摘要
〔目的〕探索一种准确、快捷适合基层单位使用的蜚蠊分类方法。〔方法〕将图像数字形态学特征提取与LVQ神经网络模式识别相结合,计设和实现一种能将3种蜚蠊自动分类的系统。该系统建立、训练了2个分类器:完整虫体的分类网络net1和残缺虫体的分类网络net2。〔结果〕样品图片送入系统后,自动完成分类过程。经检验,完整虫体分类器net1的正确率为100%;残缺虫体分类器net2的正确率为97.2%。〔结论〕对3种蜚蠊的自动分类取得成功,为开发出适用于更多种类蜚蠊甚至其他昆虫的自动分类系统打下基础。
Objective To explore an efficient, rapid and automatic method which is applicable for grass-root unit in recognition of cockroach imago. Method The digital image morphologic character and LVQ neural network pattern for three kinds of cockroach imago were combined, by which an automatic system was designed with 2 dassification devices of net 1 and net 2. Net 1： the whole insect; When the photos were sent to the system they will be identified automatically. Net 2： the deformity sample. Result The exactness ratio of Net 1 （the whole insect） and Net 2 （the deformity） were 100% and 97.2% according to the test of known-samples. Conclusion The study is the stable base for the automatic classification system of cockroach forward developing and using.
出处
《中国国境卫生检疫杂志》
CAS
2008年第1期29-33,共5页
Chinese Journal of Frontier Health and Quarantine
关键词
蜚蠊
分类
自动识别
神经网络
Cockroach
Classification
Automatic recognition
Neural network
