SUCCESS CASE: RENFE

Real-time machine vision
inspection of wagons

«AllRead makes data capture faster and easier and facilitates invoicing tasks. It offers a flexible, scalable and future-oriented solution that allows us to process both images and videos using a mobile device or tablet.».

David, Rail Operator for Renfe Mercancías.

• Company: RENFE Mercancías.

• Industry: Rail

• Success Case: Real-time machine vision inspection of wagons.

The challenge

Renfe Mercancías is the main rail freight operator in Spain. Every year, it transports more than 18 million tonnes of freight

At Renfe Mercancías rail terminals, before each train departure, it is compulsory to carry out a train check. This involves a process of verification and data capture to obtain the complete structure of the train, from which the following data is extracted:

BIC codes of the container

UIC Wagon
Codes

Dangerous goods signs

Twist Locks
.

This verification of the elements is necessary both for operational efficiency and to comply with current regulations.

From TrenLab we knew we were ambitious when we defined the challenge, in which we were looking for transformative, innovative and technological solutions that would improve the operation and competitiveness of freight transport. AllRead has perfectly matched what we were looking for in the startup ecosystem, with a solution with traction and validated in other clients in the logistics chain“, says Andrés Gómez, Head of Open Innovation at Renfe.

The problem: Manual reading and recording of data

The data collection process has to be carried out for each train manually. Once completed, the collected data has to be entered by typing it into a computer. For Renfe Mercancías and its operators, this is a slow, repetitive and error-prone task, leading to higher costs, longer review and correction times, resulting in lower operational efficiency.

In addition, the current data collection process does not provide any visibility of the physical condition of trains and containers, which consequently creates additional difficulties in the audit process in the event of an incident.

Finally, the time required for data collection implies a greater presence of employees, and for longer periods of time, in operational areas with rail traffic.

In the words of Pilar Górriz, Renfe’s Innovation Manager, “one of our commitments through our TrenLab innovation programme, in which we worked with startups for 6 months, was to work on an operational solution. This solution was not only good for Renfe staff, who work much more comfortably and much more operationally, but it also makes us earn money and be much more efficient. In other words, we are all gaining in the value chain. Moreover, in the process we are supporting the entrepreneurial fabric of our country, helping a start-up and giving it a business opportunity, thus closing a chain“.

Renfe Mercancías requirements

Based on this problem, Renfe Mercancías needed a solution that would allow it to collect the information immediately in order to reduce the duration of the process and increase the security and traceability of the information.

  • Digitisation of containers, UIC wagons, dangerous goods and twist locks.
  • Need to store images for later audit.
  • Ease of implementation and use.
  • It works in adverse conditions.

And other technical requirements:

  • No internet connection required.
  • Real-time operation.
  • Data that can be integrated into the TOS.
  • Scalable to other terminals.
  • Scalable to new visual inspection functionalities.

The solution: automatic digitisation with the AllRead APP

AllRead provided an APP to process video along the train, and extract the required data in real time. The operator points the Tablet at the places of interest, and visualises how the codes are digitised and automatically ordered to provide the structure of the train: each wagon, with its containers, twist locks and dangerous goods.

Each time a reading is made, it appears on the screen monitored by the operator. When necessary, he can correct or add information. Once the train’s journey is complete, the operator checks the list and validates it so that the data is sent to the information system immediately.

Key factors: Robust but lightweight neural reading models

Thanks to AllRead’s reading systems, based on Deep Learning and Computer Vision, we are able to offer a solution that reads data in any situation. External influences such as difficult perspectives, lack or excess of light, atmospheric conditions… They do not affect the reading quality. However, AllRead’s system requires few resources to process the images, enabling operators to obtain the results in real time.

In the words of María Cristina Rodas Cabezas, head of the marketing department at Renfe:

“We have been working with the company Allread throughout the process of the pilot test of the 4th edition of Trenlab, Renfe’s startup accelerator, and the assessment has been positive. It was a period in which adjustments had to be made, but they were flexible in understanding our needs and have been creative in finding solutions, developing a pilot with great potential to advance in the digitisation of the training of our trains”.

Results: more accurate readings

At the end of the pilot project, the AllRead software provided:

  • 95% automatic reading of wagons (i.e. in 5% of cases, the code had to be entered manually).
  • 81% automatic container readings (work in progress).
  • 100% automatic identification of dangerous goods (RID).
  • 100% automatic identification of Twistlocks.
  • Increased operational efficiency: data is sent immediately at the end of the train’s journey, and does not require the operator to return to the office to enter the data into the system.
  • Test images for audits and litigation.

The benefits expected by Renfe Mercancías

Based on these promising results, Renfe Mercancías is planning to deploy the AllRead application in production at several terminals. The benefits expected by Renfe Mercancías are as follows:

  • Operational Efficiency.
  • Traffic safety.
  • Scalability.
  • Quality of information.
  • Auditable images.
  • Respect for the rules.

For Renfe, this will mean greater operational efficiency, increased traffic safety, better information quality, improved regulatory compliance and accelerated decision-making.

Do you want to optimise the monitoring of internal traffic flows?