Research On Image Detection And 3D Reconstruction Algorithm Of Four-electric Equipment In Rail Transit

Rail transit is a complex system project.With continuous technological innovation,the demand for intelligent construction processes is increasing.Traditional technologies are no longer able to meet these requirements.To improve the standardization of equipment in rail transit,intelligent technologies such as target detection and 3D reconstruction are introduced,which will help promote the intelligent development of rail transit construction.In the field of rail transit,communication,signaling,power;and electrification equipment are generally referred to as the"Four-electric equipment." This study aims to introduce intelligent technologies such as target detection and 3D reconstruction into the research of four-electric equipment to provide technical support and assistance for the intelligence and digitization of rail transit.The main work of this article is as follows:1.A study on the Multi-Scale Dilated Convolution YOLOv3(MSDC-YOLOv3)algorithm for rail transit four-electric equipment object detection.Firstly,a target detection dataset is established for rail transit four-electric equipment.Secondly,in response to the problem of low detection accuracy caused by different target scales,the MSDC-YOLOv3 algorithm is proposed:different levels of features are enhanced by using dilated convolution and mixed dilated convolution in the YOLOv3 prediction module to achieve feature enhancement under different perspectives.By extracting target feature information of different scales,the model can express features more fully during the prediction process.The use of the SIoU loss function enables more precise shape matching between predicted and true bounding boxes.Finally,different algorithms are trained and tested on the rail transit four-electric equipment target detection dataset,and the experimental results show that the mAP of MSDC-YOLOv3 algorithm reaches 83.18%,and the detection speed reaches 53FPS,which significantly improves the detection accuracy and speed compared to other popular networks of the same type.2.A study on the Pyramid Feature Attention Pixel2Mesh(PFA-Pixel2Mesh)algorithm for rail transit four-electric equipment three-dimensional reconstruction.A three-dimensional reconstruction dataset is created by selecting three-dimensional models that meet the requirements of rail transit four-electric equipment from the ShapeNet database.Based on the Pixel2Mesh network,the pyramid feature attention mechanism is introduced into the image feature extraction network to extract high-level features at multiple scales using the pyramid feature module.Different levels of features are fused by using channel attention and spatial attention,which helps to obtain the target features of the image.Then,the perception feature pooling layer of the connection module is connected to the reconstruction network for cascaded deformation network.Finally,the optimization algorithm proposed in this paper is trained and tested based on the rail transit four-electric equipment three-dimensional reconstruction dataset.The experimental results show that the F-score and CD evaluation indicators of the proposed algorithm are better than those of currently popular mesh-based reconstruction networksIn summary,in the field of intelligent construction of the "four electrical equipment" in rail transit,the optimized target recognition algorithm proposed in this paper significantly improves the recognition accuracy compared to currently popular recognition algorithms,and the improved 3D reconstruction algorithm exhibits a significant improvement in reconstruction performance compared to the current mesh-based reconstruction algorithms.Constructing an intelligent management system for rail transit" Four-electric equipment" based on the algorithms proposed in this paper can effectively promote the intelligent construction of rail transit.