时空离散度复杂背景运动目标快速检测  

Moving Targets Fast Detection by Space-time Dispersions Under Complex Background

作　　者：王振华[1] 陈杰[1] 窦丽华[1] 

机构地区：[1]北京理工大学信息科学技术学院,北京100081

出　　处：《火力与指挥控制》2008年第10期49-52,70,共5页Fire Control & Command Control

基　　金：北京市重点实验室基金资助项目

摘　　要：复杂背景下运动目标的快速检测一直是目标跟踪领域的难点。灰度投影法可将时空域内的图像序列由三维信号压缩为二维信号来处理,且压缩后的信号噪声仍能通过高斯滤波来消除;之后计算灰度投影向量的时空离散度,并采用水线阈值分割技术查找离散度曲线的波峰,从而确定运动目标所在位置,实现运动目标的快速检测。该方法计算简单,对于摄像机抖动、目标被部分遮挡或背景中景物非完全静止等等复杂背景情况,均能快速准确地检测到运动目标。Moving target fast detection under complex background has been the difficulty in target tracking domain. The intensity projection method can reduce the image sequences from three-dimensional signals to two-dimensional signals, and also the noises of compressed signals still can be eliminated by Gaussian filter. Then the space-time dispersions of intensity projection vectors are computed and the wave crests of dispersion curves are found through watershed segmentation technique. Thus the regions existing moving targets are confirmed and the moving targets are detected rapidly. This method is simple in computing, otherwise, in the circumstances of camera dithering, target being part obscured and the sceneries in background being incompletely still, et al, it can all detect the moving targets quickly and exactly.

关 键 词：复杂背景 运动目标快速检测 灰度投影 时空离散度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]