OmniView: Region-of-Interest Aware View Selection for Point Cloud Coverage Augmentation in Connected Vehicle Networks

This paper explores point cloud coverage augmentation to address physical occlusion in bandwidth-limited vehicle-to-vehicle wireless networks. Traditional approaches involve sharing 3D point cloud data between vehicles and performing point cloud registration to merge their views for a wider perspective. However, current solutions assume unlimited bandwidth and transmit the entire point cloud for view merging, posing challenges in bandwidth requirements and computational burdens for time-critical autonomous driving applications. To address these issues, we propose a novel Region-of-Interest (ROI) aware method called OmniView, which can significantly reduce the required point cloud size for transmission while maximizing the point cloud coverage to create a “see-through” effect for bandwidth-limited connected vehicles. Most importantly, our solution ensures high object detection accuracy in obstructed areas with only polynomial complexity. Simulation results demonstrate its superior performance in terms of efficiency and accuracy compared to existing work. Overall, our study provides valuable insights into point cloud coverage augmentation for connected vehicles, making substantial progress toward the practical implementation of point cloud data merging in real-world scenarios.