基于新型特征提取的寄生虫卵图像识别研究  

Study on recognition for parasite ovum images based on new method of feature extraction

作　　者：罗泽举[1] 宋丽红[2] 伍小明[3] 詹希美[4] 

机构地区：[1]重庆工商大学计算机科学与信息工程学院,重庆400067 [2]重庆工商大学实验实习中心,重庆400067 [3]中山大学数学与计算科学学院,广东广州510275 [4]中山大学中山医学院,广东广州510080

出　　处：《计算机应用》2007年第6期1485-1487,共3页journal of Computer Applications

摘　　要：讨论了用支持向量机进行多分类的若干学习策略,提出了一种新型图像特征提取方法,以此来实现对鞭虫等九种寄生虫卵图像自动识别和分类,平均识别率优于传统神经网络,达到了93.9%,为寄生虫卵图像识别提供了一种新方法。Several learning strategies about multi-class classification by using Support Vector Machines （SVM） being discussed in this paper, we proposed a new kind of method for images feature extraction, so as to realize recognition and classification automatically for the nine kinds of parasites ovum images such as Trichuris trichiura, etc. The average recognition rate is superior to that of traditional neural network and reaches 93.9%, providing a new method for parasites ovum＇s images recognition.

关 键 词：支持向量机 新型特征提取 寄生虫卵识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]