Research Of Image Processing Algorithm On Automatic Grading Of Tobacco Leaf～2

The quality of tobacco leaf influences the technics and quality of tobacco productdirectly. Nowdays, the extemal quality of tobacco leaf is measured by senses throughoutthe worid. Although study on automatic grading of tobacco leaf is carrying out, there isno matue product. So it is essental fOr me to do further research in this field.Wth the target of measuring the external quality of tobacco leaf the color featureof tobacco leaf is detected, the correlativity of three paraxneters of RGB analyzed, aNg model based on color featUre is builded uP.In this article, the fOllowing contens are mainly illustrated in detail:l. The idea of the systemic design scheme is established correctly. The related parametersof systern are designed;2. The tranSformation from RGB color SPace via XYZ color sPace and L*a*b* color spaceto HVC color sPace Which is fit for hUInan's color percePtion;3. Analyzc the correativity of three paramthes of RGB using regression. Fit nonlinearequation using polynoAnal regression. Determining the steP of polynomial equationusing stapwise Fit linear equation.4. Use geometric operation to achieve decorrelation. Describe fethee sPace with cuboidand elliPsoid and accordingly enhance the efficiency offeatUr sPace;5. Define a kind of distance according to the characteristic of tobacco leaf The distanceconsists of the differene of tWo featde sPaces, a normalized diStance and sizeproportion of tWo featUr sPaces;6. Choose tWo pouPs of standard tobacco leaves and a grOuP of pending tobacco 1eaves tovalidae the algOrithIn. More efficien e1liPsoid featUre sPaces are Obtained afterregressing and geometric operating. The result of grading is satisfactory.