World Journal of Engineering Research and Technology

Mohammed Sharfuddin*

ABSTRACT

High-definition (HD) maps are a cornerstone for safe and efficient autonomous vehicle (AV) navigation, offering centimeter-level accuracy for road features. However, frequent changes in real-world environments necessitate constant updates to maintain reliability. This paper presents a deep learning-based framework that leverages multi-modal sensor fusion and temporal data comparison to detect and classify environmental changes. The proposed system integratesLiDAR, camera, and GPS/IMU data, and uses a Siamese U-Net architecture to generate change masks. Experiments conducted on nuScenes and KITTI datasets show high accuracy in change detection and potential for real-time application. A map update module integrates validated changes into HD maps with minimal human oversight.

[Full Text Article] [Download Certificate]