摘要
针对人工筛选蚕茧工作中漏检率高、效率低的问题,提出了一种基于颜色特征和支持向量机的蚕茧分类方法。该方法以黄斑茧、烂茧、印头茧、薄皮茧和上车茧等5种蚕茧为研究对象,首先对采集的蚕茧图像进行边缘检测、连通域提取等步骤以实现背景分割与正、反面连接,然后提取蚕茧图像HSV颜色空间模型下的颜色直方图和颜色矩作为颜色特征,并将5种蚕茧的HSV三通道像素分布详细划分、量化,最后对蚕茧图像颜色特征进行主成分分析,结合支持向量机设计4种分类器构造方案进行效果对比。试验结果表明,将颜色直方图和颜色矩组成颜色特征集后进行主成分分析,并使用支持向量机构造分类器对5种蚕茧分类效果最好,其中主成分变换最优累计贡献率为999%,分类准确率达91%,单粒蚕茧图像识别时间为012s。本方法为蚕茧智能化分选设备的研究提供了理论基础。
A cocoon classification method based on color feature and support vector machine was proposed to solve the problem of high false detection rate and low efficiency in the manual screening of cocoons.Taking 5 types of cocoons as research objects,including yellow spotted cocoon,decayed cocoon,end-stained cocoon,thin shelled cocoon and reelable cocoon,edge detection and connection area abstraction of collected cocoon images were performed to segment background and to stitch the front and back split images.Then,color histogram and color moment under HSV color space model of cocoon image were extracted as color features,and HSV three-channel pixel distributions of 5 types of cocoons were divided and quantified in detail.Finally,principal component analysis on the color features of cocoon images was carried out,and 4 different classifier construction schemes based on support vector machine were designed and compared.The experimental results showed that based on principal component alaysis on color features of color histogram and color moment,the classifier using support vector machine was the best to classify 5 cocoon types.The optimal accumulative contribution rate of principal component transformation was 999%,the classification accuracy was91%,and the image recognition time of single cocoon was 012 s.This study laid theoretical foundation for further research on equipment for intelligent selection of cocoons.
作者
孙卫红
黄志鹏
梁曼
邵铁锋
Sun Weihong;Huang Zhipeng;Liang Man;Shao Tiefeng(College of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou 310018,China)
出处
《蚕业科学》
CAS
CSCD
北大核心
2020年第1期86-95,共10页
ACTA SERICOLOGICA SINICA
基金
中国纤维质量监测中心项目(OITC-G190281374)
浙江省公益技术应用研究项目(LGG20E050014)
江西省科技计划项目(20192BBFL60037)
国家市场监管总局科技计划项目(S2019MK430).
关键词
蚕茧
颜色特征
支持向量机
图像分类
Cocoon
Color feature
Support vector ma-chine
Image classification
作者简介
第一作者及通讯作者:孙卫红(1969—),男,博士,教授。E mail:whsun@126.com
