摘要
近年来基于Adaboost的人脸检测算法因其快速和可接受的检测率得到了成功的应用,但Viola-Jones学习算法需要对级联分类器的每一个特征反复训练弱分类器显得非常缓慢。本文给出了一种新的级联检测器节点分类设计方法,首先将每个节点所有弱分类器的训练移到循环外,然后选择使强分类器有最小错误率的特征集代替选择单个最小加权误差的特征生成强分类器。实践表明该训练速度快于Viola-Jones的方法。
Recently the human face detection system based on Adaboost is successfully used in application areas because of its high speed and accepted detection rates, but the Viola - Jones learning algorithm, in which the weak classifiers are retrained once for each feature in the cascade, is very slow. This paper presents a new methor to design node classifiers in the cascade detector. First, in our method training all weak classifiers are moved out of the loop per node, then we select the features that the ensemble classifier has the smallest error rate instead of choosing a single feature with the smallest weighted error. Experimental resuits prove that our training speed is faster than Viola- Jones's methor.
出处
《嘉应学院学报》
2007年第3期85-89,共5页
Journal of Jiaying University