Research of Real-Time Algorithm for Chestnut´s Size Based on Computer Vision

A grading system was developed to classify chestnut automatically into various grades of quality in terms of size. The chestnut is scanned with a color charge-coupled-device camera and then the size is extracted by image processing. In the image processing there are two kinds of algorithm, one is the minimum enclosing rectangle (MER), and the other is the distance between centroid and border (DCB). The algorithm finds out the chestnut´s major axis and minor axis which are used to predict the size of the chestnut. By regression analysis, it´s found that the relative error of the MER is 0.7465%, and the relative error of the DBCB is 1.83%. The MER is better to predict the chestnut´s size than the DCB.