摘要
在提取多运动目标的基础上,利用目标运动的时空连续性,提出一种基于模板匹配的多目标精确分割算法。该算法首先利用混合高斯模型提取视频中的运动目标,然后利用模板匹配算法初步识别出每一个目标的区域,最后利用扩张算法扩充目标的完整边界区域。实验结果表明,与传统的静态粘连分割算法比较,所提出的算法不但能分割同一颜色的多目标视频,而且分割效果更准确更高效。
Based on the extraction of multiple moving objects,using of spatial and temporal continuity of target motion,a multi-objective accurate segmentation based on template matching algorithm is proposed.The algorithm uses Gaussian mixture model to extract moving objects in the video,then uses the template matching algorithm to identify each of the initial target area,the final goal is to use the full expansion algorithm to expand the border region.Experimental results show that the traditional segmentation of static adhesion compares with the proposed algorithm not only multi-objective segmentation of the same color,also the segmentation results are more accurate and efficient.
出处
《计算机与现代化》
2011年第6期22-24,44,共4页
Computer and Modernization
基金
广州市科技支撑计划项目(2008Z1-D291)
关键词
模板匹配
粘连
多目标
分割
template matching
adhesion
multi-objective
segmentation
