RU133082U1 - IDENTIFICATION OF MOBILE RAILWAY UNITS - Google Patents

Abstract

A device for identifying mobile units of a railway transport, including a unit for reading data from a moving unit, a unit for transmitting read data, a unit for processing read data, a unit for recognizing read data and a unit for transmitting and long-term data storage, characterized in that the unit for reading data from a moving unit (1 ) is connected to the read data transmission unit (5) and includes a photoelectric object presence sensor (DNO 2) in the inspection area and at least two cameras (3 and 4) connected to it, the read data transfer is connected (5) with the read data processing unit (10) and consists of an object identification information input unit (BVIN 6), an object body information input unit (BVIK 7) and a process information input unit (BVTI 8), the latter is connected to the automated system for the operational management of transportation by means of (Intraet ASOUP 9), the read data processing unit (10) is connected to the read data recognition unit (15) and includes a preliminary processing unit for information on the vehicle’s flight number (BPS) 11), a block for isolating the contours of each of the symbols of the side number of the object (BVN 12), a block for preliminary processing of information about the body of the object (BPOK 13) and a block for selecting the contour of the body of the object (BVK 14), made with the possibility of ensuring the selection and implementation of a certain information processing algorithm , the read data recognition node (15) is connected to the classification unit of types of moving units (20) and includes a block of a three-layer artificial neural network for identifying board number symbols (ANN1 16), a block of a three-layer artificial neuron

Description

The utility model relates to devices for the automatic identification of mobile units of railway transport (hereinafter referred to as the object) such as trains, wagons, gondola cars, tanks, etc. in the process of movement when they pass sections of railways.

The inventive identification device is intended for use at railway facilities, at enterprises using railway transport, within the production process of which reliable information is needed about the movement of objects, their arrival and departure.

The use of identification devices with the possibility of recognizing wagon numbers is necessary to solve problems that are relevant for railway enterprises, since rolling stock numbers contain the most important information for domestic and interstate transportation processes. Wagon numbers allow monitoring the progress of goods, searching for “lost” transport units and represent a common means of communication between enterprises.

The state-of-the-art accounting and control system (RF patent for invention 2317909, publ. 02.27.2008) containing control and accounting stations and identifiers at controlled sites is known. Each control and accounting station contains a reader, the output of which through the controller is connected to indicators and information input devices, and each identifier contains an antenna connected to a signal receiver, the output of which via the input-output interface is connected to one input / output bus of the microprocessor, other information buses which are connected respectively with random access memory, read only memory and reprogrammable memory. Vibration transducer of vibrations into an electric signal, a rectifier and a filter are introduced into each identifier, the inputs of the rectifier and filter are connected to the output of the vibration converter, the output of the filter is connected to the enable input of the microprocessor, and the output of the rectifier is connected to the identifier power bus.

The disadvantage of this technical solution is the need to additionally provide each mobile unit with a special sensor (identifier), which is permanently attached to the vehicle body, for example, to the carriage frame, and is maintenance free. Since international transportation is currently actively developing, and each country is implementing its own systems for identifying mobile units, there is a serious problem of providing such sensors to the entire international fleet of wagons. In addition, as claimed by the inventors, the sensors are permanently attached to the vehicle body and are maintenance-free, which creates an additional problem for the operation of the proposed system, since if such a sensor fails, it cannot be replaced at the point where a malfunction was detected. In order to eliminate the breakdown of the sensor, the entire mobile unit must be sent to the depot for repair, which significantly increases the turnaround time of the mobile unit, and makes it impossible to urgently deliver the goods to the consignee.

There is also a system of automatic identification for double-track sections of railways (patent for invention RU 2314955, publication date 01/20/2008), consisting of a plurality of coded on-board sensors and at least one reading point, which contains antennas connected to the outputs of the antenna switch the input of which is connected to the first output of the control unit, while the signal input of the antenna switch is connected to the bi-directional output of the circulator, the output of which is connected to the first input of the receiver, the second input and the output of which is connected respectively to the second output and the first input of the control unit, the third output and second input of which are connected respectively to the first input and the first output of the transmitter, the second output of which is connected to the input of the circulator, and the bi-directional output of the control unit is connected to the first bi-directional output of the modem, the second the bi-directional output of which is connected to a data transmission network, characterized in that the system contains four antennas located two on the outer sides of each double-track path road weights connected to the first, second, third and fourth outputs of the antenna switch, and four axis counters, one in front of each antenna on the path closest to it, and connected by their outputs to the third, fourth, fifth and sixth inputs of the control unit, the seventh input and the fourth outputs of which are connected to the first output and the first input of the signal processing unit, the second input of which is connected to the second output of the receiver, and the second output is connected to the input of the modem.

