Data Fusion Technique Based on Dempster-Shafer Evidence Theory

The data fusion technique of D S evidence theory is described systematically, the method of constructing basic probability assignment (BPA) of sensor is analyzed, and the dual threshold is used to divide the measurement space into confidence regions and non confidence ones. Then the Dempster's rule of combination is used to combine the belief functions of two sensors. The improvement on performance of receiver operating character (ROC) with data fusion is simulated. Double channel sequence image produced by infrared scene generator and real target sequence image of double channel imager--AGEMA THV900 are detected and fused, and the results are satisfactory.