摘要
为快速准确识别苹果叶部病害,研究基于Android移动平台的苹果叶病害远程识别系统,提出病害图像采集、存储和发送的Android客户端及用于图像接收、分析处理和返回识别结果的PC服务器端系统架构。客户端调用系统相机获取图像,以socket流实现客户端与服务器通信,服务器用最大类间方差法抽取图像病斑部位,提取颜色、纹理、形状参数,选择支持向量机实现病害识别。实验结果表明,相比BP网络,支持向量机的识别性能更佳,该系统对苹果叶部斑点落叶病、锈病以及花叶病的识别率达98.33%,识别时间小于16s,可为果农提供方便快捷的苹果病害诊断及防治技术服务。
To identify the apple leaf disease more rapidly and accurately, remote recognition system of apple leaf disease based on Android platform was presented. The system architecture including Android mobile client for disease image collecting, storage and sending, and the PC server for image receiving, analyzing and returning the recognition results were proposed. The client called system camera to capture photo. The server segmented the lesion site by using OTSU method. With color, shape, texture features, the support vector machine was used to identify disease. The system implemented communication between client and server through socket stream. Experimental results show that, compared to BP, SVM has better performance, and the accuracy rate of disease recognition for apple rust, altermaria leaf spot and mosaic can reach 98.33%, the recognition time is less than 16 s. It provides convenient and fast identification service and prevention guidance of the apple leaf disease for farmers.
出处
《计算机工程与设计》
北大核心
2015年第9期2585-2590,共6页
Computer Engineering and Design
基金
国家科技支撑计划基金项目(2012BAH29B00)