The use of the known system is impossible without the use of identification elements installed on the sides of the cars, in particular, the presence of many code on-board sensors. Due to the fact that part of the cars may not be equipped with such code sensors, either they have become unusable or have been removed from the side of the cars, the process of accounting for carriage is disrupted, and the efficiency of the process of recording freight transportation is sharply reduced.

The closest technical solution is a complex of automated carriage accounting (RF patent for utility model No. 110354, published on November 20, 2011), which contains a system for reading car numbers, a system for processing and transmitting read data, a system for recognizing car numbers, a centralized collection, transmission and long-term storage system data, the automated workstation (AWS) of the administrator, while the reading system for the numbers of the cars contains a reading point made in the form of two supports installed by both sides of the monitored railway track, onto which an optical recording unit equipped with television cameras and a lighting unit, as well as containing a sensor subsystem for positioning a car, are mounted, a system for processing and transmitting read data is connected to a system for reading car numbers, is equipped with a local server and contains them in a container controller, local data transmission module, remote data transmission module, car number recognition system connected to the processing and transmission system the data being read and at each of the reading points it contains a video server, software modules for car counting, video capture, car number recognition, an online video archive and a train of trains, a centralized data collection, transmission and long-term data storage system is a central post containing a long-term data storage, consisting of storage modules, a software module for maintaining databases and a central management server connected to reading systems, recognition systems via Ethernet and nny software module integration with AMS business.

The disadvantage of this system is the complexity and low reliability of reading board numbers, since only the board number read by a video camera is used as a source of information about a mobile unit. The comparison of the identified board number with the full-scale sheet is made at the end of the identification process, which indicates the lack of reservation of information sources at the decision-making stage for identifying the mobile unit. In addition, when inspecting an object, video cameras are used that provide the system with a video data stream. The video stream takes up much more memory space, takes longer to transmit via communication lines, and is more difficult to process by a computer compared to a photo image. Since a standard video image is formed of 25 frames, it is necessary to conduct a video storyboard, select the appropriate frame and extract the board number from it. When in the case of processing photo information, only one frame is needed and there is no storyboard process, which significantly reduces the load on the data line and the computer.

The objective of the claimed utility model is to create a device that allows with a high degree of accuracy and reliability to identify the units of mobile vehicles using remote equipment.

The technical result consists in increasing the accuracy and reliability of identification of objects in conditions of a high level of interference associated with the natural aging of paint and pollution of the identification number of the object; acceleration of the processing of read graphic images of the side number and the body of the object; reduction of technical requirements for the used computing equipment.

The problem is solved by creating a device for identifying mobile units of a railway vehicle, including a unit for reading data from a moving unit, a unit for transmitting read data, a unit for processing read data, a unit for recognizing read data, a unit for classifying types of moving units, and a unit for transmitting and long-term data storage, this node reading data from a moving mobile unit is connected to a node for transmitting read data and includes a photoelectric sensor of the presence of an object in In addition to inspection and at least two cameras connected via communication lines, the read data transmission unit is connected to the read data processing unit and consists of an input unit for information about the object identification number, an input unit for information about the body of the object and a unit for inputting technological information, the latter is connected with an automated system for the operational management of transportation through the Intranet network, the read data processing unit is connected to the read data recognition unit and includes a preliminary unit the processing of information about the side number of the object, the unit for selecting the contours of each of the symbols of the side number of the object, the unit for pre-processing information about the body of the object and the block for selecting the contour of the body of the object, made with the possibility of selecting and implementing a certain information processing algorithm, the recognition unit of the read data is connected with a node for classifying types of moving units, and includes a block of a three-layer artificial neural network for identifying board number symbols, a block of three-layer art a tween neural network for identifying the body of an object, a unit for converting and checking technological information about a moving unit, a memory block for setting synapse weights and information about encoding types of moving units, a unit for classifying a type of moving unit associated with a unit for transmitting and long-term data storage, including a unit for comparing types of moving units units obtained from at least three sources and a control unit configured to synchronize device blocks by supplying activation signals, the transmission unit and and long-term storage of data associated with an automated traffic control system operative comprising a computer unit configured for monitoring functions, control the identification process, the formation, transmission and storage of the results of identification of mobile units reports.

