Research On Eye Detection And Algorithm Of Eye Location

Since 21 st century,against the backdrop of the continuous development and improvement of computer performance,rapid progress has been made in many fields,such as information processing technology,digital image processing technology,artificial intelligence,computer vision and pattern recognition.And face detection and human eye detection,as one of important research area,have developed so unceasingly as to promote the development of related areas to a large extent,including face recognition,unmanned area,line of sight tracking,medical education and so on.The eye detection and segmentation implemented in this paper are carried out on the basis of face detection,and what's more,the recognition of eye state is carried out on the basis of human eye detection and segmentation.This paper achieved the recognition of the human eye state based on BP Neural Network,making some improvements in traditional methods of face detection and human eye detection,and go deeply into the studies of relevant theories and problems in the areas of face detection,human eye detection and human eye state recognition.The main works carried out in this paper are as follows: Face Detection and Segmentation: This paper,aiming at the brightness threshold issues existing in the traditional skin color model,puts forward an Elliptical Skin Color Model which is adaptive to lightness to carry on segmentation of human face.This method,at first,transforms the face image into the YCbCr space.And then it takes advantages of improved Elliptical Skin Color Model to segment human face.Finally fake skin area in the background is eliminated according to the morphology and area filtering based on connected region labeling.Eye Center Location and Segmentation: Considering the problem that the detection of eye region is not accurate enough as a matter of that the crests and troughs of gray integral projection curve are be susceptible to be affected by noises.In this paper,a method that combines Horizontal Gray Projection and Radial Symmetry Transform is adopted,making for realizing the eye location and segmentation.Experiments prove that the accuracy of the method has been improved a lot.Eye State Recognition Compared with eye state recognition based on geometric features and template matching,eye state recognition based on neural network has a lot higher accuracy.So,in this paper,eye state recognition is carried out on the basis of BP neural network.It has been proved by experiment results that this method has much better recognition rate.