|Face recognition; Face detection; PCA; LDA; OpenCV; Digital image processing
|Full text PDF:
Face recognition is one of the most important biometrics in computer vision and it has been broadly employed in the area such as surveillance, information security, identification, and law enforcement. Over the last few decades a considerable number of studies have been conducted in face recognition such as Principal Component Analysis (PCA), Independent Component Analysis (ICA), Linear Discriminant Analysis (LDA), and Elastic Bunch Graph Matching (EBGM), etc. What seems to be insufficient is the research in accuracy of face recognition, it could be affected by the factors such as luminance changes, pose changes, making up, complex backgrounds, head rotation, aging issues, and emotions, etc. This thesis will limit the discussions and concentrate on accuracy problem of face recognition in complex environments. The complex environments are considered as a place with a large number of people such as big office, internet cafés, airport, train and bus station, and casino etc. In these environments, the target human faces for recognizing usually mingle with moving objects. However, the face recognition in complex environments also can be described as the face recognition for several people who might be interested. Therefore, in this thesis, we target only the person (referred as the “Target User”) who is located closest to the camera and stationary. In this thesis, a new scheme is proposed to recognize human faces in such complex environments. The proposed scheme can be split into three phases. The first is Moving Object Removal (MOR). The moving object could be a pedestrian, a vehicle or other moving object. The second phase is face detection which is a technology to locate a human face in a set of images or a video. The Open Source Computer Vision (OpenCV) locates human face features such as those of eyes, ears, mouth, and nose utilizing Viola-Jones algorithm. The final stage is face recognition. If a face is detected, it will be decomposed into PCA components, and then compared to other decomposed images in a face dataset. The objective of this thesis is to propose a new scheme for human face recognition in complex environments so as to improve recognition precision and reduce false alarms. The scheme can be applied to prevent computer users against sitting too long in front of a screen.