given. This method can adjust each proposition's weight automatically based their probability. Which is more efficient than the current method and the rationality and reliability of fusion result is improved.3. The time space registration method of different sensor is researched. A new registration approach based on data fitting is proposed. This method can adjust the data of different sensors and sample interval to the same time space so as to extract feature and get the fusion result. Comparing to other methods, this method has the character of easily calculation, quick speed and preferable fusion performance.4. The asynchronous sensor fusion's applications in auto target recognition is researched. Some typical methods of basic probability assignment are discussed. An example of applying D-S evidence theory to asynchronous sensor target recognition is given.The research work of this dissertation has made certain contributions to our country's military high-tech and the development of information fusion theory, also it has important engineering value.