Real time Identification of Railway Track faults using image processing
Pakistan ranks 12th in the world where passengers prefer to travel by train. Although rail transport is the safest approach to public transport, Pakistan has a poor record of train accidents. Over the past decade, the country has faced several fatal accidents. And it seems to have increased in recen
2025-06-28 16:34:43 - Adil Khan
Real time Identification of Railway Track faults using image processing
Project Area of Specialization Artificial IntelligenceProject SummaryPakistan ranks 12th in the world where passengers prefer to travel by train. Although rail transport is the safest approach to public transport, Pakistan has a poor record of train accidents. Over the past decade, the country has faced several fatal accidents. And it seems to have increased in recent years.
Below are some accidents that happened because of minor failures in which Major casualties occurred.
On July 11, 2019: A Quetta-bound train collided with a cargo train near Saidabad in Punjab, killed 24 people and injured over 100. The accident later resulted in a delay in the exchange of train tracks.
On September 27, 2018: A Peshawar-bound train had derailed in Sehwan, Sindh because 11 bogies were overturned.
On September 16, 2018: Nine bogies of the Peshawar-bound Khushal Khan Khattak Express from Karachi derailed near Attock. Twenty passengers were injured.
According to the statistics, Accidents of rail transportation mainly caused by obstacles appear on the rail, human congestion, vandalism, signal systems failures. Still, among them, the actual cause is track faults become the reason for the train derailment. The lack of attention or resources that we use to monitor the railway track is not efficient enough. In Pakistan, workers monitor railway tracks manually time-to-time basis through buggies. Pakistan Railways lack new technologies; therefore, human error chances are more, and it is one of the significant causes of rail accidents in Pakistan. Though nothing can be foolproof with technology, it certainly reduces the chances of accidents.
Our solution enables real-time identification of railway track using an Image Processing Algorithm. This approach enhances track maintenance and helps to ease the technicians who monitor the track through heavy boogies. This detection technique uses an efficient algorithm of image processing for the model and prepares a model with defective and non-defective track images. Once the model is ready, we will provide an image of the track surface as an input to the model. The model can compare the input image with the images already provided in the dataset and inform us where the track contains any fault. The model will inform the track’s condition when our algorithm finds any defective point in the track. So, that maintenance of the track will be speedy and efficient.
Project ObjectivesWe aim to design a model that makes railway track monitoring automatic, efficient, and fast approach to detect faults in the Railway Track so that maintenance can happen at the right time to avoid unfortunate incidents and reduce human error happening during monitoring of track.
The key project objectives are stated below :
- Utilizing the Artificial Intelligence technique.
- Update existing data set with real-time images.
- Use an anomaly detection technique for image processing algorithm to Train and test on the dataset.
- Make Railway Tracking System Smart.
- Reduce railway track accidents.
- Make Railway Track Fault detection easy and cost-effective.
- Collection of defected and no-defected track images in the dataset.
- Deploy our model in traditional carts for Surveying Track.
Our Project is the Deep learning model that detects damages and faults in railway tracks in real-time. Our Solution is based on an efficient algorithm of Deep learning named Convolutional Neural Network, which is used for image processing. We will implement our network into the 3D model Proto-21 of suitable dimension that we design using autocade software, which deploys on railway train buggies or Survey vehicles. We trained the model with the dataset containing images of the defective and non-defective railway track; the training process will continue until efficient training statistics are obtained. The 3D model has IP cameras that examine the track surface using image processing techniques and provide data to the input layer. Our model will compare the input image with images already provided in the dataset after the classification model will generate the result whether the track surface contains any fault or not.
A block diagram of complete implementation process is shown in below in figure 1 :
Figure 1
Benefits of the ProjectBenefits of using a Real-time Railway Monitoring Systems are :
- By implementing this type of monitoring systems in Pakistani Railway tracks, we can prevent disastrous accidents.
- We use Anomaly detection technique which is the most efficient approach for the detection of images.
- Our model can prompt changing over the Manual Rail Inspection System in Pakistan into a real-time Rail Inspection Approach, just with the assistance of image processing.
- Our model can remove human error.
- Faster track maintenance due to the most efficient detection technique.
- Help to ease the technicians for monitoring the faults.
- Make a railway track system smart.
- And Our model will also reduce labor cost.
The final product will be in the form of a hardware model with IP cameras, which will be installed on inspection vehicles of railway train or bogies that fetch the images of the track surface. Then it inputs into the suggested system where we have used an efficient algorithm Convolutional Neural Network and detection technique of neural network called Anomaly detection. The captured footage will be enhanced and converted into grayscale using image processing techniques. The processed data provide to the input layer further to improve the efficiency model contains few hidden layers where the model will finally compute its results finally at the output model will classify according to result whether the track surface contains any fault or not and inform the system. So that maintenance will occur at the time and avoid misfortunate accidents.
Final Deliverable of the Project HW/SW integrated systemCore Industry ITOther Industries Education , IT , Others Core Technology Artificial Intelligence(AI)Other Technologies OthersSustainable Development Goals Affordable and Clean Energy, Decent Work and Economic Growth, Industry, Innovation and Infrastructure, Sustainable Cities and Communities, Responsible Consumption and ProductionRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 70000 | |||
| Proto-21 3D model | Equipment | 1 | 19500 | 19500 |
| SST Mart Mini IP camera | Equipment | 2 | 10000 | 20000 |
| NVIDIA 1060 4gb | Equipment | 1 | 24000 | 24000 |
| Glue gun | Miscellaneous | 1 | 500 | 500 |
| Poster | Miscellaneous | 1 | 1000 | 1000 |
| Thesis printing | Miscellaneous | 3 | 1000 | 3000 |
| Soldering iron | Miscellaneous | 1 | 500 | 500 |
| 12 v Battery | Miscellaneous | 1 | 500 | 500 |
| Cables of GPU | Miscellaneous | 2 | 500 | 1000 |