Segmented evaluation and multivariate temporal compensation method for thermally induced image drift in chain cameras for bridge deflection monitoring

Accurate bridge deflection monitoring is vital, yet conventional single-camera vision methods are limited in measurement range. Chain-camera systems overcome this constraint but suffer from thermally induced image drift during camera startup, which compromises accuracy. This study proposes a two-step evaluation and compensation approach for startup-induced thermal drift in chain cameras. A segmented linear model is established to characterize the drift evolution, revealing a multi-stage drift behavior and showing similarity with the experimental data. An engineering-oriented lightweight compensation model is constructed by integrating measurement time, camera temperature, and ambient temperature to characterize startup-induced thermal drift. Experimental results demonstrate that the proposed multivariate fusion method is effective and outperforms commonly used alternative modeling approaches. Field application on an in-service bridge preliminarily demonstrates the potential of the proposed compensation method for practical monitoring scenarios. Overall, this study provides a lightweight preliminary approach for mitigating startup-induced thermal drift, with potential to improve the accuracy of bridge deflection monitoring under appropriate conditions.