The memory block, hereinafter BP, is intended to store information about the encoding of object types to determine the type based on the identification results. In addition, here the memory unit stores the settings of synapse weights for three-layer artificial neural networks, and the settings of one network are different from the settings of another. During operation of the device, neural networks can learn and, accordingly, adjust the settings of synapse weights.

The control unit, further BU, is designed to synchronize the blocks of the device by supplying activation signals for the blocks. The operation of the control unit is controlled by a computer, then FDA, and by the operator.

The unit for entering information about the identification number of the object, hereinafter BVIN, receives digital information from one of the cameras, designed to read the side number from a moving object.

The input unit for information about the body of an object, hereinafter BVIK, receives digital information from a second camera designed to read the entire body of a moving object. The cameras are driven by a signal from a photoelectric sensor of the presence of an object in the inspection area, then the bottom.

The technological information input unit, hereinafter BVTI, receives electronic information about the composition of the train (train number, number of wagons in the train, side numbers of wagons, etc.) from the automated system for the operational management of transportation, hereinafter ASOUP.

The preprocessing unit for information on the vehicle’s number of the object, hereinafter referred to as BPON, implements binarization of the image — converts the received image into black and white, produces a binary image matrix, filters the image of the airborne number of the object in order to eliminate noise, then performs anisotropic filtering to smooth out Outline character contours and eventually normalizes the image;

The unit for preliminary processing of information about the body of the object, hereinafter BLOCK, carries out binarization of the image similarly to the above; forms a binary image matrix; filtering the image of the body of the object, in order to eliminate noise; further produces anisotropic filtering in order to smooth the body contour; and eventually normalizes the image.

The technological information conversion unit, hereinafter BPTI, converts the read information about the board numbers into a convenient for reading and checking by calculating and comparing the control eighth digit. As a result of the work, an array of eight characters is formed at the output, and on the basis of these data the type of object is determined (wagon, gondola, tank, etc.);

The block for selecting the contours of each of the symbols of the tail number, then the BVN, by the method of chain coding and calculation of numerical coefficients (identification signs) based on the received code using orthogonal exponential functions. For each character of the tail number, a lot of signs are formed that characterize its shape. Since the image may contain extraneous characters that are not part of the tail number, rules have been developed for refusing to highlight the current character and move on to the next. The rules are based on the length of the outline of characters and allow you to consider characters only a certain font size.

The unit for selecting the contour of the body of an object, hereinafter BVK, by the method of chain coding and calculation of numerical coefficients (identification signs) based on the obtained code using orthogonal exponential functions. As a result, at the output, a lot of signs are formed that characterize the shape and dimensions of the body of the object.

The three-layer artificial neural network unit for identification of the tail number symbols, hereinafter ANN1, is intended to identify the symbols of the side number of the object based on the input set of identification signs. As a result of the work, an array of eight characters is formed at the output, and on the basis of these data the type of object is determined (wagon, gondola, tank, etc.);

The three-layer artificial neural network unit for identifying the body of an object, hereinafter ANN2, is designed to identify the body of an object based on an input set of identification features. As a result of the work, at the output we obtain the type of the geometric figure of the body of the object (rectangle, oval, trapezoid) and based on this information the type of object is determined (wagon, gondola, tank, etc.);

The type comparison unit, hereinafter referred to as the BST, is intended to make a decision on the classification of the current facility based on data on the types of cars received from two neural networks and a technical information transfer unit. The decision is made by voting. In the case when all the data do not contradict each other, an unambiguous decision on classification is made. Otherwise, a majority decision is made, and a message is transmitted to the FDA. If all three sources are inconsistent, an error message is sent to the FDA.

