基于分割模板运动预估的相关跟踪算法  

A Correlation Tracking Algorithm Based on Template Partition Motion Estimation

作　　者：徐一鸣[1] 刘晓利[1] 刘怡昕[2] 

机构地区：[1]南京理工大学瞬态物理国家重点实验室,江苏南京210094 [2]南京炮兵学院,江苏南京211132

出　　处：《兵工学报》2009年第7期879-884,共6页Acta Armamentarii

基　　金：国防预研基金资助项目

摘　　要：为了对常规相关匹配算法实时性能进行提升,提出一种基于分割模板运动预估的相关跟踪算法。采用最小二乘法全区间等距拟合目标运动轨迹,计算出目标当前预估点;将模板图像按块运动估计算法要求分割成宏块;按菱形搜索法在预估点周围区域进行搜索,得到每个宏块的最佳运动矢量;取匹配度最佳的运动矢量对应点为模板的粗匹配点;判断该点所在宏块与搜索区域相对位置关系,决定是否进行精匹配结束搜索,或是按梯度方向建立新的搜索区域。目标跟踪实验证明,该方法比基于全搜索的归一化积相关(NProd)算法其计算时间缩短到3.31倍。A correlation tracking algorithm based on template partition motion estimation was proposed for improving real time performance of the conventional correlation matching algorithms. The implemented steps of the method are that target motion trajectory is fitted with least square total interval isometric algorithm to obtain a target prediction point; according to the requirement of a block motion estimation （BME） algorithm, the template is divided into some macro blocks; the searching process is conducted by using a diamond search algorithm for macro blocks around the prediction point, to get optimal motion vectors of all macro block; the point corresponding to the motion vector with the optimal matching measurement is set as a rough matching point of the template; the relation of relative position between the block with matching point and the searching area is used to decide whether to conduct precise matching search or to construct a new search area in the gradient direction. Target tracking experiment was performed by CCD, image acquisition card of the OK-MIOM type, computer and a controlled automobile remotely. The experimented results show that the calculating time with the proposed method decreases to 3.31 times compared with that of normalized product correlation （NProd） algorithm based on all searching.

关 键 词：信息处理技术 相关匹配 目标轨迹预估 块运动估计 菱形搜索 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]