The main contents of the thesis are shown as follows : presenting fundamental theories of statistic pattern recognition , discussing rgb ( red , green , blue ) color space , ohta color space , hsi ( hue , saturation , intensity ) color space and its converted color space , materials consistency in gray scale and the application in removing foreign bodies in tobacco flows , hence presenting recognition pattern based on " unit recognition " , designing sample machine for this purpose , which consists of material - providing system , optic system , image - grabbing system , real - time intelligent image - processing system and systems of automatically rejecting foreign bodies and self - diagnosis , analyzing and optimizing hard wares , offering concrete designs such as optic system and air - ejector driver circuit , presenting and realizing physical ram 本文的主要內(nèi)容有:統(tǒng)計(jì)模式識(shí)別基礎(chǔ)理論及它們?cè)跓煵莓愇镒R(shí)別中的應(yīng)用;討論了rgb ( red 、 green 、 blue )基礎(chǔ)顏色空間、 ohta顏色空間、 hsi ( hue色調(diào), saturation飽和度, intensity亮度)顏色空間及其變換空間、物料圖像紋理、灰度均勻性等在煙草異物識(shí)別中的應(yīng)用,并在此基礎(chǔ)上提出了"基于判別單元顏色統(tǒng)計(jì)特性"的煙草在線異物識(shí)別模型,設(shè)計(jì)并研制了煙草在線異物實(shí)時(shí)識(shí)別與自動(dòng)剔除系統(tǒng)原理樣機(jī),它由供料系統(tǒng)、光學(xué)系統(tǒng)、圖像數(shù)據(jù)采集系統(tǒng)、實(shí)時(shí)智能圖像處理系統(tǒng)、異物自動(dòng)剔除系統(tǒng)以及自診斷系統(tǒng)等組成。