家蚕微粒子病的图像识别技术研究  

Research on Image Recognition Technique for Pebrine in Silkworm

作　　者：张香琴[1] 方如明[1] 汪萍[2] 蔡健荣[1] 许俐[3] 

机构地区：[1]江苏理工大学生物与环境工程学院 [2]中国农科院蚕业研究所 [3]南京航空航天大学民航学院

出　　处：《农业机械学报》2001年第5期65-68,共4页Transactions of the Chinese Society for Agricultural Machinery

摘　　要：用计算机图像识别技术来检测微粒子病 ,以替代人工镜检。对拍摄的原始图像用直方图均衡化法进行对比度增强的处理 ;用快速二维 Ostu阈值化法进行二值化 ,实现了微粒子与背景的分离 ;用形态筛选法去掉了大量噪声及小杂质 ,实现了微粒子与杂质的初步分离 ;提取了周长、面积、圆度、凹度、内角极值、形状规则度 6个特征 ,对微粒子进行分层识别。对 4 3幅含微粒子的图像进行识别试验 ,识别完全正确率为 72 .0 9% ,识别有效率为86.0 5% ,漏判率为 1 3 .95%。Pebrine is a kind of ancient widely distributed and strong destructive silkworm disease. At present the detection method for the disease is basically microscopic method by manual work. In this paper image recognition technique was employed to detect pebrine disease instead of the microscopic method. In this research the process of image recognition of pebrine was also the process of the separation of pebrine with background and impurities in the image. The original image was boosted up in contrast by histogram equalization. A twovalued image was gained by the fast two-dimensional Ostu method, in which pebrine was separated with background. A mass of noises and small impurities were filtered by morphology filtering method and the preparatory separation of pebrine with impurities was realized. The six characters, perimeter, area, roundness, concavity, extremum of internal angles and shape regulation, were extracted to remove the remainder impurities and finally the purpose of recognizing pebrine was obtained. The total 43 pebrine images were tested in the experiment. The complete recognition exactness was 72.09%, the recognition validity 86.05% and the leakage of justice 13.95%.

关 键 词：家蚕 微粒子病 图像识别 特征提取 

分 类 号：S884.21[农业科学—特种经济动物饲养] TP391.41[农业科学—畜牧兽医]