白粉病是黄瓜常见病害之一, 传播速度极快, 严重时可造成黄瓜大量减产, 对其进行快速准确识别, 对黄瓜白粉病诊断和防治具有重要意义, 应用可见光谱技术, 结合主成分分析和支持向量机算法, 实现对黄瓜白粉病的快速识别。 配制白粉病菌孢子悬浮液, 并人工接种于科研温室内的黄瓜叶片上, 以诱发黄瓜白粉病, 待白粉病有一定面积暴发后, 利用海洋光学USB2000+型便携式光谱仪对黄瓜叶片光谱信息进行采集, 利用五点取样法采集样本, 在5个检查点, 每点选取2株黄瓜进行调查, 每株选取4枚感病叶片, 每枚叶片随机选取5个感病区域进行光谱采集, 共计采集200个感病叶片光谱样本, 同样采集200个健康叶片样本作为对照。 通过Ocean Optics Spectra-Suite软件采集漫反射标准白板信息和光谱仪暗电流实现光谱仪校正, 调节积分时间、 扫描次数以及平滑度等参数来实现光谱曲线平滑处理, 以有效抑制光谱噪声, 对光谱特征进行分类识别, 去掉首尾噪声较大的波段, 保留光谱的可见光波段进行研究, 最终选取450~780 nm波段范围作为研究对象。 利用主成分分析对所研究波段范围内的高维光谱数据(947维)进行降维处理, 根据主成分的累计贡献率, 选取前5个主成分作为分类模型的输入, 以白粉病和健康叶片的判别结果作为输出, 利用支持向量机算法, 通过对样本的分类学习训练构建黄瓜白粉病和健康叶片的分类识别模型, 随机选取120个样本作为训练集用于分类模型构建, 其余80个样本作为测试集用于模型检验, 并通过选取不同的核函数来获得最优模型。 利用混淆矩阵对分类识别模型的准确率进行评价, 当选取径向基核函数时, 分类识别模型对黄瓜健康叶片和白粉病叶片的识别准确率最高, 分别为100%和96.25%, 总准确率为98.125%, 具有较高的准确率。 结果表明, 利用可见光光谱信息并结合主成分分析和支持向量机算法, 可以实现对黄瓜白粉病的快速准确识别, 为黄瓜病害诊断提供了方法和参考依据。

Powdery mildew is one of the common diseases of cucumber, which has a rapid propagation speed and can cause a large reduction of cucumber. Quick and accurate recognition of cucumber powdery mildew has great significance for the diagnosis and control of cucumber diseases. Utilize visible spectrum technology combined with principal component analysis and support vector machine algorithm can realize the quick recognition of cucumber powdery mildew. Sphaerotheca fuliginea was used to make spore suspension and inoculated it into the cucumber leaves in a Scientific research solar greenhouse to induce powdery mildew. When the powdery mildew occurred, the spectral information of cucumber leaves was collected by the Ocean Optics USB2000+ portable spectrometer. Five point sampling method was used to collect samples, two cucumber plants were inspected at each point and four leaves were checked on each plant, and five areas were chosen randomly on each leaf to use to spectral information acquisition. Then 200 samples of cucumber powdery mildew leaves were got , and 200 healthy leaf samples were collected as contrast by the same method. the standard white plate and dark current was Utilized to calibrate the spectrometer. The integral time and the scanning times were adjusted and the smoothness parameters of Ocean Optics Spectra-Suite software was used to smooth spectral curves and suppress noise. Through classification and recognition of spectral features, the spectral bands with big noise was removed and the 450~780 nm visible light band was chosen as the research range. The principal component analysis (PCA) was used to reduce the dimension of high-dimensional spectral data (947 dimension). According to the cumulative contribution rate of principal components, the former 5 principal components were chosen as input variables and the discriminant results as the output to build the classification model. We utilized support vector machine (SVM) algorithm and randomly took 120 samples as the training set to build the classification model, and the rest 80 samples as testing set for model checking, and the optimal model was obtained by selecting different kernel functions. The confusion matrix was used to evaluate the accuracy of the classification model, when the radial basis kernel function was selected, the recognition accuracy of the classification model for cucumber healthy leaves and powdery mildew leaves were respectively 100% and 96.25%, and the total accuracy was 98.125%. The results showed that the visible light spectrum analysis combined with PCA and SVM algorithm could be used to identify cucumber powdery mildew quickly and accurately, which provides a method and reference for the diagnosis of cucumber diseases.