Vision is one of the main functionalities of human. For most people, eyes are the premier way to receive the perception from the world. Then, one big problem in implementing artificial intelligence is to let computers “see” the world. Computer …show more content…
The goal of face-detection algorithm in object-class detection is to detect all faces from any format of image data, no matter size, color, location, or other factors which affect the accuracy. The history of face detection is only about 50 years. From 1964 to 1990, face detection is treated as part of common object-class detection. The main function used in face detection is geometric feature based. This technical function focuses much on profile of human face. How to extract information from face profile determines the performance of face detection at that time. Even artificial neural net is once used in detecting human face. Secondary state of face detection is from 1991 to 1997. In this period, face detection becomes a famous field in artificial intelligence. Several significant face detection algorithms and FERET test of face detection algorithms are claimed in this period. In the aspect of business, face detection system also achieved Visionics (Identix) and Facelt system. These algorithms have improved performance under ideal image extracting, optimal object, and small human face database. Third period of face detection is from 1998 to current. In this period, scientists are more focused on how to detect face suffering various …show more content…
These algorithms recognition speed, small memory needed, but low recognition rate. In Principal Component Analysis (PCA) Face recognition, features face recognition method is based on KL transform, KL transform image compression is an optimal orthogonal transformation. High dimensional image space after the KL transform a new set of orthogonal basis, to retain one of the important orthogonal group, these groups may be spanned by a low-dimensional linear space. If we assume that the face has a projection these low-dimensional linear space of reparability, it can be used to identify these projectors feature vector, which is the basic idea Eigen faces methods. These methods require more training samples, and is entirely based on the statistical characteristics of the gray scale image. The Neural Networks can reduce resolution face image, the autocorrelation function of the local area, the local texture. Such methods also need more training samples, and in many applications, the number of samples is very limited. Elastic graph matching method defined in two-dimensional space in a normal face deformation for a certain distance invariance, and the use of property to the representative face topology, topographies any vertex contains a feature vector used to record face information in the vicinity of the