【摘要】 针对室外调车信号数据采集困难、数据集场景覆盖度低、目标较小等问题，提出一种基于WGWS与改进YOLOv8的铁路调车信号室外设备状态智能检测方法。首先，通过数据采集、数据预处理、图像合成与数据标注等操作，制作调车信号室外设备状态智能检测自定义数据集。之后，通过基于二阶段学习的WGWS模型对不良天气图像进行增强，提高图像的清晰度；其次，在基础YOLOv8模型中引入空间深度转换卷积模块和K折交叉验证法，提高模型在低分辨率下对小目标的检测精度。最终，通过在自定义数据集上进行对比实验，结果表明：相较于基础的YOLOv5、YOLOv8和YOLOv11模型，所改进的YOLOv8模型对蓝色和白色调车信号机的平均检测精度由77.5%提升到95.7%；对道岔位置的平均检测精度由45.3%上升至65%；图像增强后的检测准确率更高。更多还原

【Abstract】 To address the problems of difficult data collection for outdoor shunting signals, low scenario coverage of datasets, and small targets, an intelligent detection method for the status of outdoor equipment of railway shunting signals based on WGWS and improved YOLOv8 was proposed. Firstly, a custom dataset for the intelligent detection of the status of outdoor equipment of railway shunting signals was constructed through operations such as data collection, data preprocessing, image synthesis, and data annotation. After that, the WGWS model based on two-stage learning was employed to enhance the images under adverse weather, improving image clarity. Secondly, a spatial depth conversion convolution module and a K-fold cross-validation method were introduced into the basic YOLOv8 model to improve the detection accuracy of the model for small targets under low resolution. Finally, comparative experiments were conducted on the custom dataset, and the results show that compared with the basic YOLOv5, YOLOv8, and YOLOv11 models, the average detection accuracy of the improved YOLOv8 model for blue and white shunting signals increases from 77.5% to 95.7%; the average detection accuracy for switch positions rises from 45.3% to 65%; the detection accuracy is higher after image enhancement.更多还原