Human eye positioning and human eye state recognition method

Abstract

The invention discloses an eye positioning and eye state recognizing method, wherein first setting an self-adaptive threshold value to pretreat a face image, according to the face geometry character, designing an eye classifier, positioning eye initially, for some edge images, based on the symmetry face and eye, accurately positioning eye, thereby providing a method for positioning face symmetry axis based on lip auxiliary positioning, then dividing one eye from a positioned eye binary picture, according to the black pixel character of pupil, judging the eye state. The inventive method has real-time property, with the application for different backgrounds, lights, rotation and bias angles, and face details or the like to position eye and recognize eye state.

Description

technical field [0001] The invention belongs to the application field of image processing and pattern recognition technology, in particular to a human eye positioning and human eye state recognition method. Background technique [0002] Face detection and localization of facial organs is one of the most challenging research topics in the field of computer vision. As the most prominent feature of the human face, the human eye can provide more reliable and important information than the mouth and nose, so it is often the necessary processing object in face recognition. Human eye recognition plays an important role in face image recognition applications, such as face recognition, facial expression analysis, posture determination, visual tracking, human-computer interaction, fatigue detection, etc.; With precise positioning, other features such as eyebrows, nose, and mouth can be positioned more accurately based on the potential distribution relationship; eye positioning can al...

AI Technical Summary

Problems solved by technology

After summarizing these research methods, there are mainly the following categories: (1) template matching, which is to use the eye template to match in the image, and get the point with the largest similarity as the eye for positioning. This method has a large amount of computation and is difficult to use in real time. Eye tracking, and the weighting factor in the template is difficult to set reasonably, it is difficult to take into account the faces under all conditions at the same time, and the accuracy rate is low

(2) Edge extraction, which locates the human eye by detecting the circular feature of the pupil or the elliptical feature formed by the eyelid. This method has high requirements for the pixel accuracy of the human eye area, and when the eyes are closed, the edge cannot be extracted to make the detection fail

(3) Gray scale distribution, according to the characteristics of the eye area: the eye area is darker than the surrounding area, that is, the gray value is low; the gray scale change rate of the eye area is large, and the horizontal and vertical gray scale is performed on the face Projection is used to segment the eyes. This method has a fast positioning speed, but the distribution of peaks and valleys is very sensitive to different faces and posture changes, and the positioning accuracy is poor.

(4) Blink detection, through the difference between image frames, according to the change of eye shape to capture the blink, so as to locate the human eye, but the relationship between the blink frequency and the interval between frames is not easy to determine, and it is greatly affected by external interference

(5) Face structural features, the most commonly used is to locate the human eye by searching for symmetrical blocks in the face image according to the strong symmetry of the center of the human eye. This method is suitable for frontal standard faces with a single background, and the scope of application is narrow

(6) Neural network, this method can only give the approximate position of the eyes, and the point-to-point detection mechanism makes the calculation amount very large

Human eye state recognition based on feature analysis. The eye state is mainly determined by the inner and outer canthus, upper and lower eyelids, iris and sclera. There are three typical methods: grayscale template matching method, iris and eye white extraction method, Hough transform Pupil detection method, which requires high pixel accuracy in the human eye area, and the scope of application is limited

The human eye state recognition based on pattern classification is to judge the human eye state according to the method of automatic learning rules or knowledge of samples. At present, there are mainly methods such as characteristic eyes, neural network, SVM, HMM and so on. Such methods generally need to zoom in on the image, Complicated normalization processing such as rotation requires a large amount of calculation, and also has certain requirements for pixel accuracy.

Images