基于多重分形特征的作物叶片图像分割技术的研究

发布时间：2018-11-10 10:39

【摘要】：叶片作为作物的重要器官和营养摄取的主要组织,其形态和表面的纹理特征能够表征作物的生长情况和病虫害情况。同时叶片在作物生长过程中形态变化比根、茎的变化小,因此其纹理特征是探究作物生长状况的理想对象,但由于现阶段刻画纹理特征的有效工具很少,是一个待解决的难题。分形中的多重分形理论是一种描述图像纹理特征的重要手段在并图像处理领域得到了很好的应用。本文利用多重分形理论针对叶片图像奇异性的特点,提出了多种描述叶片灰度图像多重分形特征方法,并利用这些特征对作物图像分割问题展开了研究,旨在通过机器智能为作物叶片缺素和病害无损诊断系统提供理论基础。1.实现了基于局部容量测度多重分形的油菜病虫害叶片纹理特征的提取方法,并进一步实现图像分割技术。实验表明该分割方法对叶片的边缘、叶脉以及病害区域非常敏感,能够较好地分割出病害区域,对于叶片的各种不同病害图像用基于不同容量测度的多重分形谱分割的效果存在差异,但总的来说,基于容量测度的多重分形谱分割方法能够较准确的检测出病害关键区域。2.在去波动平均分析方法的基础上提出了一种新的二维多重分形去趋势波动平均分析方法,并对缺素叶片图像作分割实验,将新的方法应用于对油菜缺钾和缺镁叶片分割中来验证该方法的有效性。
[Abstract]:Leaf is an important organ and the main tissue of nutrient uptake. Its morphology and surface texture can characterize the growth of crops and the situation of pests and diseases. At the same time, the morphological change of leaf is less than that of root and stem, so its texture feature is the ideal object to explore the crop growth condition, but there are few effective tools to describe the texture feature at present, which is a difficult problem to be solved. Multifractal theory in fractal is an important method to describe texture features of images. It has been well applied in the field of parallel image processing. In this paper, according to the singularity of leaf image, multifractal theory is used to describe the multifractal feature of leaf grayscale image, and the problem of crop image segmentation is studied by using these features. The purpose of this paper is to provide the theoretical basis for the nondestructive diagnosis system of crop leaf deficiency and disease through machine intelligence. 1. The method of extracting the leaf texture feature of rape based on the multifractal of local capacity measure is realized, and the image segmentation technology is further realized. The experimental results show that the method is very sensitive to the edge, vein and disease area of the leaf, and it can segment the diseased area well. For different disease images of leaves, the effect of multifractal spectrum segmentation based on different capacity measure is different, but generally speaking, The multifractal spectrum segmentation method based on capacity measure can accurately detect the key areas of disease. 2. On the basis of undulating average analysis method, a new two-dimensional multifractal detrend wave average analysis method is proposed, and the segmentation experiment of leaf image with lack of prime is made. The new method was applied to the potassium deficiency and magnesium deficiency leaf segmentation of rape to verify the effectiveness of the method.
【学位授予单位】：湖南农业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S126;TP391.41

【参考文献】