Rapeseed Seeds Classification by Machine Vision

The implementation of new methods for fast classification of seeds is of major technical importance in the large-scale investigation of seeds identification. A few indices with biological significance were used to identify rapeseed type and variety. The plumpness and the plumpness ratio of rapeseed were extracted by using the variation coefficient of radius for fourteen rapeseed varieties at Chinese five locations. The equivalence diameter was extracted by using reference calibration method. The major color ratio and color saturation were extracted by using nine color HSV model and major color method. ANN-BP models were established for rapeseed varieties classification. The equivalent diameter and plumpness ratio were used to identify two classes´ rapeseed, B.compestris and B. napus L. with accuracy of 100%. The equivalent diameter, plumpness ratio and color saturation were used to identify six rapeseed varieties, with accuracy of 92.06% to 92.37%,, with average of 92.15%. By machine vision, it is feasible to identify rapeseed type and variety with a few biological indices.