Abstract: To address the issues of insufficient utilization of directional features, rigid multi-scale fusion, and imbalanced local-global information in existing crack detection methods under complex environments, this study proposes a ground fissure detection method based on dynamic collaborative adaptive directional fusion decoder (Dyn-Co ADFD). ResNet-50 is adopted as the encoder for multi-scale feature extraction, and a global context enhancement (GCE) module is incorporated to enhance fissure responses and suppress background noise. Furthermore, a Dyn-Co ADFD is designed, which realizes direction-aware feature reconstruction and multi-scale information fusion through dynamic directional feature decomposition, two-dimensional adaptive weight calculation, and an integrated fusion-smoothing collaboration mechanism. Experiments conducted on the ground fissure dataset of mining areas in Shanxi Province demonstrate that Dyn-Co ADFD achieves 85.2%, 88.7%, 85.8%, and 87.9% in terms of intersection over union (IoU), F1-score, structural similarity index measure(SSIM), and edge accuracy (EA), respectively, significantly outperforming models such as U-Net. Ablation experiments further validate the effectiveness and collaborative advantages of each core module. This study provides a solution for fissure detection in complex scenarios, and exhibits important application value for the intelligent monitoring of geological hazards and safety assessment of engineering.