Fusion of Multiple Sensors and V2V Information for 3D Object Detection

Current vehicle road 3D target detection techniques mainly rely on information from vehicle on-board sensors (e.g., LiDAR, RGB camera, etc.) for unimodal detection or fusion of multimodal sensing data for detection and have not yet considered information obtained from the internet of vehicle networks, such as V2V messages. In this paper, we consider integrating local onboard sensing data and fusing vehicle information obtained from vehicle-to-vehicle (V2V) communication, which has been shown to significantly improve the performance of 3D target detection on vehicular roads. Specifically, V2V messages are incorporated into multimodal features to improve object recognition accuracy. Unlike previous camera and LiDAR fusion only, our approach explores the fusion of V2V data into a fusion network. We have conducted extensive experiments, and the results demonstrate the utility of V2V messages and can provide useful inspiration for future 3D target detection on vehicular roads.