This study presents an improved method for classifying workpiece surface roughness through the use of techniques for processing digital images, overcoming the limitations of traditional measurement methods. The proposed method uses a combination of new feature parameters, gray level co-occurrence matrix (GLCM) optimization technique and Extreme Learning Machine (ELM) algorithm, abbreviated as OGLCM-ELM. The experimental system is built to collect images of workpiece surface with different roughness levels, under conditions of changing shooting angle, lighting and resolution. The collected data are divided into two main groups (light roughness and medium-heavy roughness) or three groups (light, medium, and heavy roughness). Experimental results show that OGLCM-ELM provides improved effectiveness relative to traditional techniques, through evaluation indexes such as accuracy, sensitivity (recall), F1-score and processing time. Notably, when performing binary classification, the OGLCM-ELM method achieved high accuracy even under complex experimental conditions. This shows that the method not only has good applicability in controlling the surface quality of workpieces but also contributes to improving efficiency in mechanical manufacturing.
Hien et al. (Sat,) studied this question.