摘要
提出一种基于D-S证据理论的多特征融合人体检测算法。该方法首先使用harr-like特征方法和HOG特征方法提取人体特征,然后利用后验概率密度函数,估计出两个基本概率分配函数,最后应用D-S证据理论中的合成法则,将两个特征的分类器进行融合决策。实验结果表明,本文提出的算法在Inria数据集上取得了较好的效果,满足了一般工程应用的要求。
This paper puts forward a multi -feature human detection algorithm based on the theory of D -S. This method firstly extract the characteristics of human body by using hart - like characteristics method and HOG characteristics method, then estimates two basic probabil- ity assignment function by use of the posterior probability density function, finally by combina- tion rule in evidence theory of the application of D-S makes a fusion decision towards the classi- fier with two features, proposed one kind based on the D-S evidence theory fusion of multiple features of human detection algorithm. The method firstly uses the Harr-like method and HOG method to extract characteristic features characteristic of the human body, and then use esti- mated two, finally the application of D-S in, two feature classifier fusion decision. The experi- mental results show that the proposed algorithm in this paper obtains the better result in Inria data sets and meets the general requirements of Engineering application.
出处
《吉林工程技术师范学院学报》
2012年第8期73-76,共4页
Journal of Jilin Engineering Normal University
