To address the problems of low accuracy， high rates of missed and false detections， and poor real-time performance of existing road defect detection methods in complex road backgrounds， this paper proposes a lightweight road defect detection algorithm named YOLO11-AKAD based on YOLOv11n. Firstly， a C3k2_AKConv module based on adaptive kernel feature extraction is designed in the backbone network of YOLOv11n to enhance the model''s local feature extraction capability. Secondly， the ADown module is introduced to optimize the downsampling process， improve the transmission of multi-scale contextual information， and reduce the number of model parameters. Finally， an adaptive image scaling method is adopted in the input preprocessing stage to solve the problem of excessive redundant information caused by the original scaling method. Experimental results on the China_RDD dataset show that the mAP@0.5 of YOLO11-AKAD reaches 81.6%， with the number of parameters and floating-point operations reduced to 2.09 M and 5.1 G FLOPs， respectively. Compared with the baseline model YOLOv11n， the mAP@0.5 is increased by 1.8 percentage points， and the model parameter scale is reduced by 19%. The results demonstrate that the proposed method can effectively improve the performance of road defect detection in complex scenarios.