Spherical transform of fruit images for on?line defect extraction of mass objects

Abstract

Defects on spherical curved objects cannot be easily identified because of the object image boundary effect. Spherical transform methods are developed to solve the problem and successfully applied in an automated defect sorter. A spherical transform is obtained by compensating the intensity gradience on curved objects. Defects below the background level can be extracted through a preservation transform. The defect extraction is enabled by a uniformly distributed plane image through two‐step transformations and the defect position is determined by allocation process. The results show the effectiveness of the processing methods for the high‐speed on‐line defect identification on fruit packing lines. © 1996 Society of Photo−Optical Instrumentation Engineers.