MH-CDNet: Map- and History-Aided Change Detection of Traffic Signs in High-Definition Maps

Recent high-definition maps play a key role in autonomous driving, providing precise environmental information for navigation and decision-making. Current approaches typically focus on either the online map construction or the timely updating of existing maps to keep them aligned with the real-world environment. However, the first approach is limited by insufficient map accuracy, while the second lacks emphasis on map currentness, especially for map change detection. Thus, we propose a novel method that integrates spatial-temporal information from both prior maps and revisit records to enable timely and accurate high-definition map change detection. The experimental results demonstrate that our method significantly outperforms traditional approaches for this task, achieving a 35% improvement in accuracy on real-world data. To the best of our knowledge, this is the first method for high-definition map change detection that combines both map-aided and history-aided components. It significantly enhances the model's ability to discern map changes in complex environments and partially addressing the limitations of previous approaches.