中文摘要：

通过对黄瓜病害图像的准确分析，有效提取了图像的底层特征，建立了8种常见黄瓜病害的高斯混合模型（Gaussian Mixture Model，GMM），并利用最大期望算法（Expectation-Maximization，EM）估计GMM的参数，精确描述了8种黄瓜病害的特征分布，从而提高了对黄瓜病害的正确识别和为害情况的准确把握，为实现黄瓜病害的实时与准确的预测和防治提供了理论依据。

英文摘要：

Based on the accurate analysis of cucumber disease images, the low level feature of images was effectively extracted, and Gaussian Mixture Model （GMM） for 8 common cucumber diseases was built. The parameters of GMM were estimated by the algorithm of expectation maximum （EM） to accurately charac- terize the feature distribution of 8 cucumber diseases, thus increased the correct identification of cucumber diseases and accurate grasping of damage conditions, and provided basis for achievement of real-time and accurate prediction of cucumber diseases.