基于机器视觉的玉米病虫害监测方法研究

发布时间：2018-01-16 16:00

本文关键词：基于机器视觉的玉米病虫害监测方法研究　出处：《安徽工程大学》2017年硕士论文　论文类型：学位论文

【摘要】：中国是个农业大国,农业收入是国民经济的主要来源之一。农作物在生长过程中时时刻刻会受到病虫害的影响,从一定程度上制约着农作物的产量。因此,及时、准确地识别农产品病虫害会有效地减少其损失。玉米作为我国最重要的粮食之一,近些年受到病虫害的影响导致产量有所下降,如何利用现代机器视觉技术作为依据对病症种类进行准确地识别,达到对病情"对症下药"具有重要意义。本文主要围绕利用计算机视觉技术对玉米叶常见四种类型的病害进行识别,并进行了相关实验研究。具体研究内容及成果包括以下几个方面:首先,通过相机采集玉米叶灰斑病、大斑病、小斑病、锈病四种常见病害作为研究对象,并利用中值滤波和聚类方法对玉米叶片图像预处理,排除无关变量(阴影,光照等)因素的影响,并在此基础之上,通过彩色空间的聚类分割方法对病斑区进行分割,能较好的保留病斑区边缘以及颜色特性。其次,利用颜色、纹理、形状特征参数的方法对玉米叶片图像进行特征提取;颜色特征得到一阶矩、二阶矩、三阶矩特征参数,纹理特征得到对比度、相关性、熵、均匀性、能量相关特征参数,形状特征得到面积、周长、矩形度、伸长度、似球度相关参数,因此,通过提取玉米四种病害间的差异性可对病害准确分类。最后,通过利用支持向量机分别对玉米叶片病害进行识别;在支持向量机原理的基础上,对数据统计分类并建立数据库,并得到最终的识别结果。实验结果表明利用单向的特征识别存在一定的误差,而结合多特征的方法能够达到较为理想的实验结果;并在此基础之上,利用vs2013设计了玉米四种常见病害的可视化界面,具有良好的交户性,为玉米叶的病害图像识别提供了一定的借鉴。综上,本文利用图像处理技术对玉米常见的四种病害特征识别,通过聚类分割方法对病斑区分割,结合图像纹理特征向量的参数进行统计与分类,以对玉米常见病斑进行精准识别,具有一定的实际应用和推广价值。
[Abstract]:China is a large agricultural country, agricultural income is one of the main sources of the national economy. Crops are always affected by pests and diseases in the process of growth, which to some extent restricts the output of crops. Timely and accurate identification of pests and diseases of agricultural products will effectively reduce their losses. Corn, as one of the most important grain in China, has been affected by pests and diseases in recent years, resulting in a decline in yield. How to use modern machine vision technology as the basis for accurate identification of disease types. It is of great significance to achieve the "right remedy" for the disease. This paper mainly focuses on the identification of four common diseases of maize leaf by computer vision technology. The specific research contents and results include the following aspects: firstly, four common diseases of maize leaf gray spot, large spot, small spot and rust were collected by camera as the research object. The median filter and clustering method were used to preprocess the maize leaf image, and the independent variables (shadow, light, etc.) were excluded, and on this basis. Through the color space clustering segmentation method to segment the disease spot area, can better preserve the disease spot edge and the color characteristic. Secondly, uses the color, the texture. The method of shape feature parameter is used to extract the feature of maize leaf image. Color features obtained first moment, second moment, third moment feature parameters, texture features obtained contrast, correlation, entropy, uniformity, energy correlation feature parameters, shape features to obtain area, perimeter, rectangle. The relative parameters of elongation and sphericity are related, so the disease can be classified accurately by extracting the difference among the four maize diseases. Finally, the maize leaf disease is identified by using support vector machine. Based on the principle of support vector machine (SVM), the statistical classification of data and the establishment of database are carried out, and the final recognition results are obtained. The experimental results show that there are some errors in using one-way feature recognition. The experimental results can be achieved by combining the multi-feature method. On this basis, the visual interface of four common diseases of maize was designed by using vs2013, which has good housekeeping, and provides a certain reference for the recognition of maize leaf disease image. In this paper, we use image processing technology to identify the four common disease features of maize, segment disease area by clustering segmentation method, combined with the parameters of image texture feature vector for statistical and classification. It has certain practical application and popularizing value to accurately identify common spot of maize.
【学位授予单位】：安徽工程大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S435.13;TP391.41

【相似文献】