利用光谱成像技术(400～720 nm)识别黄瓜白粉病、 角斑病、 霜霉病、 褐斑病和无病区域。 构建高光谱图像采集系统进行样本图像的采集, 预处理和光谱信息的提取。 由于获得的原始光谱数据量很大, 为了减少后续运算量, 提高准确率, 采用逐步判别分析和典型判别分析两种方法进行降维。 逐步判别从55个波段中选择12个波段, 典型判别从55个波段中提取2个典型变量。 利用选择的光谱特征参数建立病害识别模型。 逐步判别构建的模型对训练样本和测试样本的判别准确率分别为100%和94%, 典型判别构建的模型对训练样本和测试样本的判别准确率均为100%。 说明利用高光谱成像技术可以进行黄瓜病害的快速、 准确识别, 并为实现可见光谱范围内黄瓜病害的田间实时在线检测提供了可能。

Hyperspectral imaging(400-720 nm) and discriminate analysis were investigated for the detection of normal and diseased cucumber leaf samples with powdery mildew(Sphaerotheca fuliginea), angular leaf spot(Pseudomopnas syringae), downy mildew(Pseudoperonospora cubensis), and brown spot(Corynespora cassiicola). A hyperspectral imaging system was established to acquire and pre-process leaf images, as well as to extract leaf spectral properties. Owing to the complexity of the original spectral data, stepwise discriminate and canonical discriminate were executed to reduce the numerous spectral information, in order to decrease the amount of calculation and improve the accuracy. By the stepwise discriminate we selected 12 optimal wavelengths from the original 55 wavelengths, and after the canonical discriminate, the 55 wavelengths were reduced to 2 canonical variables. Then the discriminate models were developed to classify the leaf samples. The result shows that the stepwise discriminate model achieved classification accuracies of 100% and 94% for the training and testing sets, respectively. For the canonical model, the classification accuracies for the training and testing sets were both 100%. These results indicated that it is feasible to identify and classify cucumber diseases using hyperspectral imaging technology and discriminate analysis. The preliminary study, which was done in a closed room with restrictions to avoid interference of the field environment, showed that there is a potential to establish an online field application in cucumber disease detection based on visible spectroscopy.