Surface damage of structures is an important indicator that further inspection of the structure is needed. However, existing detection methods rarely detect three-dimensional data of damage. Aiming at the inspection of structural surface damage, this study proposes a method based on 3-D structural surface damage reconstruction techniques for reconstructing and extracting data for damage volume calculation. The surface damage of concrete specimen is three-dimensionally reconstructed using multi-view smartphone-taken images and compared with a depth camera. The point cloud data was obtained, and then the damage plane was fitted and removed by a Random Sample Consensus algorithm to obtain the damage site data, finally, the damage volume was calculated. In the experimental process, accuracy and post-processing difficulty, equipment cost, data acquisition efficiency, and overall applicability of the two methods were compared and analyzed. In conclusion, it was determined that the 3-D damage parameters obtained through the smartphone multi-view stereo reconstruction performs better, at a lower cost and is more convenient for use in inspection work. Finally, the practicability of the method was proved by the damage detection experiment of real building.

During substation inspection, operators are often exposed to hazardous working environments. It is necessary to use visual sensors to determine work status and perform action detection to distinguish between normal and dangerous actions in order to ensure the safety of operators. However, due to information security, privacy protection, and the rarity of dangerous scenarios, there is a scarcity of related visual action datasets. To address this issue, this study first introduces a virtual work platform, which includes a controller for the parameterized control of scenarios and human resources. It can simulate realistic substation inspection operations and generate synthetic action datasets using domain randomization and behavior tree logic. Subsequently, a spatiotemporal action detection algorithm is utilized for action detection, employing YOLOv8 as the human detector, Vision Transformer as the backbone network, and SlowFast as the action detection architecture. Model training is conducted using three datasets: a real dataset, a synthetic dataset generated via a VWP, and a mixed dataset comprising both real and synthetic data. Finally, using the model trained on the real dataset as a baseline, the evaluation results on the test set shows that the use of synthetic datasets in training improves the model’s average precision by up to 10.7%, with a maximum average precision of 73.61%. This demonstrates the feasibility, effectiveness, and robustness of synthetic data.

To reduce temperature-induced measurement effects and the complexity of temperature compensation in long-term structural strain monitoring, this study proposes a machine-vision-based strain measurement method with reduced temperature sensitivity and develops a vision strain sensor incorporating fused-silica sensing elements. By integrating low-thermal-expansion fused-silica sensing elements with a dual-target visual measurement configuration, together with template matching and sub-pixel localization algorithms, the sensor achieves high-precision structural strain measurement. Static loading tests demonstrate strong agreement between the sensor and a linear variable differential transformer (LVDT), with a maximum root mean square error of 4.908 με. The sensor also exhibits good zero-drift stability and repeatability. Temperature cycling tests further show that, over the range of -20 °C to 50 °C, the sensor exhibits an average equivalent temperature sensitivity coefficient of 1.32 με/°C, indicating that it can effectively suppress temperature-induced additional responses at the sensor level and reduce reliance on post-compensation. A 20-day monitoring test on a steel member further demonstrates that the strain variation measured by the sensor agrees well with the theoretical thermal strain of the member in both overall trend and magnitude, thereby verifying its stable operational capability and its ability to characterize thermally induced structural strain responses. This method provides a feasible technical route for long-term continuous strain measurement in structural health monitoring with reduced dependence on temperature compensation.