【作者】 张伟；

【导师】 王业成； 周成；

【作者基本信息】 东北农业大学 ， 工程硕士（专业学位）， 2020， 硕士

【摘要】 目前,大豆表型性状的测定基本靠人工完成,劳动强度高、测量周期长、效率低下、程序繁琐、精度较低,已经严重制约了我国大豆种业发展速度。仅靠一种人工主观科学测量和简单客观测量方法获得的多种植物主要的特征和功能性状数据十分有限,也因为缺乏一套规范化的科学表征方法,没有办法满足各类农业作物育种在基因组方面的主要功能特征研究,还有多种作物人工育种的实际应用需求问题。现在植物分析特征监测技术已经可广泛用于进行植物群体表型特征研究,能够精确化地获得一种涵盖从草本分子到植物群体的各种不同植物表型特征和形态性状。计算机信息技术的快速进步发展为有效利用管理急速生长增多的分子生物学和大数据资源提供了多种可能,而分子生物学和信息学已经成为图像处理和数据挖掘利用高通量生物数据处理信息的主要技术手段。本研究以大豆豆荚为研究对象,对其成熟期的粒用豆荚进行研究分析处理,采用数字图像处理分析技术和现代机器人学视觉处理技术的结合方法同时进行豆荚表型的研究分析,开展大豆豆荚表型数据高效快速和精准提取方法的研究,并在此基础上建立大豆豆荚数据快速测量的检测模型,开发大豆豆荚表型形态分析系统。本文主要研究如下。(1)制定统一的豆荚图像采集标准。为了能够方便处理粒用豆荚图像,搭建了特定的图像采集平台,并且采用标准的背景板,还有标准的标记正方形块。(2)对大豆豆荚的表型形态分析研究。通过对大豆豆荚采集图像的预处理,目标豆荚图像背景分割,豆荚区域完整性提取,对这批数据进行Alex Net网络进行训练,得出三种粒用豆荚形态分类参照系数,通过标准标记块的尺寸为参照,知道实际边长和面积,进行像素和实际尺寸数据的换算比例,从而得到豆荚的几何参数和颜色分量数值,为大豆选育优良品种提供基础数据参照。(3)可视化(GUI)大豆豆荚形态分析系统的设计。基本功能包括:摄像开启采集豆荚图片,豆荚图片处理选择文件,目标豆荚图像处理,豆荚数据测量,豆荚数据汇总,豆荚形态分类等。把粒用豆荚的形态分析方法运用到分析系统中去,达到初期设定的大豆豆荚目标测定分析功能,用大豆豆荚表型分析系统,完场表型数据高效快速和精准提取。更多还原

【Abstract】 At present,the determination of soybean phenotypic traits is basically done manually,with high labor intensity,long cycle,low efficiency,cumbersome procedures,and low precision,which has severely restricted the development speed of soybean seed industry in my country[1].The plant characteristics and traits obtained by manual subjective measurement and simple measurement are very limited,and there is also a lack of standardized characterization,which cannot meet the actual needs of crop genome function research and crop breeding.There are many analytical monitoring techniques available for plant phenotype research,which can accurately obtain a variety of plant characteristics and traits ranging from molecules to populations.The rapid development of computer technology has made it possible to effectively manage the rapidly increasing biological data,and bioinformatics has become the main means of processing and mining high-throughput data information^[4].This study takes soybean as the research object,researches and processes the pods at the mature stage,uses digital image processing technology and machine vision technology to study and analyze the phenotype of pods,and develops phenotypic data of soybean pods efficiently,quickly and accurately Research on extraction methods,and on this basis,establish a detection model for the rapid measurement of soybean pod data,and develop a phenotype analysis system for soybean pods.The main research of this article is as follows:（1）By formulating a unified image collection standard for pods,in order to facilitate the processing of pod image for grains,we built a specific image acquisition platform,and used a standard background board,as well as a standard marked square block.（2）In this paper,through the analysis of the phenotype morphology of the grain pods,through the preprocessing of the collected images,the segmentation of the image background,and the extraction of the integrity of the pod areas,the Alex Net network was trained on this b atch of data to obtain three types The grains are referenced by the pod morphology classification coefficient,and are quickly segmented through standard marks,knowing the actual side length and area,and converting the pixel and actual size data to obtai n the geometric parameters and color component values of the pods,and breeding excellent varieties for soybean Provide basic data reference.（3）This article designs the soybean pod morphology analysis system through MATLAB’s GUI visualization function.The basic functions include:camera start to collect pictures,file selection,image processing,data measurement,data summary,morphological classification,etc.The morphological analysis method of grain pods is applied to the system implementation,and the initial setting function is achieved.更多还原