Proving Pointly's scalability: Classification of an entire major city
published: 31.07.2023
In the era of digital transformation, cities around the world are harnessing the power of technology to create detailed and up-to-date digital twins. These virtual replicas serve various use cases, from urban planning and infrastructure management to improving the overall quality of life for residents. The city of Munich, known for its progressive approach to smart city initiatives, has been actively involved in performing yearly surveys to develop a comprehensive digital twin of its city. In order to process the aerial datasets, they reached out to us. Therefore, we were given the opportunity to support them in this endeavor. Have you ever wondered how scalable our Pointly platform is and if it's possible to process an entire city? Thanks to this cooperation you will get a clear answer: Yes! Our platform has been instrumental in classifying the aerial lidar datasets of the city of Munich contributing to the goal of creating a digital twin of the city, that can be used for various use cases.
Training the model of the Munich dataset
To train the model used in the Pointly platform, a significant amount of training point cloud data was generated by manually classifying approximately 2.5 square kilometers of the Munich aerial dataset. We were supported in this task by our trusted partners at People for AI. As a baseline model Pointly's Airborne Laser Scanning (ALS) Standard Classifier was , providing a solid foundation for the model training. In addition, the trained model was further enhanced by incorporating RGB values and optimized specifically for the Leica CityMapper-2 scanner model used for the survey. This approach ensured that the classification results were tailored to the unique characteristics of Munich's urban scenes.
Semi-automatic annotations for the training data were performed in the Pointly platform
Classification of the point cloud dataset
The point cloud classification process performed using the Pointly platform was a remarkable feat in terms of efficiency and scale. The entire dataset encompassing the city of Munich was processed, which involved handling a significant amount of data. The processing duration spanned about 5 days, showcasing our platform's ability to handle large-scale datasets effectively without interruptions.
During the classification process of the aerial datasets of Munich, over 2370 files were processed, comprising a staggering 84 billion points. Leveraging the power of Pointly's API, the data was fetched from an Azure blob storage, allowing for seamless integration with cloud-based storage infrastructure. The classification results were then saved back to the same storage, creating a streamlined workflow that minimized effort and ensured easy access to the processed data. The collaboration was a complete success, and the city of Munich is now able to leverage the classified dataset for various use cases enhancing their digital twin.
The trained classifier delivers a consistent classification accuracy for the full dataset from the city of Munich
Successful processing of large amounts of data in a short time
This project perfectly demonstrates the Pointly platform’s scalability and ability to process an entire city's worth of data to capture large-scale urban environments. By leveraging advanced algorithms and cloud computing resources, Pointly enables efficient data processing and classification, making it suitable for handling complex projects such as creating digital twins of entire cities.
The processed dataset can be used for various use cases and contributes to the creation fo a digital twin for the city of Munich
Do you have a similar project yourself or something completely different and you would like to use our Pointly platform? Or do you have any questions about the classification and ranking options? Please feel free to contact us. Our experts will provide you with comprehensive advice so that you achieve the best possible results. If you are curious about the latest Pointly news or new features, please subscribe to our newsletter or follow us on LinkedIn.
Pointly – Point out what matters.
Pointly is an intelligent, cloud-based software solution to manage and classify 3D point clouds – faster and more precise than ever before.
With Pointly Services you can get on-demand advanced 3D point cloud solutions tailormade for you.
To whom it may concern,
I hope this email finds you well. I recently came across the impressive work undertaken by Pointly in the city of Munich for the classification of aerial lidar datasets and the creation of a comprehensive digital twin. The success of processing the entire dataset of a major city is certainly noteworthy, and it has piqued our interest in exploring the capabilities of the Pointly platform for a similar project.
As we embark on our own venture involving the creation of a digital twin, we are keen to understand more about that. To gain a deeper understanding, we have a few questions:
1- As you mention, the model was optimized specifically for the Leica CityMapper-2 scanner. What was the point density used in the survey, and how does Pointly accommodate variations in point density for different survey scanners?
2- Was a Near Infrared (NIR) band added to the coloring of the point cloud during the classification process? If so, could you provide more details on how this feature enhances the classification results?
Thank you for your time, and we look forward to hearing from you soon.
Best regards,
Jose
Dear Jose,
Many Thanks for your message. I have sent you an email with a detailed reply to your questions.
Best Regards,
Sid