Multi-Sensor Information Fusion Algorithm And Application

With the development of information and science technology, from the 20th century and 70s, there is a new concept of "information fusion" which is interpreted as fusing multi-sensor obtained information and the results have more useful information than any single sensor. Multi-sensor information fusion is a hot area of academic research and it has a broad application prospects in the military and civilian.In this paper, we research on the similarity between multi-sensor data fusion, which problem is in target recognition and application. The similar measurement problems have two cases: one is that when different multi-sensors observe the same objections or characteristics without knowing any prior knowledge, we compare with the effect between raw data and fusion data by studying the similarity of observation information; another is that when the same type multi-sensors observe different objections or characteristics with the prior knowledge is known, we analyze fusion results by studying the similarity of observation information. Thesis results are as follows:(1) Summarizing multi-sensor information fusion's researching status of current domestic and international; studying the structural model of information fusion; generalizing information fusion common methods;(2) Proposing fusion algorithm based on consistent and reliable measure for the first case. From each moment consistent and reliable measure of the sensor observations, the algorithm designs fusion algorithm and gives a different time line of each sensor reliability measures. The algorithm has high accuracy;(3) For the second case, we research the theory of D-S's evidence and vague sets. The theory of D-S evidence will be failure when the individual evidence is conflict; Different similarity measures of vague sets'theory have greater impact on fusion results;(4) For the case of the theory of D-S evidence will be failure when the individual evidence is conflict, we propose new D-S evidence method based on weighted similarity, which is called WSD-S (Weighted Similarity Dempster Shafer). This method improves the accuracy of the algorithm convergence and reduces the complexity of the algorithm because of selecting objective weight values;(5) Proposing a new vague sets'model of similarity measure and giving the corresponding proof. The proposed new model is used in air motor fault recognition. Here, we used the concept of closeness degree of fuzzy theory and the fault which is to be diagnosed is to be correct classified by comparing with the closeness between the faults which is to be diagnosed with the known faults;(6) Proposing two different fusion model by combing WSD-S evidence theory and the BP and RBF neural network and having effective identification of turbine faults;(7) Researching the body sensor network which is used real-time monitoring human health. The body sensor network is not only excellent performance, but also is very safe and convenient for using. At the same time, we use the adaptive weighting fusion algorithm to monitor the parameters of physical symptoms, which can effectively analyze physical health.