An adaptive camera jitter compensation method for vision-based vibration measurement of rotating bodies

View at publisher Summarise with VIRAT
Authors: Mingyu Fan, Sen Wang, Gao Junfang, Tao Fu, Liying Zhu
Cited by: 0
References: 40

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In vision-based vibration measurement, industrial cameras are highly susceptible to irregular jitter induced by environmental factors such as airflow or ground vibrations, which severely compromises measurement accuracy. To address this challenge, this paper proposes an adaptive visual measurement method for the steady-state compensation of rotor vibration displacements under camera jitter conditions. Operating from a unified perspective, the proposed method leverages the relative motion characteristics between moving targets and stationary reference objects. Initially, the jitter-corrupted camera captures a video sequence of the target scene. A dual-branch deep learning network is then constructed, integrating RepCSP-CIB modules to enhance multi-scale feature fusion and AKConv modules to dynamically adjust convolutional kernel parameters. This architecture enables the precise estimation and compensation of camera jitter. An automatically annotated dataset is fed into the network to simultaneously extract the vibration displacements of both the measurement target and a stationary reference object. The camera jitter error signal is subsequently quantified by tracking the apparent motion of the stationary reference. Finally, the true rotor displacement is reconstructed by decoupling the jitter component from the synthesized rotor vibration signal. Experimental results demonstrate the exceptional efficacy of this method in resolving rotor vibration profiles under camera jitter. A comprehensive comparison between the regressed camera jitter signal and the ground-truth data—acquired via an eddy current displacement sensor—reveals that the proposed approach achieves the highest fidelity to the actual displacement, thereby fulfilling rigorous experimental requirements.

Fields of Study & Subjects

Machine Fault Diagnosis Techniques Image and Video Stabilization Advanced Measurement and Detection Methods

Publication details

Journal / source
Engineering Research Express
Volume
8
Issue
7
Pages
075503–075503
Publication date
31 March 2026
Language
EN