基于Chain Boosting算法的棉纤维检测  

Cotton fibre detection based on chain boosting algorithm

作　　者：熊超[1] 施国生[1] 

机构地区：[1]浙江理工大学,浙江杭州310012

出　　处：《计算机工程与设计》2009年第15期3600-3601,3693,共3页Computer Engineering and Design

摘　　要：为了更好的对混纺纤维中的棉纤维进行含量分析和质量评估,提出了一种棉纤维检测的方法,此方法使用Chain Boosting算法来提高检测的正确率。预处理棉纤维的横截面图,把体现棉纤维的特征的指标从纤维横截面图像中提取出来(包括轴度,复杂度等);然后利用训练好的BP网络作为弱学习算法,使用每一个Boosting分类器学习的历史信息,基于线性回归特征消除(RFE)策略,消除冗余,据此判别一幅纤维截面图像是否为棉纤维截面图像。实验结果表明,该方法实用有效,对检测的正确率有较大幅度的提高。In order to get more accurate data on analysis of content and assessment of quality about cotton fibre in mixed fibre, an algorithm based on Chain Boosting for cotton fibre＇ s detection is proposed. Firstly, after the image preprocessing, some features such as axis degree, circularity and central moment are extracted, then BP net is used as weak study machine, and integrate historical information into successive boosting learning. Based on linear recursive feature elimination （RFE） strategy, the redundancy of AdaBoost is removed. Experimental results shows that the method is effective and improves detective correction greatly.

关 键 词：BOOSTING算法 棉纤维检测 BP网络 图像特征 图像处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]