The inventive device is shown in the figure and includes a node for reading data from a moving mobile unit 1, including a photoelectric sensor for the presence of an object in the inspection zone 2 and two cameras 3 and 4 connected by a sensor via communication lines. The read data transfer unit 5 consists of a unit for inputting information about the identification number of an object 6, a unit for inputting information about the body of an object 7, and a unit for inputting technological information 8, the latter being connected to an automated system for operational control of transportation 9 via the Intranet network. The read data processing unit 10 includes a pre-processing unit for information on the side number of the object 11 and a contour extraction unit for each of the characters of the side number of the object 12 connected in series to it, a pre-processing unit for information about the body of the object 13 and a sequential block for selecting the contour of the body of the object 14 The read data recognition unit 15 includes a block of a three-layer artificial neural network for identifying characters of the tail number 16, a block of a three-layer artificial neural body identification networks of an object body 17, a conversion and verification unit for technological information about the moving unit 18 and a memory block for setting synapse weights and encoding information for the types of moving units 19. A classification unit for the type of moving unit 20 includes a unit for comparing types of moving units 21 and a control unit 22. Node transmission and long-term storage of data 23 is associated with an automated system for the operational management of transportation 9 and includes a unit of computer 24. Maintenance of the computer is carried out operator 25.

The principle of operation of the device is the quorum reservation of sources of information about the object. The quorum reservation is based on the simultaneous use of three or more independent sources of information about the object — images of the object’s identification number, images of the body of the object and previously entered technological information about the object (train number, number of cars in the train, side numbers of cars) received from the automated system by transportation management ASOUP, through the Intranet network.

Photographs of the identification number and body read from the object and verified technological information about

identification number, allow you to form a clear picture before deciding on the classification of the type of object. The decision on the classification of the type of object is carried out in the type comparison block.

While the train is moving, the presence detector of the DNO 2 object, installed near the railroad track, detects the presence of a moving unit and sends a signal to cameras 3 and 4. The cameras take pictures of the moving unit once, and transmit images through the communication line to the BVIN 6 and BVIK 7. At the same time, technological information, for example, a full-size sheet of a train of the DU-1 form, is transferred to the BVTI block from ASOUP 9 via the Intranet railway network. The BVIN block 6 transmits the image to the BPU 11, the BVIK block 7 transmits the image to the BPOK 13, the BVTI block 8 transmits the full-scale sheet of the train to the BPTI 18.

The image of the tail number is processed by the encoding procedure in the BPOU block 11. As a result of processing, a binary image matrix is formed. In order to improve quality, filtering procedures are applied to the binary matrix. As a result of processing, noise is eliminated and the outlines of the tail number symbols are smoothed out. The matrix is transferred to the block BVN 12, where, based on the data from the matrix, the outlines of the flight number symbols are extracted and, based on the data on the curvature of the outlines, numerical coefficients are calculated. As a result, a two-dimensional array of size [1 ... 8; 1 ... 32], storing numerical coefficients. The array has eight columns for eight characters of the tail number, thirty-two rows each for storing the numerical coefficients of each character. Next, the array is transmitted to the input ANN1 16.

The image of the body is processed by the encoding procedure in the block BPOK 13. As a result of processing, a binary image matrix is formed. In order to improve quality, a filtering procedure is applied to the binary matrix. As a result of processing, noise is eliminated and the contour of the body is smoothed out. The matrix is transmitted to the BVK block 14, in which the body contour is selected, and the dimensions of the object (length and height) are determined. Based on body curvature data, numerical coefficients are calculated. The calculation results are recorded in a data array of size [1 ... 3; 1 ... 32]. The first column contains thirty-two numerical coefficients, cell [2; 1] contains the length of the body, and the cell [3; 1] - body height. Next, the array is transmitted to the input of ANN2 17.

From the full-scale sheet of the train in block BPTI 18, the data on the side numbers of cars in the current train are read. In the process of reading, the control eighth character is checked for each side number. Mistakenly filled data is marked with a label. As a result, the output is a two-dimensional array of data Nlp [1 ... 2; 1 ... N]. The first column is used to store information about the tail number, the second column is used to store the label. Number of lines - depends on the number of cars in the train. After the data array is full, a decision is made to identify the type of object. The result of the decision is transmitted to the BST21 block.

An array storing thirty-two numerical coefficients is fed to the input of a three-layer neural network ANN1 16 for one cycle of operation. To identify the side number of an object, an eight-fold launch of a neural network is required. The input is an array of size [1 ... 8; 1 ... 32]. ANN1 16 from BP 19 loads synapse weight settings. For one operation cycle, 32 numerical coefficients are weighed in accordance with the settings of the weights of the synapses of the neural network, and in the output layer of the network a decision is made to identify one symbol. As a result of eightfold start-up of the network, the Reshenie [0 ... 1] vector, consisting of two cells, is filled. In the zero cell of this vector, the identified number of the object is stored, and in the first cell is the type of object identified on the basis of the received board number. Next, the type of object is transferred to the BST 21.

