| 基于代价敏感性AdaBoost的皮肤分类器 |
| |
| 引用本文： | 赵健成,赵逢禹.基于代价敏感性AdaBoost的皮肤分类器[J].教育技术导刊,2020,19(3):31-34. |
| |
| 作者姓名： | 赵健成 赵逢禹 |
| |
| 作者单位： | 上海理工大学 光电信息与计算机工程学院，上海 200093 |
| |
| 基金项目： | 国家自然科学基金青年项目（61402288） |
| |
| 摘 要： | 为解决不同光照条件下皮肤难以检测的问题，提出一种基于代价敏感性CS-AdaBoost算法的皮肤分类器。通过对皮肤像素提取6个基于亮度值的像素特征，并循环选取特征，使用基于CS-AdaBoost算法程序训练最佳弱分类器，通过对所有最佳弱分类器的加权线性组合得到最终的皮肤分类器。由于在算法程序中引入了代价因子θ，使分类结果偏向总错分代价较小，即提高了皮肤样本的分类正确率。使用SDD皮肤数据库评估该皮肤分类器性能，结果表明，该皮肤分类器分类正确率达到了85%，比传统皮肤分类方法提高了5%。
|
| 关 键 词： | 皮肤分类器 CS-AdaBoost 亮度特征 |
| 收稿时间： | 2019-04-21 |
A Skin Classifier Based on Cost_Sensitive AdaBoost |
| |
| ZHAO Jian-cheng,ZHAO Feng-yu.A Skin Classifier Based on Cost_Sensitive AdaBoost[J].Introduction of Educational Technology,2020,19(3):31-34. |
| |
| Authors: | ZHAO Jian-cheng ZHAO Feng-yu |
| |
| Institution: | School of Optical-Electrical and Computer Engineering，University of Shanghai for Science and Technology，Shanghai 200093，China |
| |
| Abstract: | In order to solve the problem of difficult skin detection under different illumination conditions， a skin classifier based on cost-sensitive CS-AdaBoost algorithm is proposed. The skin classifier extracts six pixel features based on luminance values for skin pixels， and cyclically selects features using the CS-AdaBoost algorithm program to train the best weak classifier， and finally passes the weighted linear combination of all the best weak classifiers to get the final skin classifier. Since the cost factor θ is introduced in the algorithm program， the classification result of the proposed skin classifier tends to be less costly， that is， the classification accuracy rate of the skin sample is improved. Finally， the SDD skin database was used to evaluate the performance of the proposed skin classifier. The experimental results show that the classification accuracy rate of the proposed skin classifier is 85%， which is 5% higher than the traditional skin classification method. |
| |
| Keywords: | skin classifier CS-AdaBoost brightness value features |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
