Dynamic Monitoring of Structures by Visual Based Techniques

Structural deterioration is a serious global challenge, as numerous constructions have exceeded their anticipated lifespan and design capacity. Traditional monitoring methods are often less efficient, unsustainable, complex, costly, and impractical. In the context of increasing digitalization in this modern era, visual-based techniques are emerging solutions and scalable alternatives to be adopted for monitoring the structures remotely. Visual-based monitoring through motion magnification has drawn growing research interest in this decade, and several studies have been conducted to explore its potential. However, some significant limitations still affect overall efficiency, and addressing these limitations is crucial to enhancing the practicality of visual-based approaches in real-world applications. This Thesis aims to address the key limitations and existing research gaps of prominent visual-based motion magnification methods (MMM) by proposing enhanced techniques for detecting frequencies, displacement, and damage in structures. A novel Hybrid approach is proposed by integrating two MM techniques for detecting motion: the Reisz phase-based steerable and the Lagrangian motion tracking. In addition, robust steerable Wavelet-based de-noising and Gaussian-kernel algorithms are utilised for effective background noise and distortion suppression. The filtering and data reduction techniques are developed to reduce overall complexity and computation time. Furthermore, an enhanced visual-based damage detection technique is proposed by employing Canny edge detection with image smoothing algorithms to visualise, detect, and precisely identify damage through region analysis and binary edge maps. Comprehensive investigations and comparative analyses were conducted against various traditional monitoring methods to validate the capability and precision of the proposed approach in both controlled and uncontrolled environments. Laboratory validations were carried out on two prototype models: a multi-storied frame and a suspended bridge. Field investigations were conducted on two full-scale structures: the Atina Bridge and the Ancient Quercia Bell Tower. The conducted validations and comparable investigations against traditional monitoring methods verified that the proposed visual-based approach is highly effective in determining the exact dynamic features of the structures under various operational and environmental conditions. The obtained results validated the potential of using the proposed visual-based techniques to detect dynamic features and damage while effectively mitigating the existing significant limitations. The proposed approach is practical, sustainable, cost-effective, and precise, with significantly reduced acquisition and computational time. The obtained promising findings substantially support the utilisation of the proposed visual-based techniques as an effective replacement to traditional SHM methods.