摘要
视频图像中运动物体分析关键的一步就是从连续的视频图像中提取出运动目标 ,即运动目标检测。传统的运动目标检测方法有 3种 :背景图像差分法、时态差分法和光流法 ,分析比较了它们的优缺点 ;在此基础上笔者采用了一种结合Sobel算子和自适应背景差分算法的运动目标检测方法 ;利用Intel公司开发的计算机视觉库OpenCV开发了一个软件 ,并进行了一系列实验。通过观察、比较实验结果可以看出 ,这种运动目标检测算法能够正确地检测出视频图像中的运动目标 。
The vital stage of motion analysis in video images is to detect moving targets from the video images. There are three conventional approaches to detect moving targets: background differencing, temporal differencing and optical flow. Their features and drawbacks are analyzed and compared. An adaptive background subtraction algorithm combined with the Sobel operators is adopted, and a demonstration software using Intel's free computer vision library OpenCV is developed and applied to a series experiments. The experiment results show that this algorithm can easily detect moving objects and the performance in detection is better than the simple adaptive background differencing.
出处
《武汉理工大学学报（信息与管理工程版）》
CAS
2004年第4期1-3,17,共4页
Journal of Wuhan University of Technology：Information & Management Engineering
基金
国家自然科学基金资助项目 ( 60 2 75 0 40 )
关键词
视频图像
目标检测
背景差分
自适应算法
video images
object detection
background differencing
adaptive algorithm
