及早检测和识别植物病害是防治病害的前提。植物病害叶片症状是发现病害和识别病害类别的重要依据之一，而利用植物病害的环境信息有助于提高植物病害识别算法的准确率。本项目旨在系统研究基于病害叶片与环境信息的植物病害识别中的关键问题。首先，提出一种基于动态标记分水岭法的复杂背景下的病害叶片图像分割算法；然后，提出一种局部与全局联合保持投影的稀疏流形学习方法，并构建基于最小覆盖原理和谱回归策略的高效求解方案，以实现大规模复杂病害叶片病斑图像实时维数约简；第三，定义一种植物病害环境影响子，用来定量地表征环境信息对植物病害发生的影响。在此基础上，提出一种基于稀疏流形学习的叶片症状和环境影响子相结合的植物病害识别方法；最后，建设包含多种背景下的常见植物病害叶片图像数据库，验证所提出算法的有效性。本项目的开展不仅将对植物病害识别的方法研究起到一定的推动作用，而且还具有很大的社会经济价值。

Early detecting and recognizing the plant diseases is the precondition of the plant disease control. The symptom of plant disease leaf is one important basis for plant disease discover and type identification, and the environmental information of the plant disease is useful for accurately recognizing the plant diseases. The proposed project will systematically study the key problems of plant disease recognition based on the plant disease leaves and environment information. Firstly, a segmentation algorithm of the disease leaf image with complex background is proposed based on the dynamic marking watershed method. Secondly, a novel sparse manifold learning method is proposed with both the local and global preserving projection, and an effective solution is devised by applying the minimum set cover theory and spectral regression techniques to perform the real-time dimensionality reduction of the large-scale disease leaf spot image data. Thirdly, an environmental impact factor is defined to quantitatively characterize the environmental impact on the plant diseases. And then, a plant disease recognition method based on sparse manifold learning is proposed by considering the leaf symptoms and the environmental impact. Finally, a database about public disease leaf images with various complex backgrounds is established to verify the effectiveness of the proposed algorithms. The proposed project will not only make a progress in studying the plant disease recognition methods, but also be of great socio-economic value.