Study On An Algorithm For Eye Detection Under Complicated Environment

The study about the characteristics of target detection and location has been a hot issue in the range of the digital image processing and machine vision problem in recent years. For the eye that is the most notable organ in the face, the corresponding detection and location have the very wide application, such as the face recognition, the human-computer interaction, the medical science, the traffic safety and the intelligent robot, and so on. Based on the current rough algorithm, the realization of the finer and more efficient face detection and location algorithm is necessary.At present, There are a lot of algorithms of the eye detection, and each has its advantages. But still have eyebrows or eyelashes can be located in practical application, and eye location accuracy is not enough. Therefore, this article focuses on face detection and eye classifier trained in-depth study, and analyze the subtle features of eyes under the eyebrows,and design an efficient new precise positioning algorithm to adapt complex conditions. In this paper, a large number of experiments are done in the human eye classifier training, the human eye local information extraction and the effect of algorithm.The effect of detection algorithm is also compared with the latest eye location algorithm to analyze the performance of the algorithm. Finally, realized the real-time accurate location of human eyes under the camera.In this paper, the innovation points are as follows:1. The further study on the structure characteristics of cascade classifier and the selection of training samples based on Adaboost algorithm, training out of a multi state human eye classifier adapt to the complex environment. The characteristics of the classifier is to join their eyebrows or eyelashes features such as obvious negative samples, and can quickly determine when cascading detection eyebrows or eyelashes features as a goal, solve the problem of the eyes to their eyebrows or eyelashes. The experimental results show that add classifier with obvious features of negative samples can get the more effective coarse position of the eye area.2. Put forword an algorithm for improving eye detection based on Adaboost. Algorithm has a deep analysis of the coarse positioning human eye region in the range of human face, give full consideration to the structural features of the gray value mutation of the eye area and the boundary region of the eye area, find the coarse position eye area of the largest connected domain, and calculate the center coordinates for the precise location of human eyes. Algorithm analyzes the attitude under different illumination, low quality of image, the adaptive threshold method of precise positioning to the human eye, the results show that the algorithm enhances the precision of precise positioning, the correct recognition rate and the efficiency has a very good performance.