Learning Defect Classifiers for Textured Surfaces Using Neural Networks and Statistical Feature Representations

Detecting surface defects is a challenging visual recognition problem arising in many processing steps during manufacturing. These defects occur with arbitrary size, shape and orientation. The challenges posed by this complexity have been combated with very special, runtime intensive and hand-designed feature representations. In this paper we present a machine vision system which uses basic patch statistics from raw image data combined with a two layer neural network to detect surface defects on arbitrary textured and weakly labeled image data. Evaluation on an artificial dataset with more than 6000 examples in addition to a real micro-cold forming process showed excellent classification results.