基于计算机视觉的大田害虫远程自动识别系统

Remote automatic identification system of field pests based on computer vision

为了实现大田害虫的快速实时识别和诊断,设计了一套大田害虫远程自动识别系统。该系统通过3G无线网络将害虫照片传输到主控平台中,在主控平台中实现远程自动识别。系统首先对害虫图像进行基于形态和颜色特征值的提取。害虫图像的形态特征由周长、面积、偏心率等以及7个胡不变矩共16个特征值组成,颜色特征值由9个颜色矩组成,然后建立支持向量机分类器。采用该系统对6种常见大田害虫进行了测试,平均准确率达到87.4%。考虑到不同的害虫姿态和大田中不同的光照条件,系统的分类效果是满意的。

In order to achieve fast real-time identification and diagnosis of field pests, a remote automatic pest identification system was designed in this paper. This system is composed of remote classification platform （ROCP） including personal computer, CMOS camera and 3G wireless communication module and a host control platform （HCP）. The ROCP sends the image data, which is encoded using JPEG 2000, to the HCP through the 3G network. The image transmission and communication are accomplished using 3G technology. The system transmits the data via a commercial base station. The system can work properly based on the effective coverage of base stations, no matter the distance from the ROCP to the HCP. The image data was decoded firstly, then the pest was segmented from background, and the morphology features and color features were extracted at last for classification. Sixteen morphology features consisted of perimeter, area, eccentricity and seven Hu invariant moments etc. Nine color features were described by color moments. The support vector machine classifier was used at last for identification. Six species of common field pests including Cnaphalocrocis medinalis Guenee, Chilo suppressalis, Sesamia inferens, Naranga aenesc, Anomala corpulenta Motschulsky, Prodenia litura were tested in the system and the average accuracy is 87.4%. Considering the different pests’pose and different field lighting conditions, the result is satisfactory. The study of the automatic pest identification system which combined with machine vision, image processing, pattern recognition technology and 3G wireless communication technology, was not reported in China. The designed system can automatically identify the field pests and can provide timely and accurate information for pest prevention. The application of the designed system can reduce prevention cost and improve the control effect. The study can provide a reference for agricultural pest prevention.