Summary: | 碩士 === 義守大學 === 資訊管理學系碩士在職專班 === 100 === Extracting human facial features is one of schemes often used in the field of face recognition. The face recognition using facial features primarily utilizes the relative size of facial organs, such as eyes and mouth, and face shape to identify one or more persons by comparing it with faces stored in a database. However, how to select facial features and the number of facial features is a key point of face recognition. The thesis proposes a new set of facial features; besides using the features of the relative size of facial organs, the features of the shape of eyes and mouth approximated by Quadratic curves also are included. In the initial stage of the study, each face is extracted thirty-two facial features; then, grouping the thirty-two facial features for finding the best combination which has the highest recognition rate is executed. However, the number of combination is very large; therefore, for resolving the problem, the study presents a rule to select candidate images of each feature by statistic and decide the result from the candidate images. The rule is based on the assumption: an effective feature should contain the correct result in a small number of candidate images. So, each feature just selects several candidate images. This will be able to save a lot of time. The experimental results show that the proposed method can achieve 86.2% recognition rate.
|