High-speed imaging and processing for rail track inspection

High speed imaging and processing for rail track inspectionDmitry V.Popov*a,Roman B.Ryabichenko*aa Moscow State Univ. of Railway Engineering (MIIT), 15 Obraztsova st., Moscow, Russia, 104075ABSTRACTThe safety of railway traffic depends on state of the track. About ten parameters are measured on Moscow Metropolitan for rail control. At present time the contact technology is used that doesn't provide required accuracy, limits speed of movement up to 25 mph and doesn't work in real-time mode.Non-contact photonic measurement system (KSIR) is developed which can works at speeds up to 70 mph.The KSIR consists of four subsystems: rail wear, height and track gauge measurement (BFSM); rail slump measurement (FIP); contact rail measurement (FKR); speed, level and car locating (USI).KSIR contains five CCD matrix cameras, four line CCD cameras, five infrared stripe lasers and four spot infrared lasers. Preliminary image processing is carried out using digital signal processor.The images from cameras are distorted because there is angle between photonic unit and rail. Additional distortions are caused by short-focus optics and small distance between camera and track. This distance is limited by structure clearance. For distortion eliminating is applied the transformation algorithms. It’s based on surfaces spline-approximation. As a result the KSIR calculates coefficients of approximating polynomials. The calibration is performed for checking accuracy of measurement in BFSM, FIP and FKR units.Keywords: CCD camera, image processing, rail track inspection, rail slump, calibration1.INTRODUCTIONMoscow city is one of the largest world capitals. Moscow has population about 17 million people. Metropolitan (subway) is the main type of transportation in Moscow. Every day metropolitan conveys more than 10 million passengers. Total length of lines is more than 280 km and it is took 4th place in the world. Rush-our headway is as little as 1 minute. Traffic safety depends on how well is organized communication between all services and devices of the Metro. One of the determining factors for metro traffic safety is condition of railway track. Measurement of track is carried out by the Track Service with help of following devices: track measurement car, defectoscope car, track measurement bogies and manual track measurement tools. Track measurement car allows measuring such rail track characteristics as gauge, slumps, level, lining and movement speed. Stakes are used for car location which are placed every 100 meters.The system of measurement is based on contact technology and works since 1954 year. The measurement results are recorded on paper tape and then will be decoded by special group of workers. Work of track measurement car is possible only in day hours because contact rail voltage is switched off in night hours. Contact technology limits car speed on rail switches down to 40 kph that increase time between trains. Automated system of received data processing is absent and accuracy doesn’t satisfy to modern requirements for track maintenance. It is necessary to develop non-contact photonic system with automated data processing for solution of this problem. New generation track measurement cars are constructed at present. In Russia it is TSNII-4 that allows measure 8 parameters of rail track at speeds up to 90 kph. Foreign systems: ORIAN1 and Electronic Bar Gauge2 which are produced in USA. In 1998 MIIT and Moscow metro began developing non-contact photonic measurement system (KSIR). The KSIR consists of four subsystems(Fig.1,2): rail wear, height and track gauge measurement (BFSM); rail slump measurement (FIP); contact rail measurement (FKR); speed, level and car locating (USI). KSIR contains five CCD matrix cameras, four line CCD cameras, five infrared stripe lasers and four spot infrared lasers.1.RAIL WEAR, HEIGHT AND TRACK GAUGE MEASUREMENTBFSM consists of four special digital CCD-cameras and four laser stripe illuminators ( two cameras and two illuminators on each rail ) which are rigidly connected and structurally fixed on a wheel set of the track inspection367 26th International Congress on High-Speed Photography and Photonics, edited byD. L. Paisley, S. Kleinfelder, D. R. Snyder, B. J. Thompson, Proc. of SPIE Vol. 5580(SPIE, Bellingham, WA, 2005) · 0277-786X/05/$15 · doi: 10.1117/12.567325vehicle in such manner that the corners of observation and the distances from the cameras up to the head of the rail during movement of the vehicle remain fixed. Preliminary image processing is carried out using digital signal processor.It performs preliminary calculations of the stripe image on the appropriate side of a rail and then compress it and transfer to the central computer. The image that transferred to the computer from CCD-camera has distortions caused by an optic system nonlinearities. To receive a frontal profile of a rail, it is necessary to perform restoring operation,which can be carried out with the help of the Euler planes’ formulas. During the simulation was detected that the obtained accuracy isn’t enough. It is connected with nonlinear distortions of an optical system. Distortions obtained when using Euler planes’ formulas can be removed by applying coordinates conformity system. A main part of the system is the test grid. The test grid is a special