Research on Target Recognition Based on Improved LSSVM

It is a process that a particular type of goal target recognition from other goals based on certain characteristics. It is one of the hotspots of current research. LSSVM is a common algorithm for target recognition. It is simple, fast and accurate. Against to the parameters difficult to be determined in the LSSVM and the disadvantage of traditional method, it is proposed an improved differential evolution algorithm to optimize the parameters of the LSSVM in the paper. To restrain the premature convergence of the traditional algorithm, the premature judgment mechanism is introduced by the way of improving the mutation strategy. The experiment results showed that the improved algorithm has the ability of jumping out of local advantages. The results are better than the traditional algorithm. It is proved with the photoelectric servo tracking turntable that the improved algorithm has fast convergence speed and high accuracy. And it is demonstrated the excellence of the algorithm that correct recognition rate can be improved from 80% to 91%.