Abstract: In order to enhance the level of intelligent monitoring of abnormal floating objects in water source channels of waterworks, a FEW-YOLO intelligent monitoring model of floating objects in water sources of waterworks based on UAV networking is proposed. In this study, the FasterNet Block is introduced to reduce the complexity of the YOLOv11n baseline. Additionally, an Efficient Multi-Scale Attention (EMA) module is incorporated to enhance feature representation capabilities. The WIoUv3 loss function is then applied to optimize the regression performance of the target bounding box, further improving detection accuracy. Based on real-world image data collected from autonomous UAV network-based inspections, Floating Objects, the monitoring dataset is constructed. Then, the model's advancement and practicality are systematically evaluated. In validation experiments conducted in a large water plant and its surrounding typical river sections in the southern region, the FEW-YOLO model achieved an mAP@0.5 of 83.6%, a 5.7% improvement over the original model, with nearly an 11% reduction in the number of parameters. The model outperforms mainstream comparative models in key performance metrics. This research enables the regularized and intelligent monitoring of floating objects in the water source of a large water plant in the southern region, with the potential for easy expansion to other monitoring types. And it demonstrates significant practical engineering application prospects and promotion value.