序列图像运动目标的新型快速检测方法  

Novel Fast Moving Target Detection Method in Image Sequences

作　　者：李春华[1,2,3,4] 李宁[2,3,4] 吴立潮[1] 

机构地区：[1]中南林业科技大学理学院,长沙410004 [2]北京师范大学地表过程与资源生态国家重点实验室,北京100875 [3]北京师范大学民政部-教育部减灾与应急管理研究院,北京100875 [4]北京师范大学环境演变与自然灾害教育部重点实验室,北京100875

出　　处：《小型微型计算机系统》2014年第8期1916-1921,共6页Journal of Chinese Computer Systems

基　　金：国家重大科学研究计划项目(2012CB955402)资助;科技部国际科技合作项目(S2012GR0231)资助;中国博士后科学基金项目(20110490304)资助;国家林业公益性行业科研专项(201104016)资助

摘　　要：运动目标检测是计算机视觉运动分析的基础,并有着广泛的应用领域.为结合帧差法运算速度快和光流法运动目标检精度高的特点,提出一种基于改进帧差法和双向光流法相结合的快速运动目标检测算法.通过改进的三帧帧差检测法,准确的获取运动目标区域.利用提出的双向光流法,在光亮度不连续及干扰的情况下,准确地提取运动目标区域特征点的光流,并对光流矢量采用阈值标注,进而检测出运动目标.实验结果表明,该方法可以有效的改善目标检测的准确性和检测速度.Moving object detection is the basis of computer vision and motion analysis, and has a wide range of application areas. To combine the characteristics of computing speed by the frame difference method and high target detection accuracy by optical flow algo- rithm, a novel fast target detection method on the basis of improved three-frame difference detection method and bidirectional optical flow algorithm is proposed. According to improved three-frame difference, the region of the moving target is extracted accurately. By the proposed bidirectional optical flow algorithm, the optical flow value of the moving target in the region is obtained exactly under the condition of the lightness discontinuity and interference. Then, the moving target can be detected by the threshold value of optical flow vector. Experiment results show that the method can improve the accuracy and computation of the moving target detection.

关 键 词：双向光流法 改进三帧差分法 运动目标检测 帧差偏移 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]