基于色彩相似度的前景有效提取算法研究  

Research on Effective Foreground Detection Algorithm Based on Color Similarity

作　　者：张国家[1] 左敦稳[1] 黎向锋[1] 史晨红[1] 

机构地区：[1]南京航空航天大学机电学院,江苏南京210016

出　　处：《机械制造与自动化》2014年第3期20-23,共4页Machine Building & Automation

基　　金：国家自然科学基金资助项目(51175255)

摘　　要：提出一种将辅助背景滤波与色彩相似度检测相结合的前景有效提取算法。该算法首先运用背景差分构建辅助背景来描述环境噪声分布,并用辅助背景对疑似前景区域进行空域滤波,以增强算法对背景变化的适应能力;然后建立色彩相似度作为前景目标评价标准,进行前景目标检测;最后对检测结果进行形态学操作,得到比较完整的前景目标。多场景实验表明,该算法检测效果良好,具有较强的鲁棒性,对阴影和光照变化具有很好地抑制作用;同时实时性好,空间复杂度低。This paper proposes on effective foreground detection algorithm that the spatial filter is combined with the color similarity. The background difference is used for building the auxiliary background in the algorithm to describe the ambient noise distribution and the auxiliary background is used for performing spatial filter to improve its adaptability of background changes. Then the color semblance is set as the foreground evaluation criterion to detect the foreground regions. After foreground detection, the morphologic operation is dane to gain the complete foreground regions. Experimental results of multiple scenes show that this algorithm not only has robustness but also good inhibition of shadows and il umination changes and its efficiency is high and its complexity is low.

关 键 词：手势识别 背景差分 辅助背景 空域滤波 色彩相似度 前景提取 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]