基于水平集的作物病叶图像分割方法

针对具有复杂背景的作物病叶图像中的叶片提取问题，提出了一种基于先验信息的水平集模型。首先，在LBF模型中引入纹理信息——结构张量，构造新的水平集模型；其次，采用水平集方法表示目标叶片形状，并将先验形状信息以能量泛函的表达形式引入到上述新的水平集模型中，得到新的基于先验信息的水平集模型；最后，利用该模型对具有复杂背景的黄瓜病叶图像进行分割。结果表明，该方法能准确地提取具有复杂背景黄瓜病叶图像中的病叶，为后续的病斑提取、识别和诊断奠定前期基础。

Segmentation method for crop disease leaf images with complex background

A new segmentation method of the level set model based on prior information was proposed in this paper and was applied to crop diseased leaves with complex background. Firstly, structure tensor information was used to improve the LBF model, so that a new level set model was constructed with structure tensor information. At he same time, target shape was represent by the level set method. Secondly, prior shape information in the form of energy function was introduced to the new level set model and got the new level set model based on prior information. Finally, cucumber disease leaf images with complex background were segmented by the model. Experimental results show that the method can accurately extract the disease leaf from cucumber disease leaf images with complex background, which can provide the foundation for extracting, identifying and diagnosing the diseased spots.