A machine vision system for egg crack detection based on a modified pressure chamber and a continuously rotating egg was designed and implemented. The proposed system successfully collected surface images of the entire egg surface using only one pressure modification, which reduced the operation time and risk of enlarged cracks. In particular, this work focused on identifying cracks on unwashed eggs that were intact or cracked combined with being dirty. A robust crack detection algorithm was developed which optimized feature selection. The experiments were validated using 750 egg surface images with 94% accuracy and 1.67% false negatives. (C) 2017 Elsevier Ltd. All rights reserved.