基于时域和梯度的运动目标识别跟踪算法  

Moving Object Recognition and Tracking Algorithm Based on Time Domain and Gradient

作　　者：韩晓微[1] 岳高峰 谢英红[1] HAN Xiaowei;YUE Gaofeng;XIE Yinghong(School of Information Engineering,Shenyang University,Shenyang 110044,China)

机构地区：[1]沈阳大学信息工程学院

出　　处：《沈阳大学学报（自然科学版）》2019年第5期388-393,共6页Journal of Shenyang University：Natural Science

基　　金：国家自然科学基金资助项目(61503274);辽宁省重点研发计划项目(2018104012);沈阳市双百工程计划资助项目(Z18-5-013)

摘　　要：针对视频存在噪声干扰及现有算法实时性较差的问题,提出了一种新型的基于时域和梯度的运动目标识别跟踪算法.考虑时间域上视频的形成,用帧差法将图片中的区域划分为变化区域和运动区域;依照2个区域的幅度变化对识别算法进行设计,为了消除背景干扰,对形成的运动区域进行均值滤波;在时域基础上增加了梯度分量,对运动区域形成梯度化轨迹,实现跟踪算法的完整设计.通过实验验证,本算法能够稳定地跟踪前景运动目标,与TLD和CamShift算法相比,在实时性上有明显提升,在耗时上分别缩减12.6%和22.7%.Aiming at the problem of noise interference in video and the poor real-time performance of existing algorithms,a new moving object recognition and tracking algorithm based on time domain and gradient is proposed.Considering the formation of video in time domain,the region in the picture is divided into changing region and moving region by frame difference method;the recognition algorithm is designed according to the magnitude change of 2 regions,and the formed moving area is filtered with mean in order to eliminate background interference;the gradient component is added on the basis of the time domain,the gradient trajectory is formed for the moving area,and the complete design of the tracking algorithm is realized.Experiments show that the proposed algorithm can track foreground moving targets stably.Compared with TLD and Camshift algorithms,the proposed algorithm can reduce the time consumption by 12.6%and 22.7%,respectively.

关 键 词：目标跟踪 干扰噪声 时域分析 梯度分量 动态目标 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]