The paper introduces IDP-YOLOV9, an enhanced object detection model designed for power construction sites under severe weather conditions. The model integrates a parallel architecture comprising the Image Dehazing and Enhancement Processing (IDP) module and an improved YOLOV9 object detection module. The IDP module, which includes the Three-Weather Removal Algorithm (TRA) and the Deep Learning-based Image Enhancement (DLIE) module, filters multiple weather factors to enhance image quality. The improved YOLOV9 detection network incorporates a three-layer routing attention mechanism to capture features from restored clear images. Experiments demonstrate that the IDP module significantly improves image quality and that the improved YOLOV9 model achieves a 6.8% increase in mean average precision (mAP50) compared to traditional single-processing models on complex weather datasets. The paper also discusses the design and implementation details of the proposed method, including the TRA and DLIE modules, the improved YOLOV9 detection module, and the loss function. The experimental results show that IDP-YOLOV9 outperforms existing methods in defogging, deraining, and desnowing, and achieves higher object detection accuracy under various adverse weather conditions.The paper introduces IDP-YOLOV9, an enhanced object detection model designed for power construction sites under severe weather conditions. The model integrates a parallel architecture comprising the Image Dehazing and Enhancement Processing (IDP) module and an improved YOLOV9 object detection module. The IDP module, which includes the Three-Weather Removal Algorithm (TRA) and the Deep Learning-based Image Enhancement (DLIE) module, filters multiple weather factors to enhance image quality. The improved YOLOV9 detection network incorporates a three-layer routing attention mechanism to capture features from restored clear images. Experiments demonstrate that the IDP module significantly improves image quality and that the improved YOLOV9 model achieves a 6.8% increase in mean average precision (mAP50) compared to traditional single-processing models on complex weather datasets. The paper also discusses the design and implementation details of the proposed method, including the TRA and DLIE modules, the improved YOLOV9 detection module, and the loss function. The experimental results show that IDP-YOLOV9 outperforms existing methods in defogging, deraining, and desnowing, and achieves higher object detection accuracy under various adverse weather conditions.