Fusion of Image and Pointcloud Data for 3D Object Detection with 4D Imaging Radar

3D object detection aims to extract the spatial positions of objects from sensor data, thereby improving environmental awareness and comprehension. With the advancement of radar technology, the 4D imaging radar offers greater robustness in adverse weather conditions compared to LiDAR sensors, thus expanding its potential applications. However, due to the relative scarcity of 4D imaging radar datasets and the inherent sparsity of the radar point clouds, traditional point cloud-based 3D detection models perform suboptimally when applied directly. This paper focuses on the 3D object detection task with multimodal sensors, exploring the integration of RGB image features and 4D imaging radar point cloud features. We propose an innovative multi-modal detection fusion algorithm that markedly enhances detection accuracy on the VoD dataset.