보행자의 2차원 영상으로부터 3차원 자세 복원 연구

다국어 초록 (Multilingual Abstract)

Gait is important motion that contains biological information. Individual gait pattern is determined by subject’s habit, body type and diseases. Motion capture system is used to measure joint motion but is limited to capture within a laboratory. Optical camera is portable, simple to configure and take photographs of movement of a person, while it produces only two-dimensional information. The purpose of the study was to estimate three-dimensional pose and joint angles from two-dimensional sequential images of walking motion. Three-dimensional trajectories of body markers from three subjects and five motions were pooled together and their mean pose and principal components were obtained. One of the motion capture data was projected to a camera plane by assuming a fixed camera position. Optimizations to calculate a weight matrix of the principal components of body pose were performed with and without a constraint of body segment length to match the plane projected marker positions. Mean three-dimensional errors (standard deviation) of estimated marker trajectory were calculated using measured marker trajectory. Three-dimensional errors with and without constraint of body segment length were 3.44(±0.65) cm and 3.25(±0.80) cm, respectively. Two-dimensional errors (standard deviation) of projected marker to image were calculated using original image. Two-dimensional errors were 1.63(±0.41) pixel and 1.40(±0.38) pixel respectively. Flexion and extension angle of knee was estimated as RMS error, 2.54degree and 2.28degree respectively.

목차 (Table of Contents)

• Abstract
• 1. 서론
• 2. 방법
• 3. 결과
• 4. 토의
• 참고문헌