A Mamba-based bridge detection method for optical remote sensing images

Accurate detection of bridges in remote sensing images is crucial for various applications, including intelligent transportation, urban planning, and post-disaster recovery. However, traditional object detection models face significant challenges when applied to complex remote sensing environments, such as low-resolution imagery and background noise. This paper proposes an improved object detection algorithm, Mamba-YOLO-SimAM, which integrates the state space model Mamba with the YOLO framework and incorporates the SimAM spatial attention mechanism. The state space model enhances the extraction of spatial dependencies, while SimAM effectively focuses on critical regions, suppressing irrelevant background interference. To evaluate the proposed method, extensive experiments were conducted on a self-built remote sensing image dataset. The results demonstrate that Mamba-YOLO-SimAM outperforms several mainstream models, including YOLOv8, YOLOv10, and YOLOv11, achieving a mAP50 of 49.3%, the highest among all tested models. This research provides a practical and efficient solution for bridge detection in remote sensing images, with significant potential for real-world applications.