本文主要研究内容
作者纪宇慧（2019）在《机器视觉在苹果疤痕识别和颜色分级中的应用》一文中研究指出：我国苹果种植面积广,产量巨大,种类繁多。但由于我国水果生产加工企业普遍采用人工分级与机械分级,与国外的分级技术相比较落后,分级精度偏低,同一批次的苹果质量参差不齐,使得我国苹果在国际贸易市场上成交量小,且价格较低。利用机器视觉技术对苹果品质进行自动分级检测,可有效地提高分级精度和效率,不仅能够满足各种消费人群的需求,为企业带来良好的效益,还可以提高我国苹果的出口量和出口价格,增加外汇收入。本文以烟台地区的红富士苹果为研究对象,基于机器视觉技术对苹果的果面缺陷与颜色两个特征进行研究,主要研究内容如下:(1)在图像预处理阶段,本文对RGB与HSI颜色模型进行分析,根据两种模型的特点,确定其适用情况。通过在灰度图像中加入白噪声,利用信噪比PSNR对几种常用滤波算法的去噪效果进行对比和评价,选定了有效性和实时性相对优越的快速中值滤波法。(2)在图像分割阶段,对比R、G、B、H、S、I六个分量空间下的灰度图以及灰度直方图,发现S分量空间下背景与目标的对比度较高且灰度直方图呈现标准的双峰状,最有利于采用全局阈值法进行图像分割。通过分析几种典型的全局阈值分割法,选用一种自适应的阈值分割方法——最大类间方差法。对于分割后的二值图像背景中由传送链条引起的噪声区域,采用形态学滤波中的删除小面积对象法进行处理,获得了良好的去噪效果。(3)提出基于Canny边缘检测算子的果面疤痕识别方法。对图像分割后获取的苹果果实图像,分别用一阶的Roberts算子、Sobel算子与二阶的LoG(Laplacian of Gaussian)算子、Canny算子对其进行疤痕检测,其中Canny算子提取的边缘最完整,且没有虚假边缘。为了将疤痕区域分割出来,获得疤痕区域面积所占果实面积的百分比,本文利用空洞填充法对疤痕区域进行填充,并引入形态学中的开运算将果实边缘造成的干扰去除。(4)考虑到外观品质高的苹果不仅着色率高,通常颜色分布也较均匀,所以在提取颜色特征时不仅选用色度分量,也选用了能够反映颜色分布的R、G、B分量的均值与方差作为特征参数。由于特征参数的个数较多不利于分级,提出了基于Fisher系数及K-means的颜色特征参数优化方法。通过计算每个特征参数的Fisher系数,按照其大小进行无类别的全局优化和分类别(着色率和颜色分布两类)的局部优化,并采用K-means算法对两种优化方式进行评估,聚类结果表明全局优化的效果更好。然后利用PS0(Particle Swarm Optimization)算法优化后的支持向量机对苹果的颜色等级进行自动划分,对全局优化方式下所保留的特征参数个数进行逐一分级测试,在保留Fisher系数较高的7个特征量时,分级正确率最高,达到92%。最后用Matlab软件设计了用于分级操作、模型参数与分级结果显示的GUI产品。
Abstract
wo guo ping guo chong zhi mian ji an ,chan liang ju da ,chong lei fan duo 。dan you yu wo guo shui guo sheng chan jia gong qi ye pu bian cai yong ren gong fen ji yu ji xie fen ji ,yu guo wai de fen ji ji shu xiang bi jiao la hou ,fen ji jing du pian di ,tong yi pi ci de ping guo zhi liang can cha bu ji ,shi de wo guo ping guo zai guo ji mao yi shi chang shang cheng jiao liang xiao ,ju jia ge jiao di 。li yong ji qi shi jiao ji shu dui ping guo pin zhi jin hang zi dong fen ji jian ce ,ke you xiao de di gao fen ji jing du he xiao lv ,bu jin neng gou man zu ge chong xiao fei ren qun de xu qiu ,wei qi ye dai lai liang hao de xiao yi ,hai ke yi di gao wo guo ping guo de chu kou liang he chu kou jia ge ,zeng jia wai hui shou ru 。ben wen yi yan tai de ou de gong fu shi ping guo wei yan jiu dui xiang ,ji yu ji qi shi jiao ji shu dui ping guo de guo mian que xian yu yan se liang ge te zheng jin hang yan jiu ,zhu yao yan jiu nei rong ru xia :(1)zai tu xiang yu chu li jie duan ,ben wen dui RGByu HSIyan se mo xing jin hang fen xi ,gen ju liang chong mo xing de te dian ,que ding ji kuo yong qing kuang 。tong guo zai hui du tu xiang zhong jia ru bai zao sheng ,li yong xin zao bi PSNRdui ji chong chang yong lv bo suan fa de qu zao xiao guo jin hang dui bi he ping jia ,shua ding le you xiao xing he shi shi xing xiang dui you yue de kuai su zhong zhi lv bo fa 。