基于图切割和阴影抑制的目标检测算法  

Object Detection Based on Graph Cuts and Shadow Elimination

作　　者：李芬兰[1] 李超[1] 

机构地区：[1]汕头大学工学院电子系,广东汕头515063

出　　处：《汕头大学学报（自然科学版）》2009年第2期56-62,共7页Journal of Shantou University：Natural Science Edition

摘　　要：针对目标检测中背景噪声影响检测精度的问题,提出在HSV高斯背景模型下,使用最小图切割算法,并结合连通性,找到最大的前景连接区域,最后使用阴影抑制技术完成目标分割.实验结果表明,算法有很好的鲁棒性,能够获得清晰、光滑的前景目标轮廓.In order to solve the problem of detection precision, which is affected by the background noise, an algorithm which is based on the HSV Gaussian background model is proposed, and the minimum graph cut algorithm is used to segment objects. It is combined with the pixels connectivity to find the largest connected component. The shadow eliminating method is used to finish the segmentation. Experimental results show the robustness of the method. Clearer and smoother contour of the objects are obtained.

关 键 词：图切割 阴影抑制 目标检测 高斯背景模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]