Change detection (CD) plays a vital role in remote sensing for monitoring urban dynamics and land use evolution. However, existing deep learning-based approaches often suffer from weak semantic differencing, insufficient multi-scale feature consistency, and limited global context modeling. To address these challenges, we propose TriCDNet, a multi-scale tri-stream interaction network for building change detection. Built upon EfficientNet-B5, TriCDNet introduces three complementary feature streams — two bi-temporal features and their normalized difference map. A multi-level Difference-guided Cross-temporal Interaction Module (DCIM) performs symmetric convolutional interaction and gated fusion between these streams to emphasize true structural changes. The aggregated multi-scale features are further refined by a Feature Pyramid Network (FPN) and a lightweight Transformer-based decoder, enabling robust long-range spatial–temporal modeling. Experiments on the LEVIR-CD dataset demonstrate that TriCDNet achieves state-of-the-art performance with an F1-score of 92.01% and IoU of 85.20%, outperforming both CNN- and Transformer-based baselines while providing more accurate and structurally consistent change maps.

High-precision and real-time deformation monitoring of open-pit mine slopes is essential for early warning and management, directly impacting production efficiency and ensuring the safety of life and property. As a relatively new monitoring technology developed in the past two decades, GB-InSAR offers all-weather, real-time, large-scale, non-contact and high-precision observations, and has become one of the key techniques for slope safety monitoring in open-pit mines. Therefore, effectively integrating GB-InSAR deformation data with advanced prediction models to improve early prediction ability has become an important research topic. To this end, this paper proposes a GB-InSAR time-series processing method based on the LSTM model. First, the initial deformation information obtained from IBIS monitoring is used as input to construct an LSTM-based prediction model for short-term slope deformation. By combining the LSTM model with ground-based InSAR data, the time-series characteristics of slope deformation are fully explored and a deformation prediction model and risk warning mechanism are established, providing effective technical support for the safety management of open-pit mine slopes.