Work

Preliminary Assessment of a Low-Cost DeepTag-Based Vision System for Structural Dynamic Testing

Year: 2026
Type: conference-paper
Abstract: Continuous and accurate health monitoring of civil structure requires dense spatial coverage of sensing points. Although piezoelectric accelerometers provide high precision measurement, their unit cost increases as the number of sensing location increases, limiting the scalability of the conventional contact-based system. Non-contact, vision-based approaches offer a scalable alternative, allowing multi-point measurement from a single imaging device. This paper proposes and validates a DeepTag-based marker vision system for structural acceleration estimation using a low-cost camera. Acceleration is derived from tracked marker displacements through a pipeline comprising Savitzky-Golay smoothing, double numerical differentiation, temporal synchronization, and FFT-guided bandpass filtering. The method is validated against simultaneous piezoelectric accelerometer measurements which were used as precision reference under controlled laboratory conditions using a mechanical shaker. The proposed vision-based pipeline achieved a mean Pearson correlation of$\mathrm{r}=\text{0. 9 4 7}$with the reference accelerometer on 30 repeated runs, alongside nearly exact agreement in the identification of the dominant frequency with a mean error of 0.0032 Hz, indicating the viability of low-cost camera-based vision detection for applications of structural vibration monitoring where the deployment of dense accelerometers is economically prohibitive.
Language: English
Open Access status: hybrid
OpenAlex ID: W7204200927
FWCI:
Cites: 10
Cited by:
Related to: 10