基于决策树的棉花病虫害识别研究

发布时间：2018-07-17 02:00

【摘要】：我国是一个人口众多的农业大国。棉花是我国重要的农业作物,不仅与民生息息相关,而且也是重要的战略物资,影响着国家的经济建设与进步。棉花在整个播种到收获的过程里会遭受40多种病害侵袭,如果不能快速准确地识别出棉花病害,就会影响对棉花的防治工作。因此能否快速诊断出棉花病害就显得非常重要。本文首先阐述了研究的背景和意义,论述了国内外研究现状,指出本文研究对象为棉花黄萎病、角斑病、枯萎病三种病害,并说明了棉花病害识别研究的重要性。其次,采用数字图像处理技术对棉花病害图像预处理。概述了数字图像处理的相关技术。采用中值滤波法对图像噪声信息进行消除,降低这些噪声的影响;利用加权平均的方法对图像进行灰度化处理;使用最大类间方差法分割图像,并提出改进措施,提升分割效果;经分割处理后,利用形态学方法处理图像以便后续操作。然后,基于RGB颜色模型和HIS颜色模型提取病害图像的颜色特征,提取R、G、B和H、S、I共六个分量的平均灰度值作为颜色特征参数;采用二维Gabor变换提取纹理特征,将图像与5尺度8方向共40个滤波器进行空间卷积运算。对40个滤波图像计算每幅图的平均幅值,再以每个尺度上8个方向的平均值作为纹理特征,采取统计分析的方法选择最终的输入特征。最后,介绍了决策树方法中的ID3算法和C4.5算法。对两种算法进行比较分析,最终采用C4.5决策树分类算法对棉花三种病害进行识别。借助weka数据挖掘平台进行实验,实验结果显著,准确率达到94.67%。本文采用基于C4.5的决策树方法对棉花病害分类识别是一种新的尝试。C4.5算法简单、运算快,可以处理离散型数据,易于提取规则,生成的决策树直观、易懂。
[Abstract]:China is a large agricultural country with a large population. Cotton is an important agricultural crop in China, which is not only closely related to people's livelihood, but also an important strategic material, which affects the national economic construction and progress. Cotton will be affected by more than 40 kinds of diseases in the whole process of sowing and harvesting. If cotton diseases can not be identified quickly and accurately, the prevention and control of cotton will be affected. Therefore, it is very important to diagnose cotton diseases quickly. In this paper, the background and significance of the research are first expounded, and the current research situation at home and abroad is discussed. It is pointed out that the research objects in this paper are Verticillium wilt, Corner spot and Fusarium Wilt, and the importance of cotton disease identification is also explained. Secondly, the digital image processing technology is used to preprocess the cotton disease image. The related techniques of digital image processing are summarized. The median filtering method is used to eliminate the noise information of the image to reduce the influence of the noise, the weighted average method is used to grayscale the image, the maximum inter-class variance method is used to segment the image, and the improvement measures are put forward. The segmentation effect is improved. After segmentation, the image is processed by morphological method for subsequent operation. Then, based on the RGB color model and his color model, the color features of the disease image are extracted, and the average gray values of six components are extracted as the color feature parameters, and the two-dimensional Gabor transform is used to extract the texture features, and the two dimensional Gabor transform is used to extract the texture features. A total of 40 filters are used to perform spatial convolution operation. The average amplitude of each image is calculated for 40 filtered images, and the average value of 8 directions in each scale is taken as texture feature, and the final input feature is selected by statistical analysis. Finally, ID3 algorithm and C4.5 algorithm are introduced. Finally, C4.5 decision tree classification algorithm is used to identify three cotton diseases. With the help of weka data mining platform, the experimental results are remarkable and the accuracy is 94.67. In this paper, the decision tree method based on C4.5 is a new attempt to classify and identify cotton diseases. C4.5 algorithm is simple, fast, it can deal with discrete data, it is easy to extract rules, and the decision tree generated is intuitive and easy to understand.
【学位授予单位】：华北水利水电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S435.62;TP391.41

【参考文献】