Vision based Roughness Average Value Detection using YOLOv5 and EasyOCR

A Rough Surface involves a lot of imperfections and is prone to friction as it offers resistance to moving objects on the surface. The roughness of a Surface is an indicator of the probable performance of every mechanical component since imperfections on the surface might further lead to the formation of nucleation sites for corrosion or ruptures. As rough surfaces have higher friction coefficients as compared to smooth surfaces, it becomes absolutely imperative to test surface roughness and take appropriate action before deployment in automobiles and other industries in order to maintain safety standards. Surface roughness is a calculation of the relative roughness of a surface profile based on a single numeric parameter, Average Roughness (RA). RA is the most commonly specified surface texture parameter measured using a Stylus based instrument wherein a small tip is dragged across any surface while its undulations are recorded which provides a general measure of surface texture in microns. This paper proposes a Machine Learning model developed to read the detected value from the RA Tester and store it in the database thereby reducing manual interference. This proposed model uses a pipeline of the YOLOv5 Algorithm and EasyOCR to detect the Region Of Interest (ROI) from the image and the RA values respectively. This paper produces a real-time solution with an Accuracy of 95.3% for an Automated Entry of the Roughness Average values read directly from the image into the database and has been implemented successfully in the Automobile Industry. This project was conceptualized and Implemented jointly by KLE Technological University and Dana Anand India Private Limited, Dharwad, India.