pattern with orifices located on parallel lines on the identical distance from each other. To establish a system of coordinates the orifices are used(Fig.3). The process of calibration was produced to define conformity of systems of coordinates between pattern and chambers. For this purpose the pattern was specially located instead of a rail profile and was captured by videocameras.The image of a rail profile receives by integration two rail halves from two cameras 3. The computer integrates two halves of a rail, evaluates an altitude of a rail, breadth of the head, compares to a reference rail, and produces outcomeon a screen and records on a hard disk. The monitoring behind two rails is executed in real time.Fig.1. Structural arrangement of inspection system.Fig.2. Disposition of subsystems on track inspection car.Y Fig.3. Calibration gauge.368 Proc. of SPIE Vol. 55802.CALIBRATION OF SLUMP SUBSYSTEM (FIP)The rail slumps are measured on a base which equal to the distances between wheel axis of rear car truck (L=2,1 m). This rail track parameter is defined as a difference between distances from wheel axis and truck centers. At present the cable system is used for slump measuring. The non-contact subsystem consists of four units – two units on each rail. There are point laser, line CCD-camera and metal plate in each unit.(Fig.4) Laser and CCD-camera are mounted on the car body. Metal plate is mounted on the car truck under laser. The laser forms small spot on the plate and the CCD-camera located with angle 45° gets image.Fig.4 The subsystem of slump measurementWhen slump value is changed the number of illuminated pixel on the line CCD-camera is changed too (Fig.5). Distance between laser and plate is 210 mm and range of measurement is 210-40,210+49 мм. The spot image is moved on CCD-line and illuminates appropriate pixel group. Resolution of CCD-camera is 0,152 mm/pixel. For slump value finding it is necessary to know dependence between distance and numbers of illuminated pixels. For solving this problem the system calibration was carried out that permits to compare number of illuminated pixel and distance in mm.CCD-cameraFig.5. Dependence between distance and number of illuminated pixelWhen system was in developing it was proposed that small spot from plate will illuminate only one pixel on CCD-line.Proc. of SPIE Vol. 5580 369But actually spread in values of number illuminated pixels achieves 20-30 elements. It is connected with that spot diameter is 2 mm and spot aureole is 5 mm and the spot image contains 5/0,152≈33 pixels.For exact and quick finding of slump value the calibration was carried out for each photonic unit. The plate was placed at the fixed distance from photonic unit for measuring number of illuminated pixel. For zero slump value the distance from plate to camera is 210 mm. Maximum distance deviations are 50 and 60 mm. Calibration was carried out at the distances from 150 to 260 mm every 5mm. For each value was made 3000 experiments. The results obtained were processed and statistical quantities were calculated: average of distribution, dispersion, quadratic deviation for number of illuminated pixel. The average of distribution M[N k ] of variate k N was calculated using formula (1):∑=⋅=n k k k N n N M 11][,(1)n=3000 – number of samples,N k – current value.The average of distribution diagram is presented in Fig.6.The average of distribution150155160165170175180185190195200205210215220225230235240245250255260mm p i x e l n u m b e r Fig.6. The average of distribution diagram.It is possible to see from this diagram that illuminated pixel number changing depends linearly from distance.For values obtained were calculated dispersion and quadratic deviation. Dispersion D was calculated using formula (2):∑=−=n k k x x M D 12)(,(2)where M x – average of distribution,x k – current value,n=3000 – number of samples.Quadratic deviation is calculated as root from dispersion (3):D=δ(3)The results of dispersion are in fig.7 and results of quadratic deviation are in fig.8.370 Proc. of SPIE Vol. 5580The dispersion2468101214161820150155160165170175180185190195200205210215220225230235240245250255260mmFig.7. The values of dispersion.The quadratic deviation150155160165170175180185190195200205210215220225230235240245250255260mm Fig.8. The values of quadratic deviations.Data analysis from Fig. 7 and 8 permits to make the next conclusion: pixel numbers obtained are adequate and the calibration was successful. Values from Fig. 7 and 8 shows that system is reliable and results of calibration can be used for rail slump measurement.Proc. of SPIE Vol. 5580 3713.SPEED, LEVEL AND CAR LOCATINGUSI subsystem solves two tasks4. The first task is line parameters measuring, namely: rail line lateral slope, i.e.modification of a relative position of rail on an altitude during a researched route. The obtained information definitely is treated, is archived and is mapped on the screen monitor of a regular on-board computer by BFSM in the table form or as the appropriate graph similar to an entry, obtained on a plotter of rail track monitoring station.The second task fulfilled by USI is the gathering and information processing. This information serves for definition of a site of the rail track monitoring station in each selected instant. It is necessary for calculation of linear coordinates of a location of a rail track