基于改进DRLSE的运动目标分割方法  

Moving object segmentation method based on improved DRLSE

作　　者：胡祝华[1] 赵瑶池[1] 程杰仁[1] 彭金莲[1] 

机构地区：[1]海南大学信息科学技术学院,海南海口570228

出　　处：《浙江大学学报（工学版）》2014年第8期1488-1495,共8页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(61261024;61363071);海南省自然科学基金资助项目(614221);海南省教育厅基金资助项目(Hjkj2013-14);海南大学青年基金资助项目(qnjj1185;qnjj1245)

摘　　要：为了得到精准、连续的运动目标轮廓,提出将帧间差分法和改进的距离规则水平集演化(DRLSE)方法结合应用于运动目标分割.采用帧间差分法得到运动目标的初始轮廓;使用在能量泛函的外力项中融入运动序列时空变化信息的DRLSE方法进行轮廓演化,避免演化受背景边缘干扰,得到精准的运动目标轮廓;根据精准目标轮廓的反馈估计运动方向,并结合帧间差分法为后续的DRLSE提供一个较佳的初始轮廓,能显著提高运动目标的分割速度.实验结果表明,与现有算法相比该改进方法能够更精确、更快地得到运动目标轮廓.In order to get an accurate and continuous contour of moving target, a novel method to segment moving object by combining frame difference and an improved （DRLSE） is proposed. Firstly, the initial contour of moving target was obtained using frame difference. Secondly, the contour was evoluted with an improved DRLSE method, in which the spatial-temporal variable information in video sequences was merged into the external force field of energy functional to avoid the interference from background edge. Finatly, the moving direction was estimated according to the feedback of the obtained accurate target contour, and by combining frame difference a better initial contour could be provided for the subsequent DRLSE, and thus the segmentation speed of moving object could be significantly improved. The experimental results show that the contour of moving target can be obtained more precisely and rapidly with the improved method introduced in this paper than with the existing methods.

关 键 词：活动轮廓模型(ACM) 距离规则水平集演化(DRLSE) 图像分割 帧间差分 运动目标分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]