Sub-pixel concentricity estimation algorithm for industrial machine vision with an accompanying dataset

Industrial machine vision has become an indispensable technology, especially in mass production environments. Many geometric features can now be measured quickly and accurately using machine vision algorithms. In this study, we propose a subpixel-based algorithm for measuring the concentricity of washer-type parts. The proposed method accurately determines the center of the part by analyzing intensity variations in the transitional regions, offering a new perspective compared to classical approaches. As a second contribution, this study addresses the limited access to datasets often due to commercial constraints by introducing an open-access dataset. The dataset includes 45 washer-type parts measured using a Coordinate Measuring Machine (CMM) and imaged under four different camera configurations. In addition to concentricity, the dataset provides comprehensive geometric measurements for each part. Experimental results demonstrate that the proposed method outperforms two existing algorithms and a commercial industrial vision system. Specifically, it achieved the best result on 45 samples. The source code and dataset used in this study are available at: https://github.com/agpoyraz/Concentricity_Measurement