Research Of Data Fusion Algorithm Based On Clustering D-S Evidence Theory

With the development and improvement of the sensing technology,sensors become more and more important in data fusion.However,the data gathered by sensors has different meanings and structures because of the different types of the sensors.And the information gathered by sensors exists uncertainty when considering the surroundings,the vicious disturbance coming from the enemy and the detection precision of the sensors and so on.It has been a hot topic to focus on the problem that how to fuse the uncertain information.Therefore,this paper puts forward a new method to measure the conflict between evidences and a method of data fusion based on clustering evidence theory with combining clustering analysis with DS evidence theory.This combination method is applied to the Automobile Collision Forewarning System.The main contents of this paper include the following three aspects:(1)A new conflict measurement method is proposed based on coarse-grained distance and fine-grained distance.Firstly,define a space distance and combine the space distance with Jousselme evidence distance to form a fine-grained distance.The fine-grained distance considers the conflict degree between evidences from the difference of basic possibility assignment of every element in evidences;Then define a coarse-grained distance by considering the entire decision difference between evidences;At last,a new evidence conflict measurement method is proposed by considering the relations between evidences from the coarse-grained angle and fine-grained angle.(2)A combination method of conflict evidences based on clustering evidence theory is proposed.Firstly,classify evidences into 3 categories:consistent evidences,non-conflict evidences and conflict evidences according to the new conflict measurement and the local conflict parameter;Then the 3 categories evidences are modified in different ways;At last,combining the modified evidences by using DS combination rules.The experiment shows that the new combination method can fuse the conflict evidences effectively.(3)It is necessary to make sure that data fusion technology is applied to the Automobile Collision System as the data coming from this system has many features including heterogeneity,redundancy,real-time and so on.Therefor,this paper designs an Automobile Collision Forewarning System Model based on data fusion and combines the data in this system by making use of the combination method based on clustering evidence theory with simulation and this supplies reliable advice for drivers.