Implementation of Real-Time Vision Processing Platform for Human Feature Recognition System

• Main Authors: Jeng-wei Chen, 陳政蔚
• Other Authors: Thong-Shing Hwang
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/63719754229912998269

碩士 === 逢甲大學 === 自動控制工程所 === 97 === The purpose of this paper is to implement a real-time vision processing platform for hunman feature recognition . In order to detect the posture of whole body conveniently , we post feature points on all joints of whole body for recognition easily . The feature points are abstracted and recognized by using image processing method , then to estimate human behavior . In the meanwhile , We use the corresponding position coordinates to drive the 3D/VR for analysis and validation of estimated behavior. First we use CCD camera to capture the image of whole body , detect and recognize the features，and then transform the 2D coordinate to 3D coordinate , Simultaneously , multi-target tracking for all captured features，and save the coordinates for each feature .Through all the coordinates , We can drive the 3D/VR model to walk and complete 3D/VR imitative walk on runway in real-time synchronous demonstration .