To address the issue that various equipment primitives cannot be converted into common data during the secondary use of railway station signal plan layout diagrams,an information extraction method based on improved YOLOv8s is proposed.Firstly,the spatial and channel reconstruction convolutional block(SCConv)is combined with the C2f unit in the neck network to reduce the redundancy of feature maps and repetitive information across spatial and channel dimensions,so as to facilitate a lightweight detection model.Secondly,the efficient multi-scale attention module(EMA)is introduced into the backbone and neck network of the traditional YOLOv8s model to enhance the model's acquisition capability,obtaining contextual information around the equipment in the diagram.Finally,the standard convolution layer in the model is replaced by the attention convolution(RFAConv)with a larger receptive field to solve the feature extraction limitation stem from the sharing of convolution kernel parameters.The results show that the proposed method achieves an average precision of 0.925,a 6.6％higher improvement over the YOLOv8s algorithm;its comprehensive evaluation index value reaches 0.917,and the model weight parameters are reduced by 16％,outperforming most standard models.This method efficiently extracts the layout information of signal equipment from drawings,providing a potent technical solution for automatically generating interlocking data configuration files and improving the efficiency of interlocking software construction.
关键词
信号平面布置图/图像识别/注意力模块/信息提取/YOLOv8s
Key words
Signal layout diagram/Image recognition/Attention mechanism/Information extraction/YOLOv8s
维普
万方数据