欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
在应用AdaBoost算法的人脸检测中,针对训练时间太长的问题,提出一种基于特征值空间划分的改进型AdaBoost快速训练算法,调整了弱分类器的评价系数.在MIT-CBCL人脸和非人脸训练库上对算法进行了实现,实验结果显示改进后的AdaBoost算法简化了训练过程,训练速度提高16倍以上,而且以区间检测代替特定样本的特征单点检测,泛化能力更好,鲁棒性强,检测精度更高.
英文摘要：
It would cost very much time in face detection using AdaBoost algorithm to train. RC-AdaBoost algorithm based on feature-value-division is developed in this paper. The value coefficients of weak classifiers are modified in the new algorithm. RC-AdaBoost algorithm could make training faster and works better. Experimental results on MIT-CBCL face ＆ nonface training data set illustrate that the improved algorithm could make training process convergence quickly and the training time is only one of 16 like before. Experimental results also show that the robustness and detection precision are both exceed the corresponding method.
同期刊论文项目
同项目期刊论文
期刊信息
位置：