Automatic Registration of 3D point cloud scans in Cruise Ship Refurbishment
MJM Marine, a leading international specialist in cruise ship refurbishment and outfitting with over 40 years of experience, provides tailored solutions to its global clients. As part of their project delivery process, MJM Marine survey client venues and create highly accurate digital twin models.
The Challenge
The data captured from multiple scan locations needs to be combined into a single digital twin model. The software automatically combines the scans but this can fail, resulting in having to manually register the scans which is a time-consuming task. This manual registration quickly bottlenecks the process when multiple surveys occur simultaneously. The challenge was to identify and test new methods of automatically combining the scans using the latest innovations in AI algorithms and open-source software. If a new process could be identified (and proven reliable) then the workflow could be deployed to run 24/7 enabling MJM Marine to scale their surveying operations significantly.
The Solution
The Hartree Centre Northern Ireland Hub conducted a comprehensive review of open-source, GPU-accelerated algorithms for point cloud registration. After evaluating several options, Fast Global Registration (FGR) and Random Sample Consensus (RANSAC) were selected for their speed and robustness. The team processed nineteen-point cloud scans, converting them to compatible formats for use with the Open3D library. Multiple registration pipelines were developed, including steps for voxel-based downsampling, noise removal, and normal estimation. The most effective pipeline combined FGR with Multiway Iterative Closest Point (ICP) to enhance accuracy and reduce Root Mean Square Error (RMSE).
Declan McAuley, Research and Application Engineer
"We really appreciate the effort the Hartree Centre Northern Ireland team put into the project. They took the time to understanding our process, to research the technologies and to build the test case."
Project Impact
The project successfully evaluated the feasibility of automating point cloud registration using open-source tools. Although software exists that implements the various registration algorithms, the testing highlighted that they were unable to accurately align the scans into a model we could use. We will continue with our current process but revisit this again as open-source software develops in the future. The insights gained through this project have been valuable to the team and we have enjoyed working with the Hartree Centre Northern Ireland Hub. We look forward to working together again in the future.
Further Information
This work was completed as one of our data projects. Our projects are up to 12 weeks in duration. We will work alongside your business to scope your project, demonstrate proof of concept, and explore options to deploy it within your business. If you would like to learn more about the Hartree Centre Northern Ireland Hub and our programme, feel free to contact us at: info@hartreeni.uk
Our Impact on UK Industry and Society
The Hartree National Centre for Digital Innovation enables businesses to acquire the skills, knowledge and technical capability required to adopt digital technologies like super computing, data analytics, artificial intelligence (AI) and quantum computing. The SME Engagement Hubs play a key role in offering targeted support for business challenges in their region, creating value and generating economical and societal impact for the UK.
