论文部分内容阅读
因其核函数的良好性态,RBF核SVM(RBF-SVM)在实际应用中表现出良好的学习性能,但是RBF核函数中的参数对SVM的性能起决定性作用。阐述了RBF-SVM的性能随着变化而变化的规律,并将RBF-SVM引入自动羽绒识别系统中。根据自动羽绒识别系统的实际需求和RBF-SVM的性能变化规律,论述了本系统中参数的选取依据和选取过程,并且给出了的相关曲线变化图。通过研究,最后得到适合本系统的识别模型,从而提高了系统的总体识别率。同时,也验证了RBF-SVM的良好特性和其受参数的约束规律。
Because of the good state of kernel function, RBF-SVM (RBF-SVM) shows good learning performance in practice, but the parameters in RBF kernel function have a decisive effect on the performance of SVM. The regularity that the performance of RBF-SVM changes with the change is expounded, and the RBF-SVM is introduced into the automatic down recognition system. According to the actual demand of automatic feather recognition system and the performance change rule of RBF-SVM, the selection basis and selection process of the parameters in this system are discussed, and the correlation curve change chart is given. Through the research, we finally get the identification model suitable for this system, so as to improve the overall recognition rate of the system. At the same time, the good characteristics of RBF-SVM and its constrained rules by parameters are also verified.