基于SPN-PF模型的行人跟踪算法改进  

Improved Pedestrian Tracking Algorithm based on SPN-PF Model

作　　者：张婧[1] 田宏[1] 

机构地区：[1]大连交通大学软件学院,辽宁大连116028

出　　处：《大连交通大学学报》2015年第1期101-104,共4页Journal of Dalian Jiaotong University

基　　金：国家自然科学基金资助项目(61074029)

摘　　要：结合粒子滤波和神经网络方法提出了一种新的自适应的行人跟踪改进算法.根据行人跟踪经常出现的遮挡,失跟,误跟等问题,提出了一个SPN-PF模型,通过Sigma-Pi网络将行人的多个特征联系在一起,经过学习和计算,进一步讲结果运用到粒子滤波方法中,从而达到可靠准确的跟踪行人目标的目的.实验结果表示该文提出的方法能够更准确的对行人进行跟踪.A self-adapting algorithm is presented for pedestrian tracking, combining particle filter with neural network, and a SPN-PF model is proposed according to the problems such as covering, missing, wrong tracking and so on. The model integrates different features by means of a Sigma-Pi network. After the study and calcula- tion, the result will be used in particle filter method to achieve the goals of reliable and accurate pedestrian tracking. Experimental results show that the purpose can be achieved.

关 键 词：行人跟踪 粒子滤波 神经网络 级联分类器 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]