| Summary: | 碩士 === 中央警察大學 === 鑑識科學研究所 === 100 === Surveillance systems provide a useful tool for police officers to investigate crime cases, such as identifying suspects. However, the imaging angles of cameras of surveillance systems are usually different from the photographing angles of mug shots. It is difficult for police officers to identify suspects directly with those images.
In this thesis, we obtain 3D face models (triangular mesh data) by using a structured light 3D facial scanner. We rotate the obtained 3D face models and obtain the 2D projected images of different angles. We use these projections to compare and identify suspects from surveillance systems. We use feature landmarks to mark facial locations like canthus, alare, chelion, and chin on the 3D and 2D facial images. We calculate the ratio distance between those feature points, and proceed cross-referencing. According to the experiments, the proposed method can identify people effectively and correctly.
|
|---|
