Extraction of Switch Information from Railway Signal Layout Drawing Based on Deep Learning

Aiming at the fact that the information of the signal layout plan of railway stations cannot directly convert the equipment information into general information during the secondary utilization process, an information extraction method based on deep learning is proposed. Firstly, the multi-scale features of the image are extracted through the ResNet50 convolutional layer and the residual block structure. Secondly, a three-level character segmentation method is introduced into the model backbone of the traditional Faster R-CNN to enhance the model's ability to solve the fuzzy adhesion between character boundaries. Finally, K-Means++ and Faster R-CNN combined with non-maximum Suppression (MNMS) are used to solve the problem of inaccurate recognition and positioning. It has been verified that the average accuracy of this method can reach 89.8% on average, which is 10% higher than that of the unimproved model, and it is superior to other models in various indicators. This method can not only accurately and efficiently identify each turnout in the drawing and the relationships between turnouts, but also present the data in the form of a table for secondary utilization.