Abstract: Wheat is one of the major grain crops in China, cultivated in large-scale, distributed in vast areas with long growing cycles and multiple varieties. However, it is easily threatened by diseases and pests. Therefore, rapid monitoring and accurate identification of diseases and pests become an important research project. Based on the wheat monitoring system platform previously developed with Internet of Things (IoT), this study designed a remote diagnosis system combining image acquisition with diagnosis methods. The diagnosis methods for 3 common wheat diseases and pests were studied preliminarily, and 4 pictures of wheat leaves contaminated with powdery mildew, rust, aphis and healthy ones were compared and studied by means of image segmentation, feature extraction and digital image classification. The results showed that the recognition rates had reached desired levels. Among them, the recognition rate for powdery mildew was 82.5%, the recognition rates for rust, aphis and healthy leaves were all above 95%. This study combined the image recognition technology with IoT technology. These technology was convenient for tele-transmission of diseases and pests images and multi-peer retrival. These merits have greatly improve our ability in remote identification and diagnosis. This technology has broad development prospect.

Key words: wheat diseases and pests, internet of things, image recognition, machine vision, remote diagnosis

摘要： 小麦是中国主要粮食作物，栽培品种多、种植面积大、分布区域广、生长周期长，容易遭受病虫害威胁，快速监测和准确识别病虫害成为一项重要的课题。基于前期构建的小麦物联网监控系统平台，研发了集成图像获取、图像识别诊断于一体的应用系统。初步研究了小麦比较常见的三种病虫害的识别与诊断方法，并利用图像分割、特征提取及数字图像分类识别技术，将物联网系统获取的感白粉病、锈病、蚜虫的不健康叶片与健康小麦叶片的图片分别进行对比实验研究。实验结果显示，识别率都较为理想，其中白粉病的识别率为82.5%，锈病、蚜虫和健康叶片的识别率都在95%以上。将病虫害图像识别技术与物联网技术结合，方便病虫害图像的远程传输、多点获取等优点，大幅度提升对病虫害远程识别和诊断能力，具有广阔的发展前景。

关键词: 小麦病虫害, 物联网, 图像识别, 机器视觉, 远程诊断