(2)zai tu xiang fen ge jie duan ,dui bi R、G、B、H、S、Iliu ge fen liang kong jian xia de hui du tu yi ji hui du zhi fang tu ,fa xian Sfen liang kong jian xia bei jing yu mu biao de dui bi du jiao gao ju hui du zhi fang tu cheng xian biao zhun de shuang feng zhuang ,zui you li yu cai yong quan ju yu zhi fa jin hang tu xiang fen ge 。tong guo fen xi ji chong dian xing de quan ju yu zhi fen ge fa ,shua yong yi chong zi kuo ying de yu zhi fen ge fang fa ——zui da lei jian fang cha fa 。dui yu fen ge hou de er zhi tu xiang bei jing zhong you chuan song lian tiao yin qi de zao sheng ou yu ,cai yong xing tai xue lv bo zhong de shan chu xiao mian ji dui xiang fa jin hang chu li ,huo de le liang hao de qu zao xiao guo 。(3)di chu ji yu Cannybian yuan jian ce suan zi de guo mian ba hen shi bie fang fa 。dui tu xiang fen ge hou huo qu de ping guo guo shi tu xiang ,fen bie yong yi jie de Robertssuan zi 、Sobelsuan zi yu er jie de LoG(Laplacian of Gaussian)suan zi 、Cannysuan zi dui ji jin hang ba hen jian ce ,ji zhong Cannysuan zi di qu de bian yuan zui wan zheng ,ju mei you xu jia bian yuan 。wei le jiang ba hen ou yu fen ge chu lai ,huo de ba hen ou yu mian ji suo zhan guo shi mian ji de bai fen bi ,ben wen li yong kong dong tian chong fa dui ba hen ou yu jin hang tian chong ,bing yin ru xing tai xue zhong de kai yun suan jiang guo shi bian yuan zao cheng de gan rao qu chu 。(4)kao lv dao wai guan pin zhi gao de ping guo bu jin zhao se lv gao ,tong chang yan se fen bu ye jiao jun yun ,suo yi zai di qu yan se te zheng shi bu jin shua yong se du fen liang ,ye shua yong le neng gou fan ying yan se fen bu de R、G、Bfen liang de jun zhi yu fang cha zuo wei te zheng can shu 。you yu te zheng can shu de ge shu jiao duo bu li yu fen ji ,di chu le ji yu Fisherji shu ji K-meansde yan se te zheng can shu you hua fang fa 。tong guo ji suan mei ge te zheng can shu de Fisherji shu ,an zhao ji da xiao jin hang mo lei bie de quan ju you hua he fen lei bie (zhao se lv he yan se fen bu liang lei )de ju bu you hua ,bing cai yong K-meanssuan fa dui liang chong you hua fang shi jin hang ping gu ,ju lei jie guo biao ming quan ju you hua de xiao guo geng hao 。ran hou li yong PS0(Particle Swarm Optimization)suan fa you hua hou de zhi chi xiang liang ji dui ping guo de yan se deng ji jin hang zi dong hua fen ,dui quan ju you hua fang shi xia suo bao liu de te zheng can shu ge shu jin hang zhu yi fen ji ce shi ,zai bao liu Fisherji shu jiao gao de 7ge te zheng liang shi ,fen ji zheng que lv zui gao ,da dao 92%。zui hou yong Matlabruan jian she ji le yong yu fen ji cao zuo 、mo xing can shu yu fen ji jie guo xian shi de GUIchan pin 。
论文参考文献
论文详细介绍
论文作者分别是来自济南大学的纪宇慧，发表于刊物济南大学2019-10-31论文，是一篇关于机器视觉论文,苹果疤痕识别论文,颜色分级论文,边缘检测论文,特征参数优化论文,济南大学2019-10-31论文的文章。本文可供学术参考使用，各位学者可以免费参考阅读下载，文章观点不代表本站观点，资料来自济南大学2019-10-31论文网站，若本站收录的文献无意侵犯了您的著作版权，请联系我们删除。
标签：机器视觉论文; 苹果疤痕识别论文; 颜色分级论文; 边缘检测论文; 特征参数优化论文; 济南大学2019-10-31论文;