Research And Realize Of Multi-sensor Image Fusion And Target Tracking Algorithm Based On DSP

As the key technology of precise guidance system multi-sensor image fusion and target tracking technology has been the hot spot in research. However, there are still many problems difficult to solve to these two techniques. To design a set of multi-sensor image fusion and target tracking algorithm with strong real-time performance, high precision and good stability has important significance. Based on the study of existing image fusion and target tracking algorithm and in view of the existing problems, the major achievements of this paper are as follows.In image fusion section, this paper proposes an improved new algorithm of image fusion based on improved NSCT transform. Firstly, using the improved NSCT transform as the tool of image decomposition on multi-scale and multi-direction, then designing different fusion rules in view of the different characteristics of high frequency components and low frequency components. For the low frequency components this paper designs the rule of combining regional energy and regional variance. And for the high frequency components, the rule of combining maximum absolute value of the coefficient and spatial frequency is designed. After the fusion of high and low frequency components, the fusion image is obtained by the inverse transform. Finally the experimental results show the image fusion algorithm this paper proposed has an ideal fusion effect from the two aspects that subjective and objective.In target tracking section, in order to solve the problem that the traditional MeanShift algorithm has a poor tracking effect when the target has scale change, rotation, noise or other complex cases and the lack of target template update, this paper designs a new method of object description based on direction characteristic of SURF feature points, then combines this feature with the gray feature of the target and proposes a multi features improved MeanShift target tracking algorithm based on gray feature and local invariant feature of SURF, this algorithm is also introduced target template update mechanism; finally the improved occlusion discrimination strategy based on multi features and Kalman prediction are embedded into the improved MeanShift tracking algorithm, this paper designs a new target tracking algorithm of self-adaptive occlusion handling. The experimental results show that the tracking algorithm this paper proposed has good tracking effect in complex tracking environment and partially solves the occlusion problem effectively.In realization of the algorithm section, this paper achieves the algorithm and optimizes the algorithm on the image processing unit based on double TMS320C6678 DSPs. Finally the testing results show that the proposed algorithm has strong real-time performance, high precision, good stability and it can meet the requirement of the system.