
基于DSP的除草机器人杂草实时识别与试验研究.pdf
73页基于DSP的除草机器人杂草实时识别与试验研究Research on Weed Real-time Recognition andExperiments of Weeding Machine Based on DSP指导教师姓陈树人20 1 1年6月江苏大学硕士学位论文摘 要精确农业是现代农业发展的必然趋势之一,作为精确农业关键技术的除草剂的变量施药能够精确地对准杂草喷洒除草剂,从而有效地减少除草剂用量本文以棉田中棉花与杂草为研究对象,运用DSP和机器视觉技术对棉田杂草的实时识别技术进行研究,并为施药系统提供杂草位置信息主要的研究工作包括以下几个方面:(1)移动装置的改进与基于DSP的视觉系统的搭建根据实验室条件,对已有的除草机器人移动装置进行设计改进,提升了移动装置的动力和稳定性;根据棉田实时图像的采集和处理要求,选用摄像头、图像处理器等,搭建机器视觉系统2)基于DSP的图像分割方法的研究对图像分割常用的五种灰度化方法(超绿法、归一化超绿法R.G色差法、G.B色差法和Cr色差法)和四种二值化方法(定阈值法、最优阈值法、迭代取阈值法、Otsu法)进行研究;通过对处理效果和处理耗时的分析研究,选择综合性能最优的图像分割方法;介绍了三种图像滤波方法(高斯滤波法、形态学滤波法和中值滤波法),综合高斯滤波和形态学滤波消除图像噪声,提高图像的品质;添加人工光源进行实验,得到更优的处理效果。
3)杂草识别方法的研究根据喷施系统中四个喷头的布置,将棉田图像平均分成四行;为了简化杂草识别过程,将田间杂草分为行间杂草和株间杂草,采取不同的识别方法,利用棉花种植信息识别行间杂草,利用叶片面积特征和形状特征识别株间杂草,并分析杂草的识别效果4)杂草信息研究与实时喷施试验对图像区域进行划分,结合摄像机标定技术,用处方图来表示杂草信息,并把生成的处方图通过串口发送给微控制器搭建室内试验平台,来模拟除草机器人和棉田的相对运动,将机器视觉子系统和喷药控制子系统安装在台架上进行实时喷施试验,并对施药结果进行分析本文研究开发了智能除草机器视觉系统,并基于此系统对棉田杂草实时识别研究通过对图像分割方法、杂草识别方法和杂草信息获取的研究,综合考虑效果和实时性,择优选择了图像处理的方法并将其应用到智能除草移动平台上进行室内模拟试验,为基于机器视觉的棉田精确喷施除草装置的研发提供了理论和实践依据关键词:精确农业,DSP,杂草,实时,识别基于DSP的除草机器人杂草实时识别与试验研究Ⅱ江苏大学硕士学位论文ABSTRACTThe precision agriculture is the inevitable trend of our country agriculturedevelopment.Variable spraying herbicide,as a key technologies of precision agriculture,could be precisely spraying herbicide on weed,thus it can effectively reduce herbicidesdosage.With cottons and weeds as the research object,this paper use machine visiontechnology to research real.time identification and location of weeds,and provide weedslocation information to spraying system.The research mainly concludes several aspects asfollows:(1)Improvement of mobile device and construction of machine vision system.basedon DSE Based on the laboratory conditions,do improved design on the existing mobiledevice,to promote the mobile devices motivation and stability.According to therequirements of real-time image acquisition and processing,choose camera,imageprocessor.etc,to build a machine vision system.(2)The study of image segmentation method based on DSE Research on the fivecommon gray processing methods(Excessive Green,Normalized Excessive Green,R。
GG.B and Cr color differential)and four binary processing methods(fixed threshold method,optimal threshold method,iterative threshold method and Otsu method);Through the studyof treatment effect and processing time-consuming,choose the best image segmentation ofcombination property;introduces three kinds of image filtering method(Gaussian Filter,Morphology Filter and Median Filter),mix Gaussian Filter and Morphology Filter toeliminate image noise and improve image quality;add artificial illuminant of experiment,obtains the better processing results.(3)The research of weed identification.According to the arrangement of foursprinklers in spraying system,divide the image into four rows averagely;In order tosimplify the identification of weed,divide field weeds into rows weeds and weeds betweenplants,and take different recognition method to them,use cotton cultivation information toidentify rows,and use area feature and shape feature of leaf to identify weeds betweenplants,and analysis identification effect and time—cost,and analysis the recognition resultof weeds.(4)The research of weeds information and real—time spraying test.Divide the image,with the camera calibration technology,Use prescription diagram to show weedsinformation,and send it to micro controller through serial port.Construct indoor testI!i基于DSP的除草机器人杂草实时识别与试验研究platform,in order to simulate the spraying effect of robot under different speeds.Put themachine vision subsystem and the spraying control subsystem on the stage,and analysisthe results of real-time spraying.This paper studied and developed intelligent weeding machine vision system,andbased on this system researched real-time recognition of weeds.Through the study ofimage segmentation method,weed identification method and acquisition of weedsinformation,considering the effect and time-cost,pick over image processing method.Andthis technology was applied to intelligent weeding mobile platform for indoor simulationtests,which provides theoretical and practical basis on the research of accurate sprayingweeding device based on machine vision.KEY WORDS:Precision Agriculture;DSP;weeds;real-time;recognitionIV江苏大学硕士学位论文目 录第一章绪 论………………………………………………………………………………………………………11.1研究目的和意义…………………………………………………………………一11.2国内外的研究现状………………………………………………………………..21.2.1杂草识别研究现状…………………………………………………………21.2.2 DSP应用研究现状………………………………………………………….41.3本文研究内容与技术路线……………………………………………………….51.3.1研究内容……………………………………………………………………51.3.2技术路线……………………………………………………………………61.4本章小结…………………………………………………………………………。
7第二章移动除草机器人系统总体设计……………………………………………………………82.1全液压驱动除草机器人总体设计……………………………………………….82.2移动装置设计改进………………………………………………………………92.2.1动力系统……………………………………………………………………92.2.2转向系统…………………………………………………………………..112.2.3视觉系统…………………………………………………………………122.2.4喷施系统…………………………………………………………………132.3 DSP视觉系统……………………………………………………………………142.3.1 DSP概述…………………………………………………………………………………………142.3.2视觉。
