为提升内河航行场景下水面小目标的检测效果，降低模型复杂度，本文提出了一种基于改进YOLOv5l的轻量化水面目标检测模型。该模型以YOLOv5l作为基础模型，首先对定位损失函数进行改进，提出了NWD-CIoU函数，提升模型对小目标的召回率，增加模型的多尺度目标检测能力；其次，将FasterBlock模块与C3模块进行结合提出了C3_Faster模块，对YOLOv5l模型的主干网络进行优化，减少网络参数，降低模型复杂度；再次，基于Slimming方法对模型进行剪枝，大幅修剪冗余连接，减少模型参数和运算复杂度，提升推理速度；最后，基于通道知识蒸馏方法，以YOLOv5x作为教师模型对剪枝后模型进行蒸馏，以提升模型的检测效果。实验结果证明，本文所提出模型对内河航行场景下的水面目标具有较好的检测效果和检测速度，且相比于原YOLOv5l模型参数量减少了81.48%，GFLOPs减少了80.69%，更适合计算资源受限的移动设备上部署，具有一定的工程意义。

In order to improve the detection effect of water surface small objects in inland navigation scenarios and reduce the model’s complexity, a lightweight water surface object detection model based on improved YOLOv5l is proposed in this paper. The model takes YOLOv5l as the base model, firstly, the positioning loss function is improved, and the NWD-CIoU function is proposed to improve the recall rate of the model for small objects and increase the multi-scale object detection capability of the model. Secondly, combining FasterBlock module and C3 module, C3_Faster module is proposed to optimize the backbone network of YOLOv5l model, reducing network’s parameters and reducing model’s complexity. Thirdly, the slimming method is used to prune the model, greatly trim the redundant connections, reduce the model’s parameters and operational complexity, and improve the inference speed. Finally, based on the channel knowledge distillation method, the YOLOv5x was used as the teacher model to distill the pruned model to improve the detection effect of the model. The experimental results show that the proposed model has a better detection effect and speed for water surface objects in inland navigation scenarios, and compared with the original YOLOv5l model, the number of parameters decreases by 81.48%, and GFLOPs decreases by 80.69%, which is more suitable for deployment on mobile devices with limited computing resources, and thereby offering practical engineering significance.