The precise detection of road diseases, which is the core of modern road maintenance, is of great Significance for ensuring the safe operation of road infrastructure. However, existing algorithms have problems such as inaccurate positioning, missed detection, and false detection in scenarios such as low-contrast textures, small-sized defects, and complex background interference. To address the aforementioned problems, this study proposes a road disease detection algorithm, ECD-YOLO, which is based on multiscale feature fusion and an extended receptive field. First, the multiscale feature fusion module (EMSC) is employed to fully integrate features of different scales, thereby enhancing the perception of texture features. Second, a compound receptive field enhancement module (CRFB) is introduced to expand the perception of context information through dilated convolution and enhance the recognition ability of small targets. Finally, by utilizing the Dynamic Task Collaborative Detection Head (DTAD), task decomposition alignment is achieved through shared convolution and dynamic feature selection, thereby effectively reducing the impact of complex backgrounds on detection accuracy. Tests were conducted on the RDD2022, MEdDatasetV4, and vq840 datasets. The experimental results show that the algorithm proposed in this study improves the mAP metric by 2.5%, 1.8%, and 1.4% compared to the benchmark model. The results show that the proposed algorithm has significant advantages for road disease detection.