Developing a Visual System for Detection and Tracking of Human Faces

• Main Authors: Hui-Wen Jeng, 鄭惠文
• Other Authors: Cheng-Yuan Tang
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/29801278210090201594

碩士 === 華梵大學 === 資訊管理學系碩士班 === 91 === In this paper, a genetic algorithm was proposed to learn the facial model consisting of eyes, nose and mouth. Based on the facial model, a visual system for detection and tracking human faces was developed. The developed visual system consists of three states, such as change detection, face detection and face tracking. In formulating the genetic algorithm, two-dimensional binary genome was used to encode the chromosomes of the facial images. Two properties, such as the eye position and edge-point preserving, were used to evaluate the fitness. In the experiment, two initial models such as blank model and average edge model were used as the initialization of the population in the genetic algorithms. This facial model was used to improve the detection rate for the face detection. For the face detection, in this paper, the maximum-likelihood head detector (ML-head detector) was used to obtain a few ellipse-like objects as face candidates, the facial model learned from the genetic algorithms was used to verify these candidates, and then the best one is regarded as the detected face. Finally, some experimental results are shown. Experimental results show the facial model learned from the genetic algorithm is helpful for face detection.