Deep Learning Based Feature Discriminability Boosted Intelligent Metal Surface Defect Detection System

Abstract

The application of computer vision and deep learning can be viewed as newlineparamount in the quality inspection processes that are carried out in enterprises. Over the newlinepast several years, there has been a growing need for automatic defect detection systems newlineduring the manufacturing or shipment stages. This is due to the fact that the identification of newlinedefects over metal surfaces and the determination of their position are both essential for newlinequality control procedures. Deep learning has been proven to be extremely effective to newlineexecute the defect detection and classification procedures. Metallic defect detection might newlinebe considered as a challenging task due to the fact that ambient elements, such as lighting newlineand the reflection of light, are the primary factors, that cause defects. Metal surfaces can be newlineimpacted by a variety of defects, and this research work specifically focuses on the newlineidentification of defects like crazing, patches, inclusions, scratches, pitted surfaces, and newlinerolled in scale. Since the fastness and accuracy in detection play a major role on the newlineproduction phase, it is desirable to enable a higher-level defect detection system in newlinemanufacturing industries. Considering the fact that quality control has a direct impact on newlineproduction and profit in multiple enterprises newline