- AI
- NVIDIA
Upptäck hur en hamn kan bli mer effektiv och säker med AI. Föreställ dig ett avancerat hamnövervakningssystem som visar all trafik i realtid. Det är precis vad vi har åstadkommit tillsammans med Siemens och Aarhus Marina.
Genom att kombinera avancerade kameror, radarteknologi och Siemens IPC med Nvidia Jetson har vi skapat ett prisvärt realtidsövervakningssystem som hanterar en rad olika utmaningar. Spåra fartygsrörelser, säkerställ säkerheten och optimera kajplatsernas användning. Vår case study visar hur maskininlärningsteknik kan förenkla hamn- och marinaoperationer avsevärt och förbättra upplevelsen för alla besökare, utan att kompromissa med reglerna från datatilsynet (Integritetsskyddsmyndigheten). Vi har optimerat alla processer, från trafikhantering till säkerhet. Resultatet? Nöjd personal, glada besökare, obruten hamnverksamhet samt insikter och analyser i realtid för ledningen.
Vill du veta mer om vår övervakningsteknologi från vårt team?
"We work daily to ensure that sailors and visitors have a safe and pleasant stay in the harbour. But we lack data on traffic at the port. We lack answers to basic questions such as how many ships call and when - and which types. This knowledge is important in relation to our work with security in the port and the entire management of the port and its traffic. It could be, for example, when we have to give access to different boat types or other participants, such as the harbor tour and GoBoat."
says harbor master Thomas Krüger Bergstrand at Aarhus Lystbådehavn
Anonymized data with AI model
A major challenge was ensuring our data collection adhered to the strict guidelines of the Norwegian Data Protection Authority. On the one hand, data must be anonymized – and at the same time be precise enough to be used for analysis. It was an exciting challenge for us, and it was the start of a good collaboration between Danoffice IT and Aarhus Marina.
"There were many challenges in the task, but we came up with a solution where we combined several different data collection sources in the form of a thermal camera, 60 GHz radar and an ordinary color camera. We then used these data sources to train an AI model that is based on various techniques in computer vision and generative AI. The important thing in our training of the model was that we never used personal information. Instead, we taught the AI model to recognize shapes and contours of different boats in the harbor", says Leif Høj, Head of AI and Development at Danoffice IT and continues:
"Cameras and radar are mounted on two light towers which stand at the entrance to the harbor. By using these data sources, we can collect data even in foggy, rainy, and nighttime conditions, as well as when two ships are sailing side by side. For that, we have a setup so that if the regular and thermal cameras fail, we can combine with 60 GHz radar sensors".
"Now I have an online overview of the traffic at the port - and this has given me a number of new insights. Data is essential to confirm or disprove myths and hypotheses. Now I know, for example, that since last week in May we have had 30,000 sailings in two months, that the traffic already starts at lunchtime, and that the majority are larger sailboats. I see great potential for solutions like this. If we roll out the solution to the other large ports in Aarhus Municipality such as Marselisborg Marina, Egå Marina and Kaløvig Boat Harbor, we can gain new knowledge and a relevant overview of the traffic in the municipality's ports.''
says Thomas Krüger Bergstrand.
AI in all weathers
The interesting part of the test process is to investigate whether the equipment can cope with reality outside the laboratory. The solution must be exposed to harsh wind and weather conditions and be able to work both offline and online. That is why it is also important that it is a very robust solution that is largely self-sufficient. The data processing itself takes place in two Siemens industrial PCs (IPC) with GPU (Graphics Processing Unit). They are manufactured to have a very long life and are certified for use in industrial environments.
"It is important that the industrial PC (IPC) can run the AI model and deliver the required performance. That is why the Siemens IPCs are important for processing the data generated by cameras and radar. Siemens is also recognized by the entire industry for quality products, and they have a high delivery reliability", says Leif Høj.
In industrial AI applications, it's vital to choose hardware and software that meets 'Industrial Grade AI' standards. This ensures factors like reliability, uptime, security, 24/7 operation, and the ability to withstand demanding environments.
"Industrial Grade AI is built for industry. This is achieved by using well-tested hardware, where industrial reliability is intended from the start. In the same way, the software platform must be able to tie the various elements together – from sensors to cloud services and advanced AI models", says Anders Ottosson, account manager at Siemens.
