Railways are one of the most important components oftransportation. It is basic need that the railways has been examine at regular periodof time in order to ensure the safety of railway transportation to prevent thedisruption of transportation, and avoid accidents. To inspect of railwaysurface by using manual techniques, both damages and defects of the railsurface and leads to disruption of railway transport.

Railway analysis methodsthat utilize contactless image processing techniques are available in theliterature in order to avoid these problems. This paper presents a comparisonof rail defect detection methods that are available in the literature. Thesemethods in the literature have been compared in terms of feasibility, performance,accuracy, elapsed time, and image processing techniques used. The pros.

and cons.of these methods relative to each other are examine in this paper. Keywords—Raildefects; Image processing techniques; Defect detection; Rail surface analysis,Crack detection, Wear defect, Rail Failure, etc.                                                                                                                                                     I.      IntroductionRailway is consider as one ofthe safest transportation medium types all over world.

The railroad isspreading rapidly over transportations. As railway has become widespreadthroughout world, the importance given to the maintenance and safety ofrailways has also increased. Railways are composed of several components. Themost important component is rail. Train accidents happened every year in theworld due to heavy task. And the train accidents resulted in seriousdestruction of property and injury or death of passengers and crew members 1.Many of the railway transport accidents happen because of driver’s tiredness,bad weather conditions, and defective rail components, etc. To prevent theseaccidents, importance is attached to the detection of faulty regions in thetracks, and other rail components.

Safety of railroadtransportation can be enhanced by utilizing intelligent systems that provideadditional information about the exact location of the train, its speed andupcoming obstacles. The rails face more and more risk of damage with theincrease of speed 2. Therefore, the rails should be closely inspected forinternal and surface faults. Rail profile analysis using manual methods bothdamages the rail surface and temporarily disrupts railway access. For thisreason, rail profile analysis for railway transportation has been done usingcontactless image processing techniques. Methods, which detect the railfailures by means of contactless image processing techniques, are available inthe literature.

Aliza Raza Rizvi, Pervez RaufKhan, Dr. Shafeeq Ahmad, 1 Introduce computer vision based method in paper. Asystem has been suggested which can periodically take images of the railwaytracks and compared with the existing database of non-faulty track images on acontinuous basis. If a fault arises in the track section, the system willautomatically detect fault and necessary actions can be taken to avoid anymishappening. Main objective is to find crack in railway track through cameraimage and fix it using computer vision based method.B.

Sambo, A. Bevan, C. Pislaru,2 introduce an intelligent image processing algorithm capable of detecting fatiguedefects from images of the rail surface. The algorithm generates statisticaldata (such as total number of detected defects per image, damage index ofentire image, specific region of interest (ROI)).

Adaptive histogram equalizationis use for the local contrast enhancement so the defect regions are clearlyvisible. Then an adaptive threshold method is employ to segment the defects andpredict crack growth rate and direction. Main objective is to detect fatiguedefects from images of the rail surface.

Also, detect crack in railway trackdue to shear stresses, hydraulic pressure and fluid entrapment and squeeze filmeffect.Shahrzad Faghih-Roohi, SiamakHajizadeh, Alfredo Nunez, Robert Babuska and Bart De Schutter, 3 introduce DeepConvolutional Neural Networks (DCNNs) for automatic detection of rail surfacedefects. Data resembles that of for visual inspection of rails. One immediateadvantage of using a DCNN is that unlike, they do not have to go into anelaborate procedure for the extraction of features. They can rather use rawimages as input to the classi?cation model, which is subsequently optimizedusing a mini-batch gradient descent method for the entire network. They comparethree DCNNs with different structures for their classi?cation accuracy andcomputation time. Main objective is deep convolutional neural network solutionto the analysis of image data for the detection of rail surface defects. S.

Sam Jai Kumar, T. JobyTitus, V. Ganesh, V.S. Sanjana Devi, 4 introduce ultrasonic sensor is used todetect the crack in the railway track by measuring distance from track tosensor, if the distance is greater than the assigned value the microcontrolleridentifies there is a crack, also it tells the exact location of the crack bythe formula “DISTANCE=SPEED*TIME”.

While the checking process is going on, thetrain may approach, it is identify by the vibration sensor and gives alert tothe microcontroller. Railways are one of the important transports in India.There is a need for manual checking to detect the crack on railway track andalways railway personnel takes care of this issue, even though the inspectionmade regularly. Sometimes the crack may unnoticed.

