This paper presents SBP-YOLO, a lightweight, fast detection framework for real-time speed bumps and potholes on roads, essential for predictive recognition of advanced suspension systems. Building on YOLOv11n, we integrate GhostConv and VoVGSCSPC modules into the backbone and neck to reduce computational complexity and enhance multi-scale semantic features. To improve small object detection, we introduce P2-level branching using the lightweight, high-efficiency detection head LEDH to mitigate computational overhead without compromising accuracy. We further improve localization accuracy and robustness through a hybrid learning strategy that combines NWD loss, backbone-level knowledge distillation, and Albumentations-based augmentation. Experimental results show that SBP-YOLO achieves 87.0% mAP, a 5.8% improvement over the YOLOv11n baseline model. After TensorRT FP16 quantization, it runs at 139.5 FPS on a Jetson AGX Xavier, demonstrating a 12.4% speedup over the P2-enhanced YOLOv11. These results demonstrate the effectiveness of the proposed method for fast and low-latency road condition recognition in embedded suspension control systems.
