Brightness-invariant image segmentation for on-line fruit defect detection

The extraction of defective segments from fruit images is a critical step for on-line fruit defect detection using machine vision. To accommodate the gradient reflectance characteristics of objects with curved surfaces such as apples and to avoid the defect inspection error due to the natural brightness variability of fruit, a novel and practical brightness-invariant image segmentation method is developed. Experi- mental results from our vision-sorting system show that the proposed method is effective for large-scale multiple-line processing of massive numbers of apples.