The Detection of the Surface Defects in Materials by Distributed and Invariant Features of Image Intensity

Detection of metal surface defects is an important and necessary quality control process for industrial products. There are numerous systems available to identify material and texture defects, most of them are complex and require ample data to initiate the process. In this paper, we consider a method of converting metal surface images into binary histogram representation for further examination of the obtained images for the presence of defects in the original images. It has been found that with the right parameters set this method can meet accuracy requirements for metallic defect detection. Finally, with further transferring calculations to the GPU, we expect working time at the level of the fastest neural networks.