小麦条锈菌孢子的在线图像获取与计数方法研究

发布时间：2018-04-24 12:41

本文选题：小麦条锈病 + 图像分割　；参考：《西北农林科技大学》2017年硕士论文

【摘要】：小麦条锈病一直是威胁我国小麦优质、高产的重要病害,由于对其早期准确预测预报技术的缺乏,使得小麦条锈病发病普遍而且严重,而夏孢子菌源数是影响小麦条锈病发病和传播的直接因素。为实现对田间空气中小麦条锈菌夏孢子数量的远程实时快速测定,本研究对实时在线图像获取技术和显微图像处理技术展开研究,主要研究内容及结论如下:(1)对小麦条锈菌孢子显微图像的预处理、分割和形态学处理方法进行了研究。利用实验室条件下所获得的小麦条锈菌孢子显微图像为对象进行了试验研究。首先对图像缩放、颜色空间转换、图像灰度化等进行了研究。然后利用四种不同的分割方法:K-means聚类法、Otsu阈值分割法、Canny边缘检测法、分水岭分割法进行了分割对比试验。对比了RGB、L*a*b*、HSV等三种颜色空间下利用不同分量进行K-means聚类的效果。通过对试验结果的分析比较选择出了最优的分割方法,本研究中最优分割方法为基于Canny边缘检测的分割方法。对分割后的图像进行了形态学处理。该研究利用基于Canny边缘检测的分割方法实现了孢子目标区域的分割,通过形态学处理得到了边缘平滑的孢子区域二值图像。(2)对小麦条锈菌孢子的识别与计数方法进行了研究。利用单个孢子形状简单、形状因子较大,粘连孢子形状较复杂、形状因子较小的特性,通过形状因子对单个与粘连孢子进行了识别。对识别后的孢子利用两种计数方法:平均面积法和角点检测法进行了计数。对Harris和SUSAN两种角点检测算法进行了对比试验,对经典Harris角点检测算法进行了改进,使其更适合于检测粘连区域的边缘交点,通过对比发现Harris角点检测法检测结果更好。利用127幅400倍放大的显微图像和342幅200倍放大的显微图像进行了孢子计数实验,实验结果表明平均面积法和角点检测法的平均计数准确率均不小于84%,能够对孢子进行较准确的计数。(3)提出了一种小麦条锈菌孢子显微图像在线获取方法。该方法利用真空泵、51单片机、数码显微镜、4G无线路由器、Intel微型机、太阳能供电系统等进行小麦条锈菌孢子的捕获、显微图像获取和图像无线传输,从而实现了小麦条锈菌孢子显微图像的在线获取。解决了田间小麦条锈菌孢子监测的实时性问题,为进一步实现小麦条锈病的预测预报提供了强有力的基础支撑。(4)设计了一种小麦条锈菌孢子计数软件。利用Matlab的GUI编译工具箱完成了小麦条锈菌孢子计数软件的设计。对其功能实现与使用进行了详细说明。该软件可利用4种不同的分割方法、2种不同的计数方法进行孢子的计数,使用直观明了,操作简单,可在Matlab环境下运行。
[Abstract]:Wheat stripe rust is an important disease that threatens the quality and high yield of wheat in China. Due to the lack of early and accurate prediction technology, stripe rust of wheat is common and serious. The number of summer spores is a direct factor to affect the incidence and transmission of wheat stripe rust. In order to measure the number of summer spores of wheat stripe rust in field air, the real-time on-line image acquisition and microscopic image processing were studied in this study. The main contents and conclusions are as follows: (1) Pretreatment, segmentation and morphological processing of the microscopic image of wheat stripe rust spore were studied. The micrographs of wheat stripe rust spores obtained under laboratory conditions were studied experimentally. Firstly, image scaling, color space conversion and image grayscale are studied. Then, four different segmentation methods, namely: K-means clustering method, Otsu threshold segmentation method, Canny edge detection method and watershed segmentation method, are used to carry out a comparative segmentation experiment. The results of K-means clustering with different components in three color spaces were compared. Through the analysis and comparison of the experimental results, the optimal segmentation method is selected. In this study, the optimal segmentation method is based on Canny edge detection. The segmented image was processed by morphology. In this study, the segmentation method based on Canny edge detection was used to segment the target area of spores. The binary image of spore region with smooth edge was obtained by morphological processing. The method of identifying and counting spores of wheat stripe rust was studied. Based on the simple shape, large shape factor, complex shape and small shape factor of a single spore, a single conidial spore was identified by shape factor. The spores identified were counted by two counting methods: average area method and corner detection method. In this paper, two corner detection algorithms, Harris and SUSAN, are compared, and the classical Harris corner detection algorithm is improved to make it more suitable for detecting the edge intersection of adhesion region. It is found that the Harris corner detection method is better than the traditional Harris corner detection method. Spore counting experiments were carried out using 127 magnified microscopic images and 342 200 magnified microscopic images. The experimental results show that the average counting accuracy of the average area method and the corner detection method are not less than 84, and the spores can be counted accurately. (3) an online method for obtaining the microscopic image of wheat stripe rust spores is proposed. This method uses vacuum pump 51 single chip computer, digital microscope 4G wireless router and Intel microcomputer, solar power supply system to capture wheat stripe rust spores, obtain microscopic images and transmit images wirelessly. Thus, the micrograph of wheat stripe rust spores can be obtained on line. The real-time monitoring problem of wheat stripe rust spores in the field was solved, and a software was designed to calculate the spores of wheat stripe rust. The software of spores counting of wheat stripe rust was designed by using GUI compiler toolbox of Matlab. The implementation and use of its functions are described in detail. The software can use four different segmentation methods and two different counting methods to count spores. The software is intuitive and easy to operate. It can be run in Matlab environment.
【学位授予单位】：西北农林科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S435.121.42;TP391.41

【参考文献】