UAVs are becoming increasingly used in the field of structural health monitoring, and the position information of them during the tasks is crucial. However, in complex scenarios such as bridges and high-rise buildings, the GNSS positioning method cannot obtain the positions of the UAV all the time due to the blockage of satellite signals and multi-path effects. This paper proposes a real-time positioning method to address the issue combining GNSS and total station. The GNSS is first used to locate the UAV when it is not in the line of sight of the total station, and the coordinates of the UAV are transmitted to the total station for blind tracking through coordinates conversion. The total station is then used to directly track the UAV when it flies to the GNSS-denied area and appears in the field view of the total station. Experiments show that the shift from blind tracking to direct tracking can be guaranteed as the coordinates conversion error is always less than the field of view range of the total station, even if only two common points are used for coordinates conversion. In addition, high positioning accuracy can be achieved in complex structural health monitoring scenarios.

Deep neural networks have been widely applied to fiber optic sensor systems, where the detection of external intrusion in metro tunnels is a major challenge; thus, how to achieve the optimal balance between resource consumption and accuracy is a critical issue. To address this issue, we propose a lightweight deep learning model, the Temporal Efficient Residual Network (TEResNet), for the detection of anomalous intrusion. In contrast to the majority of two-dimensional convolutional approaches, which require a deep architecture to encompass both low- and high-frequency domains, our methodology employs temporal convolutions and a compact residual network architecture. This allows the model to incorporate lower-level features into the higher-level feature formation in subsequent layers, leveraging informative features from the lower layers, and thus reducing the number of stacked layers for generating high-level features. As a result, the model achieves a superior performance with a relatively small number of layers. Moreover, the two-dimensional feature map is reduced in size to reduce the computational burden without adding parameters. This is crucial for enabling rapid intrusion detection. Experiments were conducted in the construction environment of the Guangzhou Metro, resulting in the creation of a dataset containing 6948 signal segments, which is publicly accessible. The results demonstrate that TEResNet outperforms the existing intrusion detection methods and advanced deep learning networks, achieving an accuracy of 97.12% and an F1 score of 96.15%. With only 48,009 learnable parameters, it provides an efficient and reliable solution for intrusion detection in metro tunnels, aligning with the growing demand for lightweight and robust information processing systems.

Urban road collapse is a common disaster in modern cities, posing a severe threat to the safety of life and property of urban residents. Effective risk assessment is crucial for preventing collapses. This study proposes a novel method for assessing the risk of urban road collapse by integrating Interferometric Synthetic Aperture Radar (InSAR) technology with Long Short-Term Memory (LSTM) networks. We conduct experimental analysis using collapse incidents and corresponding subsidence maps from five areas in Guangzhou. The results demonstrate that the proposed model successfully establishes a mapping relationship between the temporal features of InSAR data and the risk of road collapse, enabling a risk assessment of specific areas solely based on InSAR data. This model overcomes the difficulty of obtaining numerous assessment indicators in traditional risk assessment methods, providing strong support for the prevention and control of regional road collapse risks.