Research On Video-Based Analysis Method Of Train Cargo Car Information And Status

Railroad freight is an indispensable part of national economic development,with its advantages of large capacity,high efficiency and low cost,which can significantly improve logistics efficiency,reduce logistics costs and create profit margins for enterprises.The train car number is the unique identification of railroad freight trains,which plays a key role in the management of train information.The traditional method of obtaining wagon car numbers is to manually transcribe the car numbers by the car number clerk of the station preparation group.In addition,companies need to monitor and analyze the number of train cars and operating speed to better grasp the logistics transportation situation,timely adjust and develop logistics plans to ensure the accuracy and timeliness of industrial logistics transportation.However,manual statistical methods are intensive,inefficient and prone to errors.In view of the above reasons,this paper analyzes and studies the information and status of train cargo cars based on video.Through image processing technology,it intelligently identifies train carriage numbers and statistically analyzes status information such as their number and travel speed to achieve real-time monitoring of carriage operation conditions and improve the automation of logistics transportation.In this paper,the recognition of carriage characters is divided into two parts:character region localization and character recognition.The Transformer self-attention mechanism is introduced on the basis of YOLOv5 s model to improve the accuracy of the model for character region localization.The image data is enhanced before training to simulate industrial site interference information while enriching the samples to improve the generalization ability and robustness of the model.After the accurate localization of the character region of the carriage is completed,for the carriage itself at the industrial site also has character defacement,unclear handwriting,motion blur and character recognition bit error,this paper adds an attention network between CNN and Bi LSTM networks on the basis of the original CRNN model to enhance the sensitivity of the network to the characters in the character region to further improve the expressiveness of the model.After the character recognition of carriages is completed,the number,speed,and direction of travel of the moving carriages are analyzed,and the Shuffle Net V2 network is used to replace the original Deep Sort backbone network,which significantly reduces the number of parameters of the model,and the two-way counting of carriages is tested for the improved Deep Sort and the original model respectively.The results show that the improved algorithm can still ensure the accuracy of the two-way counting of carriages with a significant reduction in the calculation of network parameters,which improves the efficiency of the actual deployment of the model in industrial sites.The improved three-part model of YOLOv5 s,CRNN,and Deep Sort is fused and the interface is written and designed using Py Qt5.The test results show that the detection rate can reach about 15 fps when the presence of characters in a specific region is detected.When no character region is detected,the video processing frame rate is 30 fps,which meets the need of real-time detection of carriages in industrial sites.