基于BP神经网络的柑橘始叶螨图像识别

Image Recognition of Citrus Yellow Mite Based on BP Neural Network

针对柑橘始叶螨(Eotetranychus kankitus)自动识别问题,对柑橘始叶螨图像采用基于骨架的形态特征提取方法,自动提取柑橘始叶螨的骨架数学形态特征,以此作为BP神经网络的输入因子,较好地实现了柑橘始叶螨的识别。在设计的原型系统上成功识别柑橘始叶螨、柑橘全爪螨[Panonychus citri(McGregor)]和柑橘瘤瘿螨(Aceria sheldoni)3类叶螨,识别率均大于90%,每图像样本平均识别时间小于1 s,较传统的采用图像像素灰度的神经网络识别方法更快、更实用。

To solve the problem of identifying the citrus yellow mite （Eotetranychus kankitus） automatic ally,the skeleton mathematical morphological characteristics of the citrus yellow mite were automatically extracted by using extraction methods based on the morphological features of the skeleton,which were used as BP neural network input factors to achieve the identification of the citrus yellow mite.The prototype system successfully identified citrus yellow mite,citrus red mite[Panonychus citri（McGregor）] and citrus gall mite（Aceria sheldoni） with recognition-rate of more than 90％.Each image sample average recognition-time was less than one second.Compared with the traditional use of image pixel gray neural network approach,the recognition approach based on BP neural network was faster and more practical.