A fast object detection algorithm based on millimeter-wave radar and camera fusion

In autonomous driving operations, acquiring information about the positions and categories of objects is crucial. Visual detection involves convolutional operations on images. As the input image size increases, the inference time of neural networks rises significantly. While reducing the input image size can improve detection speed, it also notably compromises detection accuracy. This issue is particularly pronounced on low-computational-power platforms. To address this, this paper proposes a novel rapid object detection algorithm, RGD-YOLO, designed to remove non-target areas from the image by guiding the process with millimeter waves. This approach enhances the proportion of large targets in the input image, achieving good detection accuracy while reducing the image size. When the mean average precision (mAP) is 0.515, compared to YOLOv5s, this algorithm reduces inference time by 33.18% and increases video processing frame rates by 49.57%.