基于全局运动补偿的多运动目标检测方法研究  

Research on multi-moving object detection algorithm based on global motion compensation

作　　者：王洪斌[1] 赵振涌 顾彪[1] 殷永凯[1] 贾桂敏[1] 

机构地区：[1]天津大学微光机电系统技术教育部重点实验室,天津300072 [2]北京自动化控制设备研究所,北京100074

出　　处：《电子技术应用》2011年第1期110-113,116,共5页Application of Electronic Technique

摘　　要：提出了一种动态背景下的对多个运动目标检测的完整方法。利用基于宏块匹配的六参数全局运动估计方法进行全局运动补偿,有效地消除了摄像机在非稳定运动情况下对目标检测性能带来的不利影响。同时在宏块匹配前进行了预处理,通过预判提取纹理信息丰富的宏块,并在宏块匹配的过程中采用九点十字搜索算法取代传统的三步搜索算法,减少了匹配数据量。通过六参数运动模型计算得到摄像机全局运动参数,然后对运动背景进行全局补偿。补偿后经过后处理提取运动区域。实验证明,本方法能够有效地检测出动态背景下的运动目标。A multi-moving target detection algorithm is completely proposed. Global motion compensation in image sequences is realized to overcome the detrimental effects of target detection performance when camera motion in unstable situation by six-parameter global motion estimation method via template image matching. At the same time the project extracted the texture of information-rich macro block by macroblock pre-sentence,use nine-point and cross search algorithm to replace the traditional threestep search algorithm in the process of image matching.This approach reduces the amount of data matching. Calculating can obtain the camera global motion parameters,and then globally compensate the motion background by using bilinear interpolation method. Before detecting moving targets eventually,there are lots of steps results show that this method can effectively detect the moving target in dynamic background.

关 键 词：动态背景 目标检测 全局运动补偿 九点十字搜索法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]