This study presents ADSFormer, a Transformer-driven active perception framework designed for autonomous robotic damage inspection. In contrast to traditional raster scanning and passive vision methods, ADSFormer formulates inspection as a Partially Observable Markov Decision Process (POMDP), allowing the robot to actively select viewpoints based on segmentation uncertainty. The framework incorporates episodic memory and hierarchical action modeling, enabling efficient integration of multi-view information and adaptive exploration. By leveraging uncertainty cues, the system refines sub-area predictions and enhances segmentation robustness. Experiments on photorealistic metallic surface datasets demonstrate that ADSFormer achieves over a threefold improvement in damage mIoU and reduces inspection time by more than 50% compared to dense raster scanning. Furthermore, it surpasses the prior ADS-DRL baseline by approximately 45% in damage IoU while maintaining similar runtime efficiency. These findings highlight the advantages of coupling uncertainty quantification with structured policy learning for accurate, data-efficient, and scalable robotic inspection. The framework’s capabilities and limitations are further evaluated across diverse inspection environments.

As bridges age, timely and accurate identification and quantification of cracks and structural defects are essential for ensuring safety and durability. Existing 2D image-based methods are limited by the lack of reliable depth information, hindering precise description of defect shape, severity, and evolution. In contrast, 3D point cloud data provides high-resolution spatial geometric information, supporting more accurate defect analysis. This paper proposes a scale-adaptive cross-modal defect detection and quantification method that integrates 2D image data with 3D point cloud information. First, a 3D semantic segmentation model with a cross-modal registration strategy enables precise segmentation of large structural cracks, while an image-guided fusion mechanism facilitates accurate depth map generation for micro-cracks. Additionally, a 2D segmentation model-guided defect 3D point cloud segmentation algorithm is proposed for detecting small durability cracks. Finally, a 3D defect quantification framework is developed to systematically measure crack length, width, depth, and volume, providing multi-dimensional indicators for structural assessment. Experimental results show that the proposed cross-modal fusion approach significantly outperforms single-modality methods. By integrating PointNet++ and U-Net, the system achieves high-precision multi-scale segmentation with an average mIoU of 0.87–0.89. Furthermore, the framework enables reliable 3D quantification of crack length, width, depth, and volume, providing a more comprehensive and objective basis for structural health assessment than traditional 2D methods. This research provides an effective approach for overcoming the limitations of 2D methods, offering a reliable solution for bridge structural health monitoring and durability prediction.

The continuous evolution of Structural Health Monitoring (SHM) calls for innovative, high-resolution, and non-invasive sensing strategies capable of detecting early-stage damage in civil infrastructure. Among emerging technologies, liquid crystal (LC) materials present a unique opportunity due to their intrinsic ability to produce visually responsive patterns under mechanical stress. This paper investigates the feasibility of employing LC films as a contactless optical sensing layer for monitoring strain localisation and crack initiation in concrete structures. This approach uses the colour-changing and light-bending properties of liquid crystals to turn mechanical stress into visible patterns, allowing real-time, full-surface visualization of strain without the need for special surface treatment, speckle patterns, or lighting. Concrete specimens were instrumented with LC coatings and subjected to controlled loading while being monitored with high resolution camera. The experimental observations demonstrate that LC responses reliably correspond with strain distribution and microcrack development, offering high contrast and sensitivity even at early deformation stages. These results were validated against conventional digital image correlation measurements, confirming that LC-based contactless sensing can complement existing vision-based SHM methods by improving detectability, reducing dependency on surface preparation, and potentially simplifying monitoring setups. The findings highlight several advantages of integrating LC materials into SHM frameworks; enhanced sensitivity to micro-scale damage, compatibility with non-contact imaging systems, and the ability to provide visual feedback for structural assessments. Furthermore, the approach has potential for scalability to larger structures and integration with automated monitoring platforms, aligning with ongoing advances in hybrid sensing and intelligent infrastructure management. This work positions liquid crystal optical sensing as a promising, innovative tool for SHM, offering a pathway toward more robust, high-resolution, and accessible monitoring strategies. By bridging material science with advanced optical techniques, the study contributes to emerging of non-invasive SHM methods capable of supporting safer, more resilient infrastructure.

