Re-DETR: Research on Fast Detection Technology for Railway Engineering Targets in the Dark Time Domain

Fast detection of railway engineering targets under low light conditions has always been a challenging problem. Traditional target detection algorithms are limited by lighting conditions, resulting in a decrease in target visibility, which in turn affects detection accuracy. To address this issue, this study proposes a new target detection network for low-light environments (Re-DETR) that enhances the model’s detection capability for targets under low light conditions by integrating an optimized RetinexNet image enhancement network and an improved transformer for the image recognition strategy (DETR). Re-DETR uses RetinexNet for image enhancement to improve image quality and visibility and then inputs the enhanced images into the DETR algorithm with an added global channel attention (GCA) module for target detection. The experimental results show that our method can quickly and accurately detect railway engineering targets in the dark time domain, which has significant advantages over traditional methods.