Automated surface roughness classification using robot vision

A robot vision system, also known as a Machine Vision (MV) system, enables automatic inspection using image processing techniques. This research focuses on the classification of turned surface images into four categories (A, B, C, and D) using deep learning algorithms with transfer learning. Images were captured under varying machining conditions (speed, feed, depth of cut) as per a full factorial experimental design. The dataset was divided with 70 % for training, 15 % for validation, and 15 % for testing the algorithms. Surface roughness parameters were analyzed using a robot vision system, comprising a Mitsubishi articulated robot with 6 degrees of freedom and a 4 kg payload, and a Cognex In-Sight 7801 camera (1.3 MP, 1280 × 1024 resolution). The performance of the models was evaluated based on average accuracy. The system demonstrated significant potential in enhancing the surface finish inspection process in high-production industries, reducing labor costs, inspection time, operator errors, and setup requirements, thereby increasing productivity and lowering production costs .