Title: Development and validation of an optical surface quality metric for product inspection

Authors: Tamal Ghosh

Addresses: Department of Computer Science and Engineering, Adamas University, AKC Campus, Barrackpore-Barasat Road, P.O. Jagannathpore, Barasat, 700126, West Bengal, India

Abstract: Surface roughness (Ra) is a key determinant of machining performance, influencing wear, lubrication, and component life. Conventional stylus profilometers, though accurate, are costly and slow, limiting their suitability for real-time inspection. This study presents a low-cost, non-contact optical pipeline for quantitative surface roughness estimation from microscope images. The system employs a Sony IMX477 camera with coaxial LED illumination and a Raspberry Pi 4 controller, with a total system cost is approximately USD 200-450, compared with USD 1.4k-50k for stylus profilometers. A selected set of GLCM-based texture features including cluster prominence, cluster shade, and sum variance strongly correlated with measured Ra (r = -0.94, p < 0.001). A Kohonen self-organising map (KSOM) - enhanced Random Forest regression model achieved cross-validated R² = 0.82 ± 0.12 and RMSE = 0.078 μm, demonstrating near-profilometer precision. The system processes images within 0.05-0.3 s per ROI, improving two-to-three-order speed over contact-based methods. The work establishes a reproducible, open-source hardware-software framework for rapid surface characterisation applicable to precision machining, tool-wear monitoring, and smart-manufacturing quality control. The paper details hardware design, statistical analysis, predictive modelling, and benchmarking against stylus profilometry.

Keywords: product quality measure; optical image analysis; feature engineering; non-contact quality inspection.

DOI: 10.1504/IJQET.2026.154321

International Journal of Quality Engineering and Technology, 2026 Vol.11 No.2, pp.178 - 198

Accepted: 21 Nov 2025
Published online: 22 Jun 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article