Real-time 3D Point Cloud Face Modeling and Recognition System Based on Profile and SIFT 3D

• Main Authors: Jyun-You, Lin, 林俊佑
• Other Authors: Chueh-Wei, Chang
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/4zzhkf

碩士 === 國立臺北科技大學 === 資訊工程系研究所 === 105 === In recent years, because of the evolution of 3D depth sensor and software technology, not only the cost of hardware becomes cheaper, but also the depth sensor gets the acquisition of 3D image data easier and faster. The 3D face data is completely presented the real human face, and it has better efficacy to identify. But because of the large amount of 3D data, it affects the real time of recognizing. Therefore, this paper proposes the real-time 3D face modeling and recognition system. On 3D face modeling, the system collects 3D face data based on 3D scanning by depth sensor. The 3D data is converted to 3D point cloud dataset by PCL (Point Cloud Library). In order to reduce the computation of identify, doing the Filtering, Normalization, Keypoints Extraction, Face Profile Extraction, and other pre-processing. According to the experimental result, the 3D face recognition system has more than 90%success rate. It not only immediately identifies the correct face, but also decreases misjudgement on detecting the 2D face image.