An improved target detection with vision and millimeter-wave radar fusion for YOLOv5

In order to solve the problem of low target detection accuracy of a single sensor in complex road environments, an improved target detection method based on the fusion of vision and millimetre-wave radar in YOLOv5 is proposed. The millimetre-wave radar point cloud is converted into an image, so that the visual image and millimetre-wave radar are fused at the model input, and the relevant experiments are unfolded on the nuScences dataset to prove that the proposed method then has a better overall performance of target detection compared with the results trained on YOLOv5, YOLOv7^[1], and YOLOv8. The Mean Average Precision(mAP) is improved by 7.42%, 5.9%, and 4.87% over YOLOv5, YOLOv7, and YOLOv8, respectively, and compared with several other fusion algorithms, the method of this paper is more advantageous.