The task of detecting objects in images affected by blurring presents a crucial and difficult problem in the area of object detection. The presence of blurred edges and noise within such images often complicates accurate object localization and identification, making it an issue that demands effective solutions. This study proposes an optimized object detection algorithm that builds upon YOLOv11s, specifically designed to improve performance in detecting objects in blurred images. To tackle feature insufficiency during down sampling and support multi-scale feature extraction, we replace the traditional convolution module with the ADown convolution module. This modification reduces the parameter count, improves model efficiency, and preserves more image information. Additionally, to facilitate the model in focusing more precisely on the blurred regions of images, we integrate the SimAM module to C2PSA. This incorporation significantly improves the precision and reliability of object detection, and at the same time, enhances the computational efficiency. To further optimize model performance, we incorporate the Powerful-IOU loss function, which accords more weightage to arduous samples, encouraging the model to perform better in such cases. Findings from experiments using publicly shared datasets prove that the presented DAP-YOLOv11s model achieves significant improvements across various evaluation metrics. On the vehiclesCounting and NEU-DET datasets, DAP-YOLOv11s performed better than YOLOv11s, with significant improvements in recall, mAP50, and mAP50-95 while reducing the number of parameters.