基于差分最小二乘支持向量机的目标识别

Target Recognition Based on Differential Evolution Algorithm of Least Squares Support Vector Machine

目标识别是目前机器视觉、图像处理和模式识别领域的研究热点之一,广泛应用于各行各业。最小二乘支持向量机算法简便、速度快、精度高,是当前目标识别的主流算法之一。针对最小二乘支持向量机的参数难以确定,仅靠传统经验试凑的方法不易实现,且结果不理想;提出一种改进的差分进化算法实现最小二乘支持向量机的参数整定。通过改进变异策略,引入早熟判断机制,遏制了传统算法早熟收敛的问题。通过实验仿真,验证了改进算法可跳出局部最优点,结果比传统算法更优。以SM-TMSSY光电伺服跟踪转台为实验平台进行实例验证,证明了改进算法收敛速度快、精度高,正确识别率可从85%提高到92.5%,验证了算法的优越性。

Target recognition is one of the research hotspots in the field of machine vision,image processing and pattern recognition. It is widely used in all fields of life. Least squares support vector machine（ LSSVM） is one of the mainstream algorithms for target recognition. It is simple,fast and accurate. Aiming at the parameters of the LSSVM difficult to be determined,the traditional method is hard to implement and the results are not satisfactory.Thus,an improved differential evolution algorithm is proposed to realize the parameter optimization of the LSSVM.By improving the mutation strategy,the premature judgment mechanism is introduced to restrain the premature convergence of the traditional algorithm. The simulation results showed that the improved algorithm can jump out of local advantages,and the results are better than the traditional algorithm＇s. The algorithm can be verified by SMTMSSY photoelectric servo tracking turntable. It is proved that the improved algorithm has fast convergence speed,high accuracy,and the correct recognition rate can be increased from 85% to 92. 5%. The result verifies the superiority of the algorithm.