Research And Implementation Of Human Eyes Location In Complex Condition

As an important branch of the biometric identification technology, Face recognition, which has a wide range of applications and broad application prospects, is a vital component of pattern recognition. Eyes are the most import and most obvious characteristic of a person's face. They provide more reliable information than nose and mouth, they also have more distinguishing features hence they become necessary treated objects in face recognition. The precise positioning of eyes has important implications; it can enhance recognition correctness, helping to the extraction and registration of facial features.At present there are all kinds of methods for eyes identification, such as horizontal and vertical projection method based on prior rules, binary location method and symmetry method, deformable template method based on geometry shape information, the color information method based on different color spaces, the neural network method based on statistics information and Support Vector Machine approach, the probabilistic net method based on relevance information, the dynamic links modeling method and the wavelet network method and so on. However, at this stage most of the algorithms are under restrict condition such as simple background, frontal face without tilting, appropriate light, no shelter (like glasses), no influence by expression and so on. Assume that the position of face is known or easy to get. On the other hand, the facial features are vary and hard to analysis or extract, has a much more computing complexity.This article reviews various complex conditions of locating human eyes methods. And the corresponding solution is presented. By analyzing, integration and comparison, an improved method of locating human eyes based on complicated condition are put forward. And, several optimization methods are used in this paper. the compute speed of tracking eye is raised too. The precision and the flexibility of the algorithms are further improved.At last, utilizing the eyes identification method of this thesis, This subject has completed a eyes identification system. The experiment results were analyzed using a common face image database. The experimental results demonstrate that the algorithm is accurate, qualified with excellent robustness, and could be applied to practical application.