计算机视频处理技术的运动姿态自动识别研究  

Research on motion posture automatic recognition based on computer video processing technology

作　　者：吴先武 秦菲 WU Xianwu;QIN Fei(Guizhou Normal University,Guiyang 550001,China)

机构地区：[1]贵州师范大学,贵州贵阳550001

出　　处：《现代电子技术》2021年第5期89-93,共5页Modern Electronics Technique

基　　金：贵州师范大学博士课题(11904/0514173)。

摘　　要：受到运动姿态初始视频图像质量的影响,导致传统的运动姿态自动识别方法存在明显的识别误差问题,为此利用计算机视频处理技术实现对运动姿态自动识别方法的优化设计。首先设置运动姿态的识别标准,分别定义不同运动姿态的运动特征向量。通过安装的摄像机设备,捕获运动姿态视频数据;利用计算机视频处理技术,通过运动视频图像提取、视频变换以及补偿运动残差等步骤,实现对初始视频数据的预处理;在此基础上检测运动目标,并从轮廓、运动周期等方面提取运动姿态特征,并得出融合特征向量。通过与设置的识别标准比对,得出最终的运动姿态自动识别结果,经过与传统识别方法的对比发现,设计方法的平均正确识别率提升了6.1%。Obvious recognition error occurs to the traditional motion posture automatic recognition method because of the in⁃fluence of the initial video image quality.Therefore,the computer video processing technology is used to realize the optimization design of the motion posture automatic recognition method.The recognition criteria of the motion postures are set.The motion fea⁃ture vectors of different motion postures are defined respectively.The installed camera equipment is used to capture the video data of the motion postures.The initial video data are preprocessed with computer video processing technology according to steps of moving video image extraction,video transformation and motion residual compensation.On this basis,the moving target is detected,and the motion pose features are extracted from the contour and motion period to obtain the fused feature vectors.By comparing with the set recognition criteria,the final automatic recognition results of motion posture are obtained.The average correct recognition rate of the designed method is increased by 6.1%in comparison with that of the traditional recognition method.

关 键 词：运动姿态识别 视频处理技术 自动识别方法 视频数据预处理 运动目标检测 姿态特征提取 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]