An Intelligent Fusion Method of Sequential Images Based on Improved DSmT for Target Recognition

It is proposed that a sequential images object recognition method combining a BP neural network with the fast mass functions convergence algorithm based on DSmT. The revised Hu invariant moments are used as the image features. And the sequential images are fused in time domain in the view of information fusion. The basic belief assignment function is created by the initial recognition result from a BP neural network. It completes the decision-level fusion with the fast mass functions convergence algorithm based on DSmT. Simulation result shows that the proposed method can improve the accuracy significantly for three-dimensional aircraft images target recognition.