Image Segmentation for Fruit Based on Convexity of Illumination Distribution Features of Light Sources

CID known as Convexity of Illumination Distribution features of light sources can be utilized to recognize objects from the background, especially when it comes to the situation that the illumination is not uniform and the surface of the background is not very even or the edge of the object is not clear enough. In this paper, models of illumination distribution have been created to prove the convexity of illumination distribution on the background of fruit. Thus we are able to distinguish rotten places from the background by selecting control points from background points on the convex hull of the plot of a scan line of the image. Therefore, finally, we can utilize this control points to reconstruct the background by cubic spline and subtract this background from the original picture to get object. In order to test the efficiency, quantitative experiments are performed and they yield promising results.