基于融合图像轮廓矩和Harris角点方法的遮挡人体目标识别研究  

Human Body Target Recognition Under Occlusion Based on Fusion of Image Contour Moment and Harris Angular Points

作　　者：杨云[1] 岳柱[1] 

机构地区：[1]陕西科技大学电气与信息工程学院,陕西西安710021

出　　处：《液晶与显示》2013年第2期273-277,共5页Chinese Journal of Liquid Crystals and Displays

基　　金：陕西省教育厅专项科研计划项目(No.2010JK419)

摘　　要：针对遮挡人体目标识别中人体特征描述容易受到遮挡物影响,但人体的头肩部分不易受到遮挡的特点,提出了一种基于人体头肩部位混合轮廓特征的遮挡人体目标识别的新方法。首先利用背景差分法把人体从复杂的背景中分离出来,接着利用人体的先验知识提取出头肩部位,然后提取出头肩部位轮廓的矩不变量和Harris角点特征,然后将2种特征融合构成混合特征,最后将新特征输入BP神经网络进行识别。实验表明:此方法具有较好的鲁棒性和较高的识别率。For the human object recognition under occlusion, the human characteristics de- scription are vulnerably effected under occlusion, but the head and shoulder of the human body are less susceptible to the characteristics of the block, put forward a new method based on the human head and shoulder mixed contour features for human object recognition under occlusion. First, using background subtraction method to separate the human body from the complex background, and then using a priori knowledge of the human body to extract the head and shoulder parts, then the contour moment invariant and Harris angular points are extracted and fused into a mixed feature, finally put the mixed feature into BP neural net-work to recognize. The experiments show that the method has better robustness and higher recognition rate.

关 键 词：遮挡人体识别 矩不变量 HARRIS角点 BP神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]