An array storing thirty-two numerical coefficients is fed to the input of the three-layer neural network ANN2 17. To identify the body of an object, a single start of the network is necessary. The input is an array of size [1 ... 3; 1 ... 32]. ANN2 17 from BP 19 loads synapse weight settings. Numerical coefficients are weighted in accordance with the settings of the weights of the synapses of the neural network, and a decision is made in the output layer of the network to identify the shape of the body of the object. Based on information about the shape and dimensions of the body, a decision is made to identify the type of object. The result of the work is a filled vector Reshenie2 [0 ... 2], consisting of three cells. The object type is stored in the zero cell of this vector, the length of the object in the first cell, and the height of the object in the second cell. Next, the type of object is transferred to the BST 21.

Thus, before making a decision on the classification of an object, the BST 21 stores information about the type of object obtained from three independent reserve sources (INS1 16, INS2 17 and BPTI 18). In reliability theory, this backup method is called the “Quorum”. After all the information is received, the control unit 22 gives an activation signal for the classification process of the object based on the results of comparing the three types of object. The decision is made by voting. In the case when all the data do not contradict each other, an unambiguous classification decision is made, the results are transmitted to FDA 24. Otherwise, a majority decision is made, and a message is sent to FDA 24. If all three sources are contradictory, a message is sent to FDA 24 about the error. FDA 24 displays the classification results on the display.

The operator 25 monitors the operation of the identification device. In case of errors, it has the ability to view the photo images of objects and make a decision on the classification yourself. Based on the identification of mobile units of the entire train, a decision is made to adjust or confirm the received full-scale sheet of the train of the ДУ-1 form. The results are transferred to ASOUP 9.

The claimed device allows you to implement the algorithm of quorum reservation of sources of primary information about mobile units of railway transport in the process of identification. As a result of the calculations, it was found that in the case of applying redundancy, the increase in the probability of success of the identification of the object is 28%.

Also, the device allows you to speed up the identification process by speeding up the processes of reading, transmitting and pre-processing the original graphic information through the use of digital photo images of the side number and the body of the object. Accelerated processing of information occurs due to the acceleration of the process of allocating the body of the object and the tail number by applying the method of chain coding of their contours and the formation of many signs of identification based on the chain code.

Claims (1)

A device for identifying mobile units of a railway vehicle, including a data reading unit from a moving mobile unit, a read data transmission unit, a read data processing unit, a read data recognition unit, and a data transmission and long-term storage unit, characterized in that a unit for reading data from a moving mobile unit (1) is connected to a unit for transmitting read data (5) and includes a photoelectric sensor for the presence of an object (BOT 2) in the inspection area and at least two cameras (3 and 4) associated with it, the read data transmission unit is connected (5) with the read data processing unit (10) and consists of an object identification number information input unit (BVIN 6), an object body information input unit (BVIK 7) and a process information input unit (BVTI 8) while the latter is associated with an automated system for the operational management of transportation through (ASOUP 9) network Intranet, the read data processing unit (10) is connected to the read data recognition unit (15) and includes a unit for pre-processing information about the side number of the object (BPON 11), a block for allocating contours of each of the symbols of the side number of the object (BVN 12), a block for preliminary processing of information about the body of the object (BPOK 13) and the block selection of the contour of the body of the object (BVK 14), made with the possibility of ensuring the selection and implementation of a specific algorithm for processing information, the read data recognition unit (15) is connected to the classification unit of types of moving units (20) and includes a block of a three-layer artificial neural network for identifying the tail number symbols (INS1 16), a block of a three-layer artificial neural network for identifying the body of an object (INS2 17), a conversion and verification unit technological information about a moving unit (BPTI 18), a memory unit (BP 19) of synapse weight settings and information about encoding types of moving units, the data transmission and long-term storage unit (23) is connected to the automated system for the operational management of transportation (ASOUP 9) and includes a computer unit (24) that is designed to monitor, control the identification process, generate, transmit and store reports on the results of identification of mobile units, wherein the device further comprises a unit for classifying the type of mobile unit (20), associated with a unit for transmitting and long-term data storage (23), and including a unit for comparing types of moving units (BST 21) obtained from at least three sources, and a control unit (BU 22) made with the possibility of synchronizing the blocks of the device by supplying activation signals.

Images