Because of this the trainaccident or derailment may occur. In order to avoid this situation and automatethe railway crack detection has been need to implement.Gaolong Hu, Ling Xiong,Jianqiao Tang, 5 introduce the basic idea of mathematical morphology is touse geometric template (circular, square, rhombus, linear, etc.) of structuralelements with certain shapes to measure and extract the corresponding shape ofthe image which in order to achieve the purpose of image analysis andrecognition. this method owns strong anti-noise performance, can detect thesmall defect edge accurately under noise, and the peak signal to noiseratio(PSNR) is 24.SdB in the condition of without reducing the detection speed.Main objective is to detect heavy rail surface defect due to uneven brightnessand noise.

Orhan Yaman, Mehmet Karakose,Erhan Akin, 6 introduce An FPGA based method is proposed for rail surfacedetection in railway. The propose method is realized by image processing withFPGA. The image taken on the railway line with the camera attached to the FPGAdevelopment board. Preprocessing is perform on the obtained image. Edgeextraction is apply to the image after pre-processing. The rail surface isdetect using the image obtained because of edge extraction. The propose methodworks in real time to monitor and diagnose faults.

It detects many defects onthe track surface. The propose method is quite advantageous because of itsreal-time operation. Main objective is to check continuously the componentsconstituting the railway line.V. R. Vijay Kumar, S.

Sangamithirai, 7 introduce Binary Image Based Rail Extraction (BIBRE)algorithm to extract the rails from the background. The extracted rails areenhance to achieve uniform background with the help of direct enhancementmethod. The direct enhancement method enhance the image by enhancing thebrightness difference between objects and their backgrounds. The enhanced railimage uses Gabor filters to identify the defects from the rails. The Gaborfilters maximizes the energy difference between defect and defect less surface.Thresholding is done based on the energy of the defects.

From the thresholder image,the defects are identified and a message box is generated when there is apresence of defects. Main objective is to detect the surface defect on railheads.In order to identify the defects, it is essential to extract the rails from thebackground and further enhance the image for thresholding.Yuvashree G, S. Murugappriya8 introduce System captures the video of the track from the vehicle that hascamera on the base of the vehicle. This system detects the rail cracks andmisplaced bolts in the tracks.

The system the monitoring and structuralcondition for railway track using vision based method and calibration to searchthe fault location on the track. The percentages of abnormalities are sent tothe maintenance vehicle Driver by hardware unit placed on the drivercabin.  To prevent such scenario, thepropose system will automatically inspect the rail crack, misplaced bolts anddeadheaded spikes in the railway track. Vision based method camera will be usedto capture the video.Zehui Mao, Yanhao Zhan, GangTao, Bin Jiang, Xing-Gang Yan, 9 introduce the suspension system states areaugmented with the disturbances treated as new states, leading to an augmentedand singular system with stochastic noises.

Using system output measurements,the observer is designed to generate the needed residual signal for faultdetection. Existence conditions for observer design are analyzed andillustrated. In term of the residual signal, both fault detection threshold andfault detectability condition are obtained, to form a systematic detectionalgorithm. Simulation results on a realistic vehicle system model is present toillustrate the observer behavior and fault detection performance. Main objectiveis to develop a sensor fault detection scheme with some detection rates for thesuspension system, in which track irregularity is modeled as unknown externaldisturbance, and processing and sensor noise are modeled as stochastic zeromean white noise. Fig. 1.   Blockdiagram of rail failure detection in the literature 9.

In this paper, we comparedseveral methods that exist in the literature to each other. It is examined railfault detection algorithms, the performance of these algorithms, and theadvantages and disadvantages of these algorithms in relation to each other.Moreover, these algorithms have been compared in terms of feasibility. Comparativetables are presented in the following sections.

                                                                                                                                         II.    RailSurface DefectsFailures that occur in therails can expressed as wear, scour, breakage, undulation, head check, andoxidation. Horizontal and vertical abrasions occur on the surface where therails are exposed to the wheel. If the amount of wear on the rails is greaterthan 33 degrees, railings will be changed or curbing will be done because ofclimbing. Rail erosion occurs in horizontal curves, in scissors tongues. Rawabrasions are divided into vertical and lateral wear.

Vertical wear are erosionin the form of spreading and crushing, which occurs in the rail mushroom ofcurves, in the corners of the scissors and on the rail heads in the seals.Lateral abrasion occurs on the inner cheeks and scissors tongues of the outerrail under the influence of centrifugal force in the curves 2, 3, and 7. Headcheck defect is foundaround the gauge corner of outer rail and this fault ascending inclines tohappen when cracks reach 30 mm in surface length. The undulation failure can beexpressed that different collapses happen in the rail surface 1, 4. The scourfault that can happen in the rail is one or several places of the rail due tothe spinning of the locomotive. It should be exchanged rails exceeding theamount scour 4. Rail oxidation is that crusting, decay, rust and small holesoccur in the rail by effecting humidity, soil and water 1.                                                                                                                  III.

