Xiaoman Wang, Dan Yuan, Xin Liu, Yike Zhao, Xiaoxiao Zhang, Xizhi Chen, Yunshi Lan. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

With the rapid urbanization and proliferation of high-rise buildings in China, facade damage such as cracking, peeling, and tile detachment has become increasingly prevalent, posing severe safety risks to the public. Traditional visual inspection methods are labor-intensive, subjective, and lack quantitative standards, often resulting in missed or inaccurate damage detection. This study proposes a systematic framework that leverages high-precision three-dimensional (3D) laser scanning and Revit model to detect facade anomalies efficiently and quantitatively. Using a dormitory building in Yibin City, Sichuan Province as a case study, we employed the Z[Formula: see text]F IMAGER 5010C laser scanner to capture dense point cloud data of the building facade. This model was then quantitatively compared with the Revit solid model to establish a spatial information database for facade damage. This database was further calibrated with actual site conditions. The case study demonstrated that using a deviation threshold of [Formula: see text] [Formula: see text]mm as the demarcation point effectively identified hollowing and peeling damage, and other issues on a building facade. A total of 18 bulges on the example building facade were identified via the proposed framework; the detection results allowed for quantification and visualization of the facade issues. The results were integrated into a digital building management platform that supports damage classification, spatial mapping, task scheduling, and closed-loop maintenance tracking. This integrated approach enhances facade safety monitoring by providing a quantifiable, repeatable, and automated detection workflow, providing a scalable solution for digitalized facade condition monitoring in urban environments.

Few-shot learning can find the latent structure information between the support and query data by the similarity metric of meta-learning to construct the discriminative model for recognizing the new categories with the less labeled samples. Most existing methods attempt to model the similarity relationships among samples within meta-tasks for achieving this goal. However, the relationship assessment among samples from distinct meta-tasks is challenging due to the differing metric relationships inherent to each respective meta-task. To address this issue, high-order structure-preserving graph neural network (HOSP-GNN) can deeply explore the rich samples structure of the different meta-tasks to predict the label of the queried data based on the graph. HOSP-GNN can mine high-order structures to enhance their relevance by triple samples. In addition, it can also generate the updating rule of the different-order structures for node and edge representation optimization under manifold constraints. Notably, HOSP-GNN eliminates the need for retraining the learning model to recognize new classes, thanks to its high generalization of high-order structure that ensures model adaptability. The experiments demonstrate that HOSP-GNN outperforms state-of-the-art methods in four benchmark datasets, as well as on a self-built dataset focusing on endangered animals. The available code ishttps://github.com/yangfeifei02/HOSP.

To improve the accuracy of electric vehicle charging station fault prediction, a method based on improved Stacking with multiple model fusion is proposed. Firstly, the operational parameters of the charging station are preprocessed, and the processed data is used as input for the models. Subsequently, multiple algorithms, optimized using particle swarm optimization, are employed to build fault prediction models. The algorithm selection is based on the importance from random forest and Pearson correlation coefficient, considering the importance and predictive correlation of each model. Unnecessary models are eliminated, and the final ensemble of base learners is determined. The outputs generated by the base learner models through 5-fold cross-validation are used as inputs for training the meta-learner in the second layer, establishing the Stacking prediction model. Finally, a real dataset of charging station faults is used for simulation analysis. The simulation results demonstrate that the proposed method performs exceptionally well, achieving a fault prediction accuracy of 96.38%.

The virtual white cane is a range sensing device based on active triangulation, that can measure distances at a rate of 15 measurements/second. A blind person can use this device for sensing the environment, pointing it as if it was a flashlight. Beside measuring distances, this device can detect surface discontinuities, such as the foot of a wall, a step, or a drop-off. This is obtained by analyzing the range data collected as the user swings the device around, tracking planar patches and finding discontinuities. In this paper we briefly describe the range sensing device, and present an online surface tracking algorithm, based on a Jump-Markov model. We show experimental results proving the robustness of the tracking system in real-world conditions.