摘要
运动图像幅度准确检测是提高对人体运动目标图像远程识别和跟踪的基础。传统方法中对人体运动图像的幅度检测采用运动帧补偿的运动幅度检测方法,当图像处于快速运动状态出现单帧视觉误差时检测性能不好。提出一种基于单帧视觉差的运动幅度检测算法,并进行仿真实现。进行快速运动图像的视频帧图像采集与预处理,采用基于单帧视觉差分析的方法进行快速运动图像的幅度特征提取,通过电子稳像方法进行防抖动处理,得到运动图像的目标函数运动幅度特征点,构造人体运动的位姿表换微分方程,进行单帧运动图像全局运动估计,提高了对运动幅度的检测。仿真实验分析得出,该算法具有较好的快速运动幅度检测性能,对于运动剧烈的Foreman序列,系统性能提高了1~2 d B,对于运动缓慢的序列Claire,系统性能提高了2.5~5 d B,展示了算法的优越性。
Motion image detection based on amplitude accuracy is to improve the human motion target recognition and tracking of image distance. The traditional method uses frame motion compensation range of motion detection method, when the image is in rapid motion state of single frame visual error, detection performance is not good. Motion amplitude of the single frame detection algorithm is proposed based on visual difference, and carries on the simulation to achieve. Fast moving image frames of the video image acquisition and preprocessing, the amplitude characteristics of fast moving image method based on single frame visual difference analysis of extraction, through the electronic image stabilization method for anti jitter, the objective function range of motion feature points of moving images is obtained, the structure of human motion pose changing differential equations of single frame image motion global motion estimation, improve the detection of motion range. Simulation and experimental analysis, the algorithm is fast motion amplitude better detection performance for Foreman sequence of vigorous exercise, the system performance is improved by 1~2 d B, the sequence of Claire slow motion, the system performance is improved by 2.5~5 d B, it shows the superiority of the algorithm.
出处
《科技通报》
北大核心
2015年第4期52-54,共3页
Bulletin of Science and Technology
关键词
帧
视觉差
运动图像
幅度检测
frame
visual difference
image motion
amplitude detection
