Freight train number real-time identification system based on convolutional neural network under big data

Abstract

The invention discloses a freight train number real-time identification system based on a convolutional neural network under big data. The system comprises the following steps of collecting a lot of freight train images and constructing a train number data set; designing a convolutional neural network model structure of train number identification and parallelizing a training algorithm; on a big data platform, training the convolutional neural network; and using a big data real-time processing technology to process the collected images, after train type and train number area positioning and train type and train number character segmentation, using a trained model to carry out identification, and outputting an identification result. By using a freight train number real-time identification method based on the convolutional neural network under a big data environment, a big data technology, a deep learning technology and an image processing technology are combined, training time of the convolutional neural network is shortened, freight train number identification accuracy is increased and a freight train number can be identified in real time.

Description

technical field [0001] The invention relates to the field of car number recognition, in particular to a real-time recognition system for freight train car numbers based on a convolutional neural network under big data. Background technique [0002] Railway freight transportation is one of the main modes of modern freight transportation, and the train number of a freight train, as the unique identification of each carriage, plays a vital role in the railway freight process: in planning, scheduling, settlement and contract management In the whole tracking, statistics and analysis of transportation and sales, without exception, it needs to be associated with the train number. It is particularly important to study the vehicle number recognition technology. [0003] The existing car number recognition technology generally adopts video recognition for car number recognition, that is, relying on the train image acquisition device to obtain the train logo, and then identifying the ...

AI Technical Summary

Problems solved by technology

But this method has the following disadvantages: (1) Although the real-time identification of freight train numbers belongs to weak real-time performance, it is still difficult to achieve better results in the real-time identification of vehicle numbers due to the lack of massive video storage technology and real-time processing technology; (2) Due to the lack of a better freight train number recognition model, the accuracy of car number recognition is difficult to achieve the desired effect; (3) When the data set is large, the recognition model training time takes a long time

[0005] At present, there is no car number recognition method based on big data deep learning. It is an urgent need to establish an efficient and accurate car number recognition method for existing freight train car number recognition research.

Images