计算机工程
計算機工程
계산궤공정
COMPUTER ENGINEERING
2015年
2期
242-247
,共6页
运动捕获%关键帧提取%帧间距%球面线性插值%重构误差%视觉概括力
運動捕穫%關鍵幀提取%幀間距%毬麵線性插值%重構誤差%視覺概括力
운동포획%관건정제취%정간거%구면선성삽치%중구오차%시각개괄력
motion capture%key-frame extraction%inter-frame gap%spherical linear interpolation%reconstruction error%visual summary performance
运动捕获数据中有大量的冗余数据,不利于运动数据的压缩、存储、检索以及进一步重用。为此,提出一种基于帧间距的运动关键帧提取方法,提取代表运动捕获数据内容的关键姿势。利用四元数之间的距离表示人体姿态差异,将人体各个关节上的总变化作为帧间距,以运动数据首帧作为第一个关键帧。通过不断计算当前帧同最后一个关键帧之间的差异,消除差异小于阈值的帧,差异超过阈值的帧被当作为新关键帧。对提取关键帧集合采用四元数球面插值方法重构。为表现人体运动特征,在重构误差中引入关节速度分量,用人体姿势误差位置与人体关节运动速率之和表示原始运动与重建运动序列之间的重构误差。实验结果表明,该方法对原始运动既有较高的压缩率,又有较强的视觉概括效果。
運動捕穫數據中有大量的冗餘數據,不利于運動數據的壓縮、存儲、檢索以及進一步重用。為此,提齣一種基于幀間距的運動關鍵幀提取方法,提取代錶運動捕穫數據內容的關鍵姿勢。利用四元數之間的距離錶示人體姿態差異,將人體各箇關節上的總變化作為幀間距,以運動數據首幀作為第一箇關鍵幀。通過不斷計算噹前幀同最後一箇關鍵幀之間的差異,消除差異小于閾值的幀,差異超過閾值的幀被噹作為新關鍵幀。對提取關鍵幀集閤採用四元數毬麵插值方法重構。為錶現人體運動特徵,在重構誤差中引入關節速度分量,用人體姿勢誤差位置與人體關節運動速率之和錶示原始運動與重建運動序列之間的重構誤差。實驗結果錶明,該方法對原始運動既有較高的壓縮率,又有較彊的視覺概括效果。
운동포획수거중유대량적용여수거,불리우운동수거적압축、존저、검색이급진일보중용。위차,제출일충기우정간거적운동관건정제취방법,제취대표운동포획수거내용적관건자세。이용사원수지간적거리표시인체자태차이,장인체각개관절상적총변화작위정간거,이운동수거수정작위제일개관건정。통과불단계산당전정동최후일개관건정지간적차이,소제차이소우역치적정,차이초과역치적정피당작위신관건정。대제취관건정집합채용사원수구면삽치방법중구。위표현인체운동특정,재중구오차중인입관절속도분량,용인체자세오차위치여인체관절운동속솔지화표시원시운동여중건운동서렬지간적중구오차。실험결과표명,해방법대원시운동기유교고적압축솔,우유교강적시각개괄효과。
To solve the problem that the motion capture data has a large number of data redundancy,this paper proposes a key frame extraction method based on inter-frame pitch. In order to compress, storage, reconstruct and further reuse motion data,key-frame is needed to be extracted which represents the content of the motion. Quaternion is introduced to represent the difference between two rotations. The distance between two frames is defined by the total rotation differences and first frame is regarded as the first key frame. Then,calculate the difference between the current frame and the last key frame continuously. The frame is eliminated when the difference is smaller than the set threshold or the opposite is reserved for the new key frame. Spherical linear interpolation is used to reconstruct the sequence. To express the characteristics of human motion, the joint velocity is introduced. Reconstruction error is defined by the human body posture error of position and the motion speed error between the original frame and the reconstructed frame. Experimental result demonstrates that the original motion capture can be compressed in a high ratio and gives a good visual summary performance.