摘要
针对传统的运动检测算法不能很好地适应前景中物体运动的问题,设计了一种视频前景运动跟踪和感兴趣区域(ROI)捕获方法.视频预处理使用高斯模糊消除视频中的噪声,然后使用形态学处理操作结合基于Sobel边缘检测算法的优化轮廓搜索算法初步定位和跟踪ROI;随后平滑位置变换以识别目标的运动模式,选择有效的ROI;最后基于肤色状态获得更准确的ROI及其定位.实验结果表明:视频中的行为判断准确率和ROI捕获准确率分别达到77.5%和91.0%,能对ROI进行有效跟踪捕获;在单帧中最耗时模块的运行时间为0.006 s,满足实时性要求.
To solve the problem that traditional motion detection algorithms cannot adapt well to the motion of targets in the foreground,a method for foreground motion tracking and region of interest(ROI)capturing in video was designed.The Gaussian blur was used in video preprocessing to eliminate noise.And then the ROI was preliminary located and tracked using morphological processing operations combined with an optimized contour search algorithm,which was based on Sobel edge detection algorithm.The position transformation was then smoothed to identify the motion pattern of the target,based on which effective ROI was selected.Finally,a more accurate ROI and its position were achieved based on skin state.The experimental results show that the behavior judgment accuracy and the ROI capture accuracy in the videos reach 77.5%and 91.0%,respectively,which can effectively track and capture the ROI.The running time of the most time-consuming module in a single frame is about 0.006 s,which meets the real-time requirements of this paper.
作者
刘政林
朱盛瑜
翟剑坤
张海春
LIU Zhenglin;ZHU Shengyu;ZHAI Jiankun;ZHANG Haichun(School of Optical and Electronic Information,Huazhong University of Science and Technology,Wuhan 430074,China)
出处
《华中科技大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2020年第6期1-5,12,共6页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(61874047)。
关键词
图像处理
运动跟踪
行为判断
肤色检测
感兴趣区域
image processing
motion tracking
behavior judgement
skin detection
region of interest