基于DM642的快速目标检测与跟踪算法  

An algorithm for fast detection and tracking of moving targets based on DM642 platform

作　　者：杨毅[1] 王正勇[1] 卿粼波[1] 

机构地区：[1]四川大学电子信息学院图像信息研究所,成都610064

出　　处：《四川大学学报（自然科学版）》2011年第4期839-843,共5页Journal of Sichuan University(Natural Science Edition)

基　　金：四川省科技支撑计划项目资助项目(2010GZ0167)

摘　　要：基于TI DM642硬件平台,设计了运动目标快速检测与跟踪算法.首先改进了RunningAverage背景实时更新建模的方法,消除了原有算法背景更新时产生的运动目标重影.采用背景差分方法检测运动目标,速度较快.使用Otsu算法自适应求取最佳二值分割阈值分割目标,提高检测精度.在此基础上,提出了一种快速的运动目标形状中心线性预测算法,估计运动目标中心,再结合MCD(Maximum Close Distance)高精度模板匹配,进行目标的精确定位.Based on T1 DM642 hardware platform, an algorithm for fast detection and tracking of moving targets has been designed. First, the background modeling of running average background real time updating was improved to remove the target shadows generated in running average algorithm. Then, a background subtraction was applied to detect moving targets quikly, and an adaptive threshold was calculated by Otsu method to segment the targets accurately. After that, an algorithm of fast linear prediction was proposed to estimate the center of the target, and combining with MCD（Maximum Close Distance） template matching algorithm to locate the targets accurately.

关 键 词：DM642硬件平台 目标检测跟踪 最佳分割阈值 形心预测 MCD模板匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]