To address the issue of increasing tunnel surface defects due to prolonged service periods. This paper proposes an online recognition method using the YOLOv10 object detection algorithm for identifying tunnel surface conditions and intrinsic textures. Initially, tunnel inspection equipment was designed and deployed on a field operation inspection vehicle to facilitate the rapid acquisition of high-definition images of the tunnel surfaces. The collected images were meticulously annotated to highlight intrinsic textures such as cracks, water leaks, pipelines, bolt holes, and distribution boxes, resulting in 5,813 authentic tunnel defect images serving as the datasets for this study. Utilizing the YOLOv10 object detection model, the developed model achieved an accuracy of 80.6%, a recall rate of 77.8%, and an mAP0.5 of 81.3% on the experimental datasets. The model size is just 5.46 MB, and it achieves an FPS of 64.32, which is sufficient to meet the requirements for online detection and recognition tasks of tunnel surface defects.