Timely detection of road cracks is crucial for reducing traffic accident risks, extending road lifespan, and improving transportation efficiency. This study introduces a novel object detection method, named Shift-Wise Cross-Scale Dynamic Head YOLO (SCD-YOLO), tailored specifically for road crack detection. SCD-YOLO builds upon the YOLOv8n architecture and seamlessly integrates the Shift-Wise Convolutional Feature extraction module (SWC2f), Cross-Scale Attention Fusion Module (CCAFM) for feature fusion, and Dynamic Head (Dyhead) detection module to enhance its structure. Inspired by the Shift-Wise Operator, this study devises and implements the SWC2f module, significantly expanding the receptive field and enhancing feature extraction efficiency through sparse mechanism and shift standard convolution. Furthermore, this study incorporates Channel Fusion Deformable Convolution within the CCAF module, effectively blending multi-scale features and enhancing the model's representation capability. Additionally, the Dyhead framework substantially enhances localization and classification accuracy by integrating multiple self-attention mechanisms, thus yielding more precise results for road crack detection. Moreover, this study employs the latest RDD2022 dataset and conducts extensive data augmentation to ensure the model's generalizability and robustness across diverse road damage scenarios and geographic regions. Experimental findings reveal a 4.1% enhancement in mean average precision (mAP) with only a marginal increase in computational cost for the proposed SCD-YOLO. Compared to current mainstream object detection algorithms, SCD-YOLO demonstrates superior performance, affirming its efficacy in road crack detection. This study not only presents an efficient and accurate method for road crack detection but also offers novel insights and directions for related research fields.