复杂场景下遗留物体的检测

摘要

研究安全视频监控系统,对异物或是危险品的遗留检测尤为重要.为了有效的解决了光照突变和阴影所引起的虚警问题,提出了一种基于多条件判断的检测方法.采用从帧间差分图像中提取出运动物体的轮廓特征,针对光照突变这一特殊情况提出了全局亮度变化的检测方法,引入了图像的不变矩特征实现对区域图像的显示.基于上述研究设定报警条件,实现报警功能.对遗留物体检测的平均虚警率为零,平均漏警率为2.72%.实验结果表明,算法能够满足实时性监控报警的要求,且很好地解决了光照突变和阴影等情况下的虚警问题.%It is especially important to detect the foreign matter or dangerous goods in Security Vedio Monitoring system.in this paper, to effectively solve the false alarm problem caused by light mutation and shadow, a detecting method based on multi-condition judging was proposed.Firstly, contour feature was extracted from inter-frame differential images.Secondly, detecting method based on global light change was proposed.Thirdly, constant momentum feature of images was quoted to describe region image.Last, alarm factors based on the previous researches were determined to realize alarm function.Totally the average of false alarm is zero and the lost alarm is 2.72 percent in three conditions.The result shows that this algorithm can fill the condition of real-time alarm, and false alarm on light mutation is effectively solved.