| Summary: | 碩士 === 國立臺北教育大學 === 資訊科學系碩士班 === 100 === Biometric is used to confirm the identity of an individual. The face has more characteristics to recognize a person’s identity. As we known, the technology of facial recognition is very common. Although facial recognition has been around for more than ten years, the Government has not been able to use the technology in fighting crime because of its wide range of error. This is all about to change.
This thesis is mainly based on the 2D and 3D of facial recognition. Hence the background of image is simplified by using white backdrop and fluorescent light when taking the pictures. There are three parts in this thesis. The first part is the detection of skin color using RGB color code. In order to reduce the colors red and green which are most sensitive to light, the Normalized Color Coordinate (NCC) method is chosen to pick up the correct skin color without the effects of light. Then the correct face range is found by morphological image processing with oval-shaped plate which is similar to the shape of the face and the circle plate which is similar to the shape of the eye. Second, to choose the important characteristics of the face, the Principle Component Analysis (PCA) uses the wavelength separation technique to make 3D images. The third part is about identifying the face. An improved PCA is used through a transfer matrix to get optimal total scatter grid of the face. The optimal total scatter grid represents the eigenvalue of face characteristics. Finally, the recognition rate and process performance between 2D and 3D images are compared via Euclidean Distance. The efficiency and recognition rate of 3D images are superior to 2D images. The recognition rate of 3D images attains to 92% and costs 0.39 second to recognize each image. It has improved 28% compared with the recognition rate of 2D images.
Lastly, this thesis introduces the uses of facial recognition in police agencies, using footage from ATM, market, and other surveillance cameras. It would be very useful except most surveillance cameras have rough quality, making facial recognition unsuccessful in most cases. If the quality is good, the successful rate of a positive match is almost 100 percent. Using this system, we can match unidentifiable persons, bank information, prior records, and many other things
|
|---|
