Eye detecting is an important part of computer face recognition. Eye detecting is vulnerable to impact of different gesture and light under complex environment. In order to solve the complex environment of the human eye positioning problem, use a Based on Adaboost and gray-scale information of the human eye detection algorithm. Firstly, collecting different eye samples carries out the balance of gray. From a database select a small number of Haar-like features produce a strong and efficient classifier. Secondly, using a method of cascade constitutes a more complex cascade classifier. And then, using gray-scale threshold means carries out preprocessing. At last, applying a classifier carries out eye detection.