摘要
针对如何高效、准确地从视频图像中提取相关特征向量,完成基于视频的人体运动分析,构建了基于视频信息的人体下肢运动系统。系统包括人体运动轮廓的提取、噪声处理和人体下肢建模及分析3个模块。人体运动轮廓提取中采用改进的光流算法,通过阈值设置改善了轮廓提取的清晰度和完整性。噪声处理模块运用单个中值滤波器与人体四周去噪算法,不仅有效解决了多中值滤波引起的人体轮廓模糊问题,同时使人体活动区域外的噪声去除率达到100%。通过系统分析,视频中人体行走的速度为0.687 m/s,髋关节垂直方向上下起伏幅度为4.71 cm,行走步态正常。
Aiming at how to efficiently and accurately extract relevant feature vector from the video image to complete the analysis of human motion,a video based human lower limb motion analysis system was constructed by means of extracting body contour and modeling human lower limbs. The system includes three modules,namely human motion contour extraction,noise processing and human lower limb modeling and analysis. The human motion contour extraction module uses the improved optical flow method,which improves the clarity and completeness of the extracted contour by threshold settings. The noise processing module uses a single median filter and around de-noising. The results show that the noise process method not only effectively solves the problem of blurry outlines,caused by times of median filtering,but also makes the noise removal rate outside the area of human activity reach 100%. The analysis of the system shows that the walking speed is 0. 687 m / s,vertical motion amplitude of the hip is 4. 71 cm,and the gait is normal.
出处
《微型机与应用》
2016年第12期59-61,66,共4页
Microcomputer & Its Applications
基金
山西农业大学科技创新基金(20142-17)
关键词
轮廓提取
人体建模
噪声处理
运动分析
视频
contour extraction
human modeling
noise processing
motion analysis
video
