基于Adaboost和CART结合的优化分类算法

提出了一种基于Adaboost算法和CART算法结合的分类算法。以特征为节点生成CART二叉树,用CART二叉树代替传统Adaboost算法中的弱分类器,再由这些弱分类器生成强分类器。将强分类器对数字样本和人脸样本分类,与传统Adaboost算法相比,该方法的错误率分别减少20%和86.5%。将分类器应用于目标检测上,实现了对这两种目标的快速检测和定位。结果表明,改进算法既减小了对样本分类的错误率,又保持了传统Adboost算法对目标检测的快速性。

Optimization of classification based on combination of Adaboost and CART algorithm

This paper presents a method based on combination of Adaboost and CART algorithm. The method firstly uses CART binary tree as a weak classifier, and then combines these weak classifiers to generate a strong classifier. Compared with the conventional Adaboost algorithm, using the strong classifier in face and digital number classification, the error rates are reduced by 20% and 86.5%. Using the strong classifier on object detection, targets＇ positions are quickly found out in pictures. The resuhs show that the improved algorithm can not only reduce the classification error, but also maintain the rapidity feature in object detection of traditional Adboost algorithm.