Abstract: Objectives: With the rapid development of domestic high-resolution remote sensing satellites, satellite imagery has become an important data source for urban natural resource monitoring, urban renewal, and refined urban governance. However, existing change detection methods are still dominated by pixellevel prediction and often suffer from fragmented boundaries, object adhesion, and incomplete change regions when applied to large-scale high-resolution imagery. Moreover, conventional methods generally provide deterministic predictions without explicitly characterizing the reliability of detected changes, which limits their applicability in operational monitoring. Wuhan-1 (Luojia-3 02), a high-resolution and high-precision remote sensing satellite developed by Wuhan University, provides timely sub-meter-resolution imagery and offers new opportunities for intelligent urban surface monitoring. The objective is to develop an intelligent change detection approach for Wuhan-1 imagery that can simultaneously achieve accurate change-object extraction and quantitative reliability assessment, and to evaluate its applicability to large-scale urban monitoring. Methods: A multi-category change detection dataset was first constructed from multi-temporal Wuhan-1 imagery over Wuhan, covering five representative urban change categories: Buildings, transportation facilities, water bodies, cultivated land, and urban green spaces. An uncertainty-aware instance-level change detection (UICD) model was then developed. Unlike conventional pixel-level change detection methods, UICD treats individual change regions as instances and directly predicts their categories and masks through a query-based instance decoder. A hierarchical Siamese encoder and cross-scale deformable attention mechanism are employed to extract multi-level bi-temporal features and enhance change-aware representations. Furthermore, uncertainty is modeled from two complementary perspectives. Evidential learning is introduced to estimate category uncertainty according to the amount of evidence supporting each category prediction, while mask uncertainty is estimated from pixel-wise entropy and overall mask confidence to characterize spatial ambiguity. The two uncertainties are subsequently integrated into an instancelevel reliability metric to quantitatively assess each detected change object. Results: UICD achieved the best F1 scores among the compared methods for all five change categories, with an average F1 score of 76.06%, exceeding the second-best method by 6.88% percentage points while maintaining moderate com putational complexity. Ablation experiments further demonstrated the effectiveness of both uncertainty modeling and instance-level decoding. Replacing instance-level decoding with conventional pixel-level decoding reduced the F1 scores for building and cultivated-land changes by 3.94% and 2.59%, respectively. Statistical analysis showed that correctly detected instances were predominantly concentrated in the low-uncertainty and high-reliability ranges, whereas false detections exhibited broader uncertainty distributions and a larger proportion of low-reliability predictions. When the comprehensive uncertainty threshold was set between 0.4 and 0.6, filtering unreliable instances improved precision by 1.52% – 6.65% and F1 score by 0.78%–2.17%, demonstrating that the proposed reliability assessment can effectively identify potentially unreliable change objects. In addition, UICD was deployed in the Wuhan Satellite Data and Intelligent Interpretation Public Service System for large-scale real applications. Conclusions: The proposed UICD method combines instance-level change interpretation with uncertainty-aware reliability assessment, improving both the accuracy and usability of high-resolution remote sensing change detection results. Rather than providing only deterministic change maps, UICD assigns quantitative reliability information to individual change objects, facilitating the screening of low-confidence predictions and subsequent manual verification. The experimental and application results demonstrate the potential of Wuhan-1 imagery for large-scale intelligent urban surface change monitoring and provide a practical technical reference for the operational application of domestic high-resolution remote sensing satellite data in natural resource monitoring and refined urban governance.