Occlusion-Aware Camera Selection in Vehicular Networks

Camera sensors are widely used to perceive traffic environments, understand traffic condition, and help avoid accidents. Since most sensors are limited by line-of-sight, the perception data from vehicles can be uploaded and shared through the edge server. To reduce bandwidth, storage and processing cost, we propose an edge-assisted camera selection system that selects only necessary camera images for sharing. The selection process is based on camera metadata which describes each cameras' coverage in terms of locations, orientations, and fields of view. Different from existing work, our metadata-based approach can detect and locate occlusions by leveraging depth sensors, and then precisely and quickly calculate the actual camera coverage and identify the coverage overlap. Based on camera metadata, we study a camera selection problem that aims to select a limited number cameras to maximize total coverage, and solve it with an efficient algorithm. To further reduce bandwidth consumption, we first introduce similarity-based redundancy suppression and sector-based selection techniques. We then propose a Redundancy-Aware Sector Selection algorithm, which incorporates image redundancy into the sector selection process to improve bandwidth efficiency. Extensive evaluations demonstrate that our algorithms can effectively maximize coverage with bandwidth constraint.