摘要
常用的几种背景提取算法在车流量较大的情况下提取的背景效果较差。在某些目标检测区域较少的场景中,若将所有像素进行检测,会浪费许多时间。针对这些问题,提出一种新的背景提取算法。先将视频帧进行分割,再对分割出的检测带依次进行车辆存在检测,最终自动选取视频中没有车辆的车道块并将其拼接成完整背景帧,最后利用Lab空间色度与亮度相互独立的特性提取目标。该算法能够充分提取前景图像,不会丢失车辆目标。相比传统算法,该算法准确性较高。
The commonly-used background extraction methods may not bring about satisfactory effects in the case of heavy traffic,because a large amount of time is consumed when all of the pixels need to be detected,even though less object detection areas exist in images.To solve the problems,we propose a new background extraction method.Firstly,the video frames were divided into blocks.Secondly,vehicles were detected in those blocks through a series of processes.Then,frame lane blocks,with no vehicles detected,was selected automatically to constitute the complete background frame.Finally,objectives were extracted by using quality of independence of brightness and color in Lab color space.This method could fully extract foreground without losing vehicles and has higher accuracy comparing with other traditional methods.
作者
巨志勇
彭彦妮
JU Zhi-yong;PENG Yan-ni(School of Optical-Electrical and Computer Engineering University,Shanghai 200093,China)
出处
《软件导刊》
2018年第5期183-186,190,共5页
Software Guide
基金
国家自然科学基金资助项目(81101116)
关键词
背景建模
区域分割
Lab色彩空间
背景相减
形态学处理
运动目标检测
background modeling
region segmentation
Lab color space
background subtraction
morphological processing
moving object detection