Abstract:

In order to enhance the ensemble ability of the traditional Adaboost algorithm, an improved Adaboost algorithm was proposed, which was based on the correlation of classifiers. In the algorithm, the correlation estimation of classifiers was added in the weak training classifiers. Every weak classifier was related not only to the current classifier, but also to previous classifiers as well. The experimental results in Carnegie Mellon University (CMU) show that the algorithm is of better detection rate and lower false alarm rate, compared with traditional Adaboost algorithm.

Key words: face detection, correlation of classifiers, Adaptive Boosting (Adaboost) algorithm

摘要：

为了提高传统Adaboost算法的集成性能,提出一种基于分类器相关性的Adaboost算法。该方法在弱分类器的训练过程中加入分类器的相关性判定,使每一个弱分类器的生成不仅与当前分类器有关,而且与前面若干个分类器相关,并将由此生成的弱分类器组合成新的强分类器。在CMU正面人脸检测集上的仿真结果表明,较传统的Adaboost算法,基于分类器相关性的Adaboost人脸检测算法具有更好的检测效率,同时降低了误检率。

关键词: 人脸检测, 分类器相关性, 自适应提升算法