Preliminary results for the evaluation of the influence of noise in Computer-Vision Sub-Pixel Algorithms for Displacement Monitoring
F. Allegrezza, F. Micozzi, Michele Morici, A. Zona, A. Dall’Asta
TL;DR: This study investigates and compares three different algorithms for real-time displacement extraction, evaluating their performance under controlled conditions through synthetic video sequences with exactly known imposed motion, and identifies their strengths and limitations.
Vision-based displacement measurement has gained considerable attention in Structural Health Monitoring (SHM), where the need to detect small structural motions with non-contact instrumentation has driven the development of subpixel estimation algorithms capable of achieving resolution well below the nominal pixel size. This study investigates and compares three different algorithms for real-time displacement extraction, evaluating their performance under controlled conditions through synthetic video sequences with exactly known imposed motion. Particular attention is devoted to the systematic errors inherent to each algorithm and to how these are influenced by external factors typical of SHM real-world applications, namely optical blur and target size. The analysis is carried out using AprilTag markers as targets identifying the points whose displacements are tracked. The presented study assesses each algorithm in terms of accuracy, robustness, and computational efficiency, with the goal of identifying their strengths and limitations, thus, offering practical guidance for their application in real-world monitoring scenarios.