基于图像识别的小麦腥黑穗病害诊断技术研究

传统的检疫小麦腥黑穗病害的方法效率较低影响检测的稳定性和客观性.提出一种基于图像识别的小麦腥黑穗病分类诊断技术.以显微镜下采集的小麦病害图像为研究对象,对其进行滤波增强及病害区域分割,再提取单个病害区域图像的颜色、形状和纹理等特征参数;最后利用归一化后的特征值,通过BP神经网络分类器实现了小麦腥黑穗病害的诊断.将计算机图像识别结果和实际小麦腥黑穗病类型进行对比,表明了该诊断技术的可行性和有效性

Study on diagnosis of Tilletia based on image recognition

The traditional quarantine treatment of Tilletia has low efficiency,which is difficult to ensure the stability and objectivity of results.Therefore,a kind of diagnosis technique was proposed to classify Tilletia based on image recognition.Regarding wheat disease images collected from microscope as research subjects,they were processed with filtering enhancement and region-segmentation of diseases and collected characteristic parameters of a single disease’s image,such as color,shape and vein.Then the diagnosis of this disease was completed after the operation/classification of normalized eigenvalues by BP neural network classifiers.It was demonstrated feasibly and effectively by comparing image recognition results of computers with the disease’s features.