4 min read
The Port of Antwerp is the second largest port in Europe. Thanks to its situation at the upper end of the tidal estuary of the Schelde river, its docks have excellent connections. However, with its size often comes congestion.
Already an innovator in its field and keen to use new technologies, the port had invested in a digital 3D version of its docks simulated in real-time. But it still relied heavily on manual operations due to limitations of current localisation technology, such as GPS and radar.
The port asked Deloitte Belgium to help come up with a pioneering solution that could determine the exact location, arrival and departure time of a ship so it could make optimal use of its docks.
Which technologies would be most suitable to get an accurate view of how the shipping berths were used, taking into account current operational and legal constraints? Our business and technology consultants worked together with the port, reviewing everything from cameras and flying or floating drones to pressure and proximity sensors.
The team decided to work with the existing camera infrastructure as vision sensors, in combination with the Automatic Identification System (AIS) of ships coming into the port. They then developed a proof of concept around image recognition and object detection using artificial intelligence (AI) and neural networks to provide accurate, real-time information on a ship’s location.
Collecting thousands of images from camera feeds at the port over the course of two weeks presented a challenge: how was it going to process and label them all? The AI application had to be trained to understand the ships and other objects in images before it could make any judgements on a ship’s location.
The answer was a collaboration with Spectrum VZW, a local organisation that supports adults with learning disabilities. "A team of eight people took on the challenge of working on the data sets and labelling the ships. For two of them, it was the first time they had used a computer," said Wouter Travers, part of Deloitte Belgium’s analytics team.
“This project was a new experience for our organisation and our people. Some only had limited computer experience but turned out to have a real talent for the work. Being able to contribute to a project for an external company was very motivating.”
With all of the images labelled, the team could improve the AI application’s accuracy. Results were promising, with the model being able to work under various lighting and weather conditions. But more was needed to define a ship’s location to within less than a meter – a prerequisite to make the most of berths.
The team built a fake dock using lunch boxes as ships to help improve accuracy. “We wanted to try placing a fluorescent strip along the quay,” explained Tim Paridaens, our engagement director on the project. “We figured the model would be able to detect the strip under any weather conditions and could then determine which parts of the quay are occupied based on the parts of the strip that are blocked by docked ships.” Using a scale model allowed the team to experiment and establish which type of strip would be most precise.
“This project shows the benefits of AI as an important tool in today’s world. For the Port of Antwerp, rather than replacing people’s jobs, it’s helping to create them.”
The port is currently investigating whether it can roll out the AI application from its pilot location to cameras across the port. While it is still in testing, it looks set to help the port take the step to digital, real-time representation of its docks and optimise the use of berths.
For Wouter, the highlight of the project has been the partnership with Spectrum. “This was one of the best experiences I’ve had working for Deloitte,” he said. “It was meaningful to collaborate with Spectrum and open up new opportunities for the team as AI trainers. The people who helped us experienced the kind of work fulfilment that we often take for granted. It made us go to work happy.”
“As part of our innovation strategy, we want to discover the potential of new technologies at an early stage. Computer vision, which turns camera feeds into actionable data, is one that holds huge potential for us. The use case we worked on with Deloitte was not easy, given that ships needed to be detected day or night and in all weather conditions. What I really appreciated about this collaboration was the creativity and tenacity Deloitte demonstrated to overcome the many challenges.”