The solution to the problem of automated recognition of the parameters of rolling stock units in real time is considered. A method for solving the problem of recognizing the main identification marks of a railway carriage in real time is selected. A specific example of the implementation of the method proposed in the article is given. The results of the developed program are presented. The application of the developed system will allow to completely solve the problem of automated recognition of the parameters of rolling stock units, and save money by avoiding downtime, delays in return of wagons and conducting reasoned work with contractors. identification of standard signs of freight railway wagons, effective recognition methods, neural networks Models and Methods for Researching Information System
Today, an important aspect of the efficiency
of using railway transport and monitoring its
condition is the identification of rolling stock
(RS) cars by their inventory number. This is due
to the fact that a unique inventory number
assigned to a railway car can determine its main
characteristics and provides code protection for
reliable reading of the railway car number [
metrology services, security services, logistics departments and railway shops.
The main problem of optimizing production processes at the station is that the process of monitoring the movement of objects of the rolling stock of railway transport, including their identification, is not automated at the processing station.
Today, hundreds of employees are involved in the control of rolling stock, who ensure the appropriate production and technological processes. Considering the fact that the survey of the cars is carried out manually, the final operations with the car take a lot of time and this process is economically expensive.
2020 Copyright for this paper by its authors. Use permitted under Creative the software package is to recognize not only the car number, but also such parameters of railway rolling stock units as: boiler calibration sign, administration code, etc. The article presents a solution to the problem of automated recognition of the main identification marks of railway RS units in real time based on the use of neural networks.
An automated system for recognizing the parameters of rolling stock units must create 3 images in three projections of each passing car (on both sides and on top of the car); and 2 photos of the beginning and end of the car.
The problem statement can be presents as
follows: an automated system for recognizing the
parameters of rolling stock units, consisting of the
next components: a frame with cameras, a server
with a recognition program, and an automated
workstation (AWS) of the user. The work is
carried out as follows: when the train passes
through the installation with cameras, each car is
photographed from several angles, then all
images are transmitted in real time to the server,
where the images are stored and processed
(search and identification of identification marks
applied to the railway carriage). After the train
passes through the installation with cameras and
the processing of all received images is
completed, the information about the train is
transmitted to the user AWS. A general view of
the system operation process is shown in Figure
1. The model “Automated system for recognition
of rolling stock units” is built on the basis of the
objects of the use case diagram.
machine translation, chat bots and task bots
programs that can replace, for example, ticket
sellers [
In addition, neural networks are used in the
analysis of scientific data, games, in finance to
assess customers and risks of various types, and
in many other areas. As a result, neural networks
work well with the data, in which there is a
correlation of the measurements (images, sound,
text, time series) [
In view of the above, as a method for solving the problem, a solution using neural networks was chosen.
In the process of modeling and building a neural network, it is important to consider the quality of the source data. Stencils applied to the railway carriage, have a very large variety of colors, fonts, size and location. This is due to the fact that the place where the stencil is applied is not fixed and depends on the model of the tank, and the boilers of the tanks are always painted in a different color. While tanks boilers intended for the second class of goods, colored in light gray (silver) colors, coppers tanks for transportation of methanol - highly toxic flammable liquid of the third class are colored yellow. In addition, you can find railway carriage painted in blue, green, orange and other colors.
In addition, when recognizing the identification marks applied to the railway carriage, the classical analytical algorithms detector finds in the frame any character sets similar to numbers and tries to recognize them.
To improve the accuracy of recognition of identification marks of freight cars, it is necessary to solve such a subproblem as localization of objects.
The current project is based on the MaskRCNN-based Nomeroff Net project. Nomeroff Net is system for recognizing the licence plate number of a car, written in Python and open source. The entire project recognition of the parameters of rolling stock units consists of three blocks: two main and one additional.
The main blocks are neural network detectors. The first detector is designed to localize objects (search for their areas in the image), the second for their identification (recognition of numbers). The third, auxiliary unit performs the function of monitoring the correctness of the railway car number recognition.
A variety of labels on railway tanks prevents determine the necessary parameters correctly. In such cases it is necessary to apply the detector is designed to localize objects. The objects localizer successfully works where objects of the required type "overlap" each other in the frame or there is a large amount of visual interference. Finding areas correctly will allow you to exclude other labels. After training, the detector will filter out unnecessary "noise" and due to this, the accuracy of parameter recognition will significantly increase.
To solve the subtask described above, a
neural network localizer was designed on the
Mask-RCNN architecture. This architecture has
several advantages: recognizes objects in the
entire image, efficiently consumes computing
resources [
To build a localization model for the railway car number, in the project added to the class AreaSearchNetwork. In the __init__ block, constants and a detector with a default configuration file are initialized, directories are being configured. The "work" method was written to work with the localizer: the input is an image; the output is an array of images, each of which is the parameter of a railway carriage (number, administration code, boiler calibration sign, etc.).
After localization, each parameter is identified using a simple neural network detector, which is trained to search and recognize numbers. It works like this: the input comes from the image area, which has a parameter; at the output - the number of the recognized parameter.
After the railway train passes through the installation with cameras and the processing of all the images obtained, information about the train is transferred to the user AWS. The subsystem "Program for recognizing the parameters of rolling stock units" is shown in Figure 2. Class diagram objects and use case diagram objects are used for display.
An important property of neural networks is their ability to learn from environmental data and, as a result of training, increase their performance. Productivity increases over time according to certain rules.
A short neural network training algorithm is as follows: 1. Data preparation - includes splitting all images that will be used for training into test, validation and training sets matches the real number of the car.
The initial image of the car arrives at the entrance to the program for recognizing the identification parameters of rolling stock units (fig. 5, a). After the work of the localizer, the areas in which the parameters of a railway carriage are located are highlighted (fig. 5, b). Further there is identification of parameters and forming a data block with information about the freight railway wagon (fig. 5, c). The correctness of the recognized railway car number is checked by the checksum. After similar processing of all received images, the collected information about the rolling stock is transferred to the user AWS.
The article discusses an automated system for recognizing the parameters of a railway carriage (railway tanks). The proposed solution is universal and does not depend on the recognized objects. It has a rational structure that ensures efficient use of computing power.
As part of solving the problem, the possibility of using neural networks for solving problems of finding and recognizing given objects (signatures) with the required accuracy is shown. Application of this solution allows automating the process of monitoring rolling stock at the processing station, which reduces the time of final operations with the freight railway wagons; creates conditions for optimizing the time spent on processing the wagons, as well as the emergence of material savings due to the optimization of production processes at the station.