deviation, for further elimination by their appropriate services to ensure for track maintenance are detected. Therefore, the realization of this task is the important factor in activity of all BFSM system. It is decided through measurement of two parameters: velocity of rail track monitoring station and number of stakes. Taking into account rather small length interoffice stages in the underground it is possible easily to determine on these parametersFig.9. Disposition of USI and sensors in car.The stakes in the underground are arrested special post, installed on crossties(Fig.9). On the dolly of rail track monitoring station the stake sensor is installed which being banged about post closes an electrical link of the compression type sensor and it arrests transiting the next stake, transmitting a commutation signal in USI. On rail track monitoring station there are such two sensors located from each side of a car. Except of measurement passed stakes sensors USI consists of two optoelectronic sensors of angular transition HEDS-5645 I14 of the Hewlett Packard corporation with resolution 512 and a unit of the interface ensuring signal processing from all sensors and feeds obtained information in the central computer. The first sensor ensures definition of lateral slope magnitude (interaltitude) of groundwork and is mounted on the gyroscopic system arbor of rail track monitoring station. Second –the speed sensor is mounted on the paper strip arbor of a plotter of rail track monitoring station. The pulse signals appears from sensors only in case of magnitude modification of traversed path (second sensor). The signals from sensors goes in the unit of the interface, where they are filtered, there is a direction finding of modifications (in “ +” or in “ - ”) and are counted for deriving a deflection rate.After that signals stored in the series output register. The pre-set (dumping in a zero) happens in the beginning of driving on pressing the “RESET” button.372 Proc. of SPIE Vol. 5580USI is implemented on the basis of the optoelectronic sensor and reprogrammed integral gate or in other words is Complex Programmable Logic Device (CPLD) of the XILINX Corporation. The connection with the central computer is carried out on a bit-serial interface RS-232C with a velocity of 57600 bits per second. The series port is customized on reception in a following condition: 1 launching bit, 8 data bits, without parity, 1 stop bit. At an angular variation of the speed sensor USI sends 2 bytes in the format shown on Fig. 10.Fig. 10. The format of the transmitted dataFor a reliability augmentation first byte transmitted from USI differs from second in a top digit. First byte stores seven low bits of a slope level (D0-D6); second byte stores last five bits of a slope level - (D7-D11), stake presence bit (stake at P=1) and bit of a slope level sign (at S=1 an angle is negative).In a fig. 11 the flow chart of USI is shown. The voltage limiter and optron decoupling are intended for a guard CPLD from noises of the relay assembly stake sensor of rail track monitoring station and galvanic isolation. The remaining blocks are structurally carried out as one chip CPLD of the corporation XILINX.Fig. 11. A functional diagram of the interface unit.4.RESULTSIn the fig. 12 BFSM and USI subsystems results are represented. The observed data of parameters of a rail track obtained are shown as the graphs at joint trials BFSM and USI, installed on rail track monitoring station. The form of representation of outcomes meets to the requirements of entries on paper strips of rail track monitoring coaches of an underground. Actual testing confirmed followed inspection accuracy:•rail head’s wear 0.2 mm,•rail track width 1 mm,•rail height0.2 mm,•track to track position at 1 degree.Proc. of SPIE Vol. 5580 373Fig. 12. KSIR output information.5.HARDWAREIn Fig. 13, 14, 15,16 illustrated KSIR devices. The general view of a manufactured sample USI is shown in a Fig. 16, where 1 - velocity sensor, 2 - sensor of slope, 3 - cable connector for connection to an onboard computer, 4 - control unit USI, 5 - cover of the USI unit.Fig.13 . BFSM in the car.Fig. 14. BFSM preprocessing unit374 Proc. of SPIE Vol. 5580Fig. 15. Inside the FIP unit. Fig. 16. A general view of a sample USI.6.CONCLUSIONThe KSIR systems of railway measurement are developed using photonics technologies that allows to define rail profile using non-contact way. Calibrations of BFSM and FIP subsystems are required for achieving required accuracy. Optics nonlinearities include aberrations are removed in calibration process.7.REFERENCES1.Daniel L. Magnus, “Non-contact technology for track speed rail measurement (ORIAN)”, SPIE Vol.2458, pp.45-51.2.Gordon S. Bachinsky, The Electronic BAR Gauge (a customized optical rail profile measurement system for railgrinding applications), Opt. Ing., vol 2458.3. D. Popov, O. Smoleva, “Algorithms and software to determine rail profile using moving distorted images” in 2Conf on Photonics for Transportation, Proc. SPIE Vol. 4761, pp. 48-554.S. Larionov, S. Kabov, D. Popov, E. Titov “Optoelectronic sensors to measure velocity and level for MoscowMetro rail track monitoring station” in 2 Conf on Photonics for Transportation, Proc. SPIE Vol. 4761, pp. 70-75* popov@rapsmps.ru;ryabichenko@rapsmps.ru; phone 7(095)369-56-28; fax 7(095)631-66-20Proc. of SPIE Vol. 5580 375。