An innovative approach to analyse point cloud – Machine learning powered Scan-to-BIM

As LIDAR Scanners and Depth Cameras become more affordable and widely used, Scan-to-BIM (Building Information Modelling) changes the way members of the AEC (Architecture, Engineering & Construction) and Construction industry are viewing projects and delivering them.

Scan-to-BIM has positively impacted two main aspects of daily efforts – saving on time and saving on money among many other advantages. 

Recently, an Ontario based office of a multinational organization originally commissioned to study and produce detailed floor plans of mixed facility office spaces succeeded in classifying and therefore transforming billions of individual measurements of a construction site dated back to the 1970s. Two pressing challenges in this project were the processing of data worth 100s of GB. The software would often crash owing to the size of the point cloud. The second was deciphering the multiple objects at the project site. Engineers were often seen manually mapping the walls, floors, ