基于Faster R-CNN的铁路扣件状态检测研究

Research on Railway Fastener Status Detection Using Faster R-CNN

针对铁路扣件缺陷种类繁多、检测困难等问题,提出一种基于目标检测的扣件状态检测方法。对扣件图像数据集进行构建并标注;通过搭建深度学习环境,将数据集输入Faster R-CNN模型进行训练和识别;选择合适的评价模型和评估标准进行结果分析。结果表明:该方法对铁路扣件的检测准确率达到97.3%,检测精度和检测效率明显提高,可有效实现对扣件图像的识别和分类。

In view of the difficulties on detecting the defects of railway fasteners and the wide varieties of defects,this paper introduces a state detection method of fasteners based on object detection.First,the fastener image data set is constructed and labeled;then,the data set is input into the Faster R-CNN model for training and recognition by building a deep learning environment;finally,the appropriate evaluation model and evaluation criteria are selected for result analysis.The results show that the detection accuracy of this method can reach to 97.3%,with which the detection accuracy and efficiency are improved obviously and the recognition and classification of fastener images can be realized effectively.