The core of developing a “transportation powerhouse” lies in establishing a sound, modernized, and intelligent transportation system. Enhancing the level of intelligence in bridge inspection contributes significantly to accelerating the construction of such a system. At present, a wide range of object detection models based on machine vision exist, yet there is a lack of models specifically designed to address the complex environments, small defect targets, and real-time detection requirements found on the undersides of steel bridge decks. Therefore, this study adopts an object detection algorithm based on YOLO11 enhanced with SimAM-HTB to improve detection accuracy and robustness. Additionally, image preprocessing based on OpenCV is employed to extract defect features. Together, these components form an intelligent inspection algorithm tailored for steel bridge structures.