   TheRail Defect Detection MethodsMany methods that detect rail surface defects exist in theliterature. These methods employ contactless image processing techniques. Sothat the rail surface is not damaged.

Besides, possible accidents are preventedby early detection of many rail failures. Rail fault detection methods that arefrequently used in the literature are presented below. A.    RailDamage Detection using Neural NetworksAn onboard measurement systemis for measuring the rail robot’s excursions from the rails midlines and therail-robot’s heights above the rail 10. In this method, to deal with thenonlinearity of the measurement models, the coupling between the outputs, andthe noise contamination, a neural network method is for building high precisionmeasurement models 10. In addition to different measurement models fordifferent types of rail tracks are also built based on the proposed neuralnetwork module. Signal processing and neural network module of the method used in10 appear in Fig. 8.

Fig. 8.   Signal processing and neural networkmodule in 1. B.    RailFault Detection based on the Morphology of Multi-scale and Dual-StructureElementsHeavy rail surface defects detected based on themathematical morphology of multi-scale and dual-structure elements according tothe characteristics of heavy rail surface defects, uneven brightness and noisein 5.

When this method is compared with the traditional edge detectionoperators, the results show that this method owns strong anti-noiseperformance, can detect the small defect edge accurately under noise. Using themorphology of multi-scale and dual-structure elements, defects such asscratches, rolled-in scale, and uneven rolling on the rails are detected 5. C.    RailDefect Detection using Gabor filtersIn the 7, Binary Image Based Rail Extraction (BIBRE)algorithm is use to extract the rails from the background.

The extracted railsare enhance to achieve uniform background with the help of direct enhancementmethod 7. The direct enhancement method enhance the image by enhancing thebrightness difference between objects and their backgrounds 7. The enhancedrail image uses Gabor filters to identify the defects from the rails. The Gaborfilters maximizes the energy difference between defect and defect less surface.Thresholding is done based on the energy of the defects. From the thresholderimage, the defects are identify and a message box is generate when there is apresence of defects 7. The faulty rail image taken as input and the faultyregion detected those are showing in Fig.

9 7.                                                        IV.   ComparisonStudy of The Rail Defect Detection Methods ?n The LiteratureStudies in 1 to 9 have been compared in severalrespects. These respects are algorithm’s accuracy rate and operation time, feasibility,techniques used in fault detection, detectable failures, hardware resourcerequirement, used software development environments, and images used in thealgorithm.  The advantages anddisadvantages of these algorithms relative to each other are given in thefollowing table.

 TABLEI.              A COMPARISON OF THESE METHODS IN LITERATURE Techniques used in fault detection Detectable failures Hardware resource requirement Used software development environments Feasibility Algorithm’s performance criteria Neural Networks Proximity Sensors   Detect nonlinearity of the measurement models (defects) Hardware required (Proximity Sensors) – High High accuracy rate Mathematical morphology of multi-scale   Scratch defect Backfin defect Uneven rolling Rolled-in scale Hardware required (CCD Camera) – Medium Strong anti-noise performance Peak signal to noise ratio is 24.5 dB Binary Image Based Rail Extraction (BIBRE) algorithm Gabor filters Thresholding Texture analysis Scour defect Hardware required (Digital camera of 12 megapixels) Matlab High Accuracy rate 89.9% Otsu method Canny edge detection Hough transform Gauss Filter Wear defect Hardware required (Special light sources and laser camera) Matlab High 84.3 millisecond elapsed time Standard deviation 1.5 The approximate speed of the system for 1 frame is 12fps Morphological feature extraction Hough transform Edge detection Laplacian filter Gradient computing Headcheck defects Breakage defects Apletilic defects Undulation defect No hardware required Matlab Medium 0.

6 sec. Accuracy rate 85.3% Axle box acceleration (ABA) measurements Wavelet power spectrum Low-pass filtering Squats defect No hardware required – Medium Accuracy rate 78% for light squats 100% for severe squats     ConclusionIn this paper, we review various methods to detect crack detectionin railway track to avoid accident and mishappening. There are list of methodscan provide better result on their own review but all methods have somelimitation due to uneven action held in railway track. In this paper, we found thatneural network is provide high accuracy rate than other method. It is feasibleto implement method using only sensors.

Neural network is better than othermethod in terms of feasibility, performance, accuracy, elapsed time and imageprocessing techniques used.