Structural health monitoring (SHM) has advanced considerably in recent years, supported by developments in sensor and data acquisition systems, and computational algorithms. Among these innovations, computer vision has emerged as a particularly powerful tool for enhancing the understanding of bridge behaviour. Its integration into monitoring systems and maintenance strategies enables non-invasive, high-resolution assessments that complement traditional measurement techniques and strengthen diagnostic confidence.

The Pirâmides Railway Bridge, located on the freight rail branch connecting the Port of Aveiro to the Northern Line, spans the Pirâmides channel of the Aveiro lagoon. This prestressed concrete cable-stayed bridge has a total length of 175 m between joints and consists of four spans: a 75 m central span and three 25 m lateral spans. The deck features a constant U-shaped cross-section, 7.80 m wide, with 1.20 m-wide and 1.60 m-high longitudinal girders and a variable-thickness slab between 0.40 m and 0.45 m. The structure main span is supported by pyramidal piers (P116, P117) at midspan, where 15 m-high masts anchor the suspension and back-stay cables. The suspension and back-stay systems comprise four 37-strand cables per line, arranged in two vertical planes aligned with the longitudinal girders.

Over time, cracking has been observed in several structural elements such as the towers, raising concerns about the presence and progression of expansive reactions within the concrete. To evaluate this behaviour, a novel computer-vision-based methodology was developed. The approach adapts established procedures for assessing concrete expansibility and applies them to high-resolution images of the crack patterns on affected elements. By enabling non-contact monitoring of crack geometry, distribution, and evolution, the method provides detailed insights into the progression of damage that would be difficult to achieve using traditional techniques alone.

This study demonstrates the potential of computer vision as a reliable and efficient tool for structural diagnostics, particularly when conventional methods are limited, intrusive, or insufficiently sensitive. The integration of image-based analysis with engineering standards offers a promising path for early detection, continuous assessment, and long-term monitoring of expansive phenomena in concrete bridge structures.

Vision-based structural health monitoring of stay cables has predominantly relied on dynamic imaging to extract vibration characteristics for tension estimation, typically requiring video sequences, high-speed cameras, and stable close-range deployment. This study presents an alternative static-image approach that estimates stay-cable tension from gravitational sag captured in a single static digital image obtained from far-field photography. The method integrates a parabolic sag-tension relationship for inclined cables with a lightweight image-processing workflow, in which global thresholding and Canny edge detection are used to extract the cable profile, followed by quadratic fitting to determine sag and cable force. The procedure enables marker-free measurement and calibration-free deployment, allowing rapid field application with minimal on-site preparation. The proposed approach is intended for inspection scenarios in which cable arrays on one bridge side approximately lie in a common vertical plane and near-orthogonal far-field imaging can be achieved. Under such conditions, basic image segmentation techniques provide sufficient accuracy, although complex backgrounds or low-contrast scenes may reduce robustness and motivate the use of more advanced segmentation strategies. Laboratory strand experiments using a 61 MP customer-grade camera demonstrated small tension estimation errors under near-orthogonal imaging, and sensitivity studies showed that moderate viewing-angle deviations introduce limited bias. Field validation on the Kao-Ping-Hsi Bridge was conducted using far-field images acquired from distances of approximately 550 m. For the longest cables with larger sag amplitudes, estimated tensions agreed with independent reference values within 2%, while errors for the shortest and steepest cables remained below 8%. The results demonstrate that single-image gravitational sag extraction offers a practical, low-intervention alternative to dynamic vision-based techniques for remote stay-cable force assessment when sensor installation or vibration testing is impractical.