muqian，guoneiwaiyanjiurenyuanyijingguangfandikaizhanleguoshucaizhaijiqirendeyanjiu。zaiziranhuanjingtiaojianxia，shixianguoshidezhunqueshibieyudingwei，shiguoshucaizhaijiqirenshijiaoxitongdedejichuheguanjianjishu

，zhijieyingxiangcaizhaijiqirendezuoyexiaolv。suoyi，kekaoxinghao、精jing度du高gao的de視shi覺jiao識shi別bie技ji術shu，能neng夠gou檢jian測ce出chu所suo有you成cheng熟shu果guo實shi
，並bing對dui其qi進jin行xing精jing確que定ding位wei已yi經jing成cheng為wei智zhi能neng采cai摘zhai機ji器qi人ren非fei常chang重zhong要yao的de一yi個ge環huan節jie。從cong現xian有you文wen獻xian看kan，果guo蔬shu采cai摘zhai機ji器qi人ren的de視shi覺jiao識shi別bie係xi統tong主zhu要yao原yuan理li是shi利li用yong顏yan色se
、形狀、紋理和近紅外反射等從背景中區分植株、水果和蔬菜，多數研究在室內和可控光照條件下進行。

我們結合前人研究成果，提出在自然環境下
，利用R-G色差圖像進行改進 OTSU 算suan法fa分fen割ge，對dui區qu域yu進jin行xing標biao記ji，計ji算suan出chu連lian通tong區qu域yu的de麵mian積ji和he質zhi心xin，進jin而er推tui出chu同tong麵mian積ji質zhi心xin圓yuan。自zi然ran場chang景jing下xia，成cheng像xiang條tiao件jian多duo變bian，光guang線xian明ming暗an不bu一yi

，場chang景jing複fu雜za多duo樣yang。要yao去qu除chu的de背bei景jing既ji有you樹shu枝zhi、樹葉和未成熟蘋果，也有天空

、caodideng。weimanzupingguocaizhaijiqirenzhugecaizhaichengshupingguodeyaoqiu，xuzhugeshibiechengshupingguobingtiquxiangguantezheng。zaiduipingguojinxingpaisheshi，youyuhuanjinghejiaodudengyinsudeyingxiang，xiangjixitongnanmianshoudaoyingxiang，suoyituxiangzhongcunzaizaosheng，duituxiangzhengqueshibiechanshengganrao
，yaoduituxiangjinxinglvboquzaochuli。zhifangtushuangfengfayuzhishourenweijitongjiyangbendeyingxiangjiaoda，duiyuyouxietuxiangrongyizaochengguodufenge。diedaifajinkaolvdaoxiangsubenshendehuidu，erhulveleqilingyudekongjianxinxi，qiediedaijisuandeyuzhijieguoshouchushiyuzhideyingxiang。

OTSU fajiangxiangsubenshendehuiduyijilingyupingjunhuiduzonghepanduan，genjuzuidaleijianfangchazidonghuoquzuijiayuzhijinxingtuxiangfenge，yidingchengdushangxiaochulehuanjingguangxianbianhuadeyingxiang，fengejiaoweiwending。danzaishiyanzhongfaxian，dangtuxiangzhongmubiaoyubeijingchayixianzhushi
，qizidongquedingdeyuzhiyibanjunnengjiaohaodijiangmubiaocongbeijingzhongfengechulai，dangmubiaoyubeijingjiaoweifuzahuomubiaowuyubeijingchayibumingxianshi，huichuxianjiaoweiyanzhongdeguodufengehuodakuaiheisequyu，fengexiaoguobulixiang。houxudepingguodingweizhuyaoshitezhengdiandexuanquhepipei
，youyumianjitezhengchengxiansuijibutong，qiejisuanliangda; 圓度特征對分割算法的依賴性很大，容易造成誤匹配，所以麵積
、圓度等特征匹配不適合作為匹配特征。由於蘋果的形狀接近於球體，對於每個成熟蘋果采集的左、右圖像上的質心是一對極好的匹配點，選擇其作為匹配特征點可以使問題簡化，且易於實現。中國糧油儀器網  http://www.98fo.cn/