Artificial intelligence supervises operations at maritime terminals
Artificial intelligence technologies based on machine learning facilitate the implementation of changes in operations at container terminals. This is an important expansion in applying the optical character recognition systems (OCR) at the terminal gates.
To find out how it is important for intermodal transport to process information quickly, you can imagine a container loss at a major maritime terminal, which tranships 115,000 TEU per day. In this case, it is impossible to find the lost unit on the same day or even a week. For this reason, digital technologies for monitoring traffic at the terminals are used on a large scale to avoid losses.
Optical character recognition systems (OCR) were first implemented in Asian ports in the 1990s, and later in other regions. The devices read information such as the container number, ISO code, container capacity and weight, registration numbers of lorries entering and leaving the terminal, VIN code and registration number of a semi-trailer, etc. These data transferred to the central terminal management system allow improving the operations and minimising the risk of mistakes in transhipment. The information scanning devices are installed at the entrance gates of terminals, as well as on cranes and other reloading equipment.
Dawn of OCR systems
The first generation of the OCR systems was based on image pixel recognition and statistical analysis, so that they could convert an image into a text file (just like in scanners). These devices ensured a fairly high accuracy in reading the information, although there were mistakes if the inscription was unclear. The second generation of scanning software was able to recognise such features as shape, colour and texture. Due to this, it became possible to develop optical function recognition (OFR) systems that provide information such as the position of a container door, type of freight, the position of a container on a platform, etc. At the same time, work continued on the development of convolutional neural networks that could be used to analyse images from scanners.
In traditional OCR systems, all shapes and other object features must be defined and coded manually. As a result, implementing changes and extending the software functionality is time- and labour-consuming. Moreover, the OCR systems sometimes generate false positives. Convolutional neural networks, which allowed using deep learning functions to develop a system for analysing scanned objects, was a revolutionary technology that overcame these difficulties. This technology has been quickly developed since 2010, thereby artificial intelligence can be used today in many practical solutions. The neural networks do not require manual definition coding, but they analyse thousands of images themselves and create algorithms to recognise the desired information.
Large database of images
Belgian company Camco, which supplied the OCR systems to many container terminals around the world, was one of the pioneers in the practical implementation of AI solutions in the transport sector. Camco collected information from the scanners and used them to learn neural networks reading key information on what is going with containers and freight at terminals.
“Our large database of high-quality images has made us a leader in AI-driven terminal automation. New applications assist terminal employees to work better and safer. Automation is entering the container shipping industry at the speed of light,” comments Jan Bossens, CEO of Camco Technologies.
Artificial intelligence has revolutionised how computers work and how we can work with software. Neural networks work much like the human brain – they learn through experience. Programming new functionalities involves learning under supervision. The developer supplies a large number of images and checks whether the system learns to recognise the necessary information.
Reliability improvement
With the deep learning function, implementation of changes and new options to the software that analyses data from terminals is much faster and more reliable. The number of the OCR misreadings has also been reduced. For instance, after implementing the systems based on machine learning for analysis of images from terminal gates, the percentage of correctly read lorry license plates at terminals in the US and South America increased from 84-84.8% to 98-99.4%. This is a big improvement as you can see!
However, increasing the reliability in reading information is not enough. Artificial intelligence provides terminal operators with completely new possibilities, such as container damage recognition. This is a complicated task, but Camco is currently developing automatic damage detection technology. AI is also able to predict the movement of objects such as containers, which can be used to speed up the crane operations. The network is also able to recognise if a container is properly secured or if there are workers under the crane.
“Artificial intelligence is the most effective technique that uses deep learning and neural networks to analyse images. We also use AI in other applications: in predicting the movements of crane grippers. This information controls our moving rail-mounted cameras, but this AI application, in particular, is currently used only at maritime container terminals,” explains Alain Buyle, Global Marketing Communication Manager at Camco Technologies.
AI systems require the preparation of appropriate computer equipment on which they can work. Remarkably, they use the computing power of graphics processors in contrast to classic software. Camco works with NVIDIA, a well-known graphics card manufacturer that provides GPUs dedicated to AI software. Camco uses camera-mounted GPUs. This solution ensures high speed and eliminates the issue of delays in sending information, which occurs, for example, when using cloud computing. Thanks to this, the OCR technology supported by AI systems can accelerate the work of reloading equipment.