低照度视频序列运动目标的检测与提取  

Moving Object Detecting and Extracting in Video Sequences with Low Illumination

作　　者：欧扬[1] 张宗杰[1] 

机构地区：[1]苏州科技学院电子与信息工程系,江苏苏州215011

出　　处：《微计算机信息》2005年第12S期91-92,118,共3页Control & Automation

基　　金：建设部建设领域信息技术科技项目(05-k6-11)

摘　　要：由于低照度环境下的成像质量存在比较突出的问题,使得低照度视频序列的运动目标检测与提取成为一项相当困难的工作。本文结合运动信息和梯度信息,提出了一种新的低照度视频序列运动目标检测与提取方法。该方法首先经帧间差分、滤除噪声得到运动区域的初始检测模板,针对初始检测模板中由于照度过低出现的目标漏检现象,采用提取函数法进行低灰度值的运动区域检测,最终形成完整的运动区域检测模板。采用多尺度形态梯度算子进行边缘检测,这种梯度算子抗噪能力强。实验结果表明这种方法能有效地实现低照度视频序列运动目标的检测与提取。Imaging under low illumination exists outstanding problems, so moving objects detecting and extracting in video sequences with low illumination becomes a rather difficult study subject. Through integrating motion information with gradient information, one novel method of moving objects detecting and extracting in video sequences with low illumination is proposed in this paper. Basing on inter-frame difference and filtering noise, initial detecting template is formed at first. Because partial objects may be lost in initial detecting template due to low illumination, an extracting function is used to detect moving regions with low gray level, finally whole moving region detecting template is formed. Multi-scale morphological gradient operator is used to detect the edges. This operator is robust to noise. Exoerimental results demonstrate the performance of our algorithm.

关 键 词：运动目标检测 图像分割 梯度图像 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]