Plumpness Recognition and Quantification of Rapeseeds using Computer Vision

摘要

The plumpness is an important index of crop seed.However, traditional measurements are time-consumingand labor intensive. The computer vision technology, whichmay offer more efficient and non-destructive methods formeasurement, has recently appeared. But it is very difficultto accurately estimate the plumpness of single seed by theratio between area and perimeter because of the diversity ofrapeseed seed’s size. This paper focused on rapeseed seedplumpness recognition and quantification, based oncomputer vision. A new method, the coefficient of variationof radius (CVR), was used to estimate seed plumpness. Therecognition and quantification model for plumpness insingle seed were established by using the fuzzy C-means(FCM) clustering and fuzzy math method. The plumpness ofthe seed is full if plumpness is greater than or equal to 0.6.Some correlative index are calculated and analyzed to verifythe validity of this method. The tests show that there is nocorrelation between plumpness or plumpness ratio, and1000-seed weight or equivalence diameter. But there aresignificantly partial correlation between plumpness orplumpness ratio, 1000-seed weight and equivalence diameter.Finally, plumpness ratio index is significantly differentamong the 12 varieties rapeseed was determined. With themean value of plumpness ratio of rapeseed variety, theplumpness degree was plotted 10 grades. The results showthat the application of computer vision technology issignificantly valid for quantitative determination ofplumpness in rapeseed seed.