摘要
人眼检测在表情识别和人脸识别中起着非常重要的作用,作为一种预处理的手段,人眼检测和定位可以有效地提高表情识别和人脸识别的识别率。提出了一种基于Adaboost算法的实时人眼状态检测的方法。Adaboost是一个构造准确分类器的学习方法。它将一簇弱分类器通过一定的规则结合成为一个强分类器,再把这些强分类器级联成为一个快速、准确的分类器。分析和讨论训练阶段不同的人眼特征选择对最终检测的影响,并实验测试各种特征方法对特定目标的检测率,给出一个理想的分类器。
As a preprocess method, eyes detection plays a very important role in the facial expression recognition and face recognition. It could improve the recognition rate in both regions. This paper presents a method based on Adaboost algorithm for eyes state detection. Adaboost is a learning algorithm for constructing accurate classifiers. It can construct a strong classifier by combining a series of weak classifiers through some rules, and a fast and accurate cascaded classifier could be obtained by cascading existing strong classifiers. In this paper, several eyes features in the eyes state detecting process are analyzed and discussed,the detection rates of all the methods are compared, and the best feature method is chosen to construct the classifier.
出处
《计算机仿真》
CSCD
2007年第7期214-216,341,共4页
Computer Simulation
关键词
自举算法
目标检测
检测率
Boost algorithm
Object detecting
Detection rates