基于Floatboost算法的人眼定位

Eye Location Based on Floatboost Algorithm

Floatboost算法相较于传统的Adaboost算法、Boosting算法具有计算量少和运算速度快的优点,目前在机器学习中应用广泛。针对人脸与人眼精确定位问题,采用Haar-like提取人脸特征,再运用Floatboost学习算法得到人脸检测器,在人脸准确定位的基础上使用Floatboost训练的人眼分类器定位人眼。Floatboost算法具备回溯性的优点,去掉了较弱特征,减少计算量,提升了运算速度。

Compared with the traditional Adaboost algorithm and Boosting algorithm, Floatboost algorithm has such features as lowcomputational complexity and high operational speed,and now it is widely used in machine learning.ln this paper,aiming at face and eyepinpoint,the face feature extraction is implemented by using Haar-like,a face detector is obtained by using Floatboost learning algorithm,andthe eye classification implements eye location by using Floatboost training based on face accurate location. Floatboost algorithm has theadvantage retrospection ,which removes the weaker features ,reduces computational complexity and enhances operational speed.