Study On Information Fusion Based On Evidence Theory And It's Application In Water Quality Monitoring

As the appearance of different kinds of multi-sensor fusion systems for complicated application background, the research on fusion algorithms is attended. In the thesis, D-S evidence theory and its key technologies are analyzed and studied. Their disadvantages are pointed and corresponding improved methods are proposed.Information fusion technology is first introduced to the process of water quality monitoring data processing. As viewed from information fusion, the methods proposed in the thesis are applied to water quality monitoring data processing, and their feasibility and validity are analyzed. The main contents and conclusions are as follows.Firstly, a combination method for conflict evidences based on a modified model is proposed. The combination rule of the method satisfies commutative law and associative law, fully uses information of conflict evidences, and solves three kinds of problems appeared in conflict evidences combination process.Secondly, a similarity measure is introduced and a fuzzy-evidence reasoning method based on the measure is proposed. Compared with other generalized methods, the method is more effective in avoiding the insensitivity problem of belief function to the change of fuzzy focal elements and catching information from the change of fuzzy focal elements.Thirdly, with the characteristics of Vague focal elements, methods to measure belief function and plausibility function of Vague focal elements are proposed, and characteristics of the functions are discussed and proved. Based on the similarity measure of Vague sets, a combination rule of Vague focal elements is proposed. By experiments, the rationality and feasibility of the Vague-evidence theory are analyzed.Fourthly, to decision-making problems based on multi-source information fusion results with D-S evidence structure, a decision-making method based on distance measure is proposed. By experiments, the method is analyzed and compared with other methods. The results illustrate the method is effective in decision-making based on the fusion results of conflict evidences or non-conflict evidences, and overcomes the disadvantage of others only effective on that of conflict evidences.Fifthly, as viewed from system, information fusion efficiency index is defined to measure the information fusion efficiency of multi-sensor systems. Using the quantification index, the efficiency of multi-source information and fusion algorithms are analyzed. The experiment results indicate that the index can express the influence of information selection, information correlation and fusion algorithms on information fusion efficiency.Lastly, based on D-S evidence theory and its improved methods, surface water quality monitoring data are fused. The fusion results are compared with that on other methods, and the comparison indicates the fusion methods are feasible to assess water quality status. Using complementarity of remote sensing information and ground monitoring information on water quality, a data fusion method based on neural network and evidence theory is proposed. Experiment results explain that the method is more effective than neural network on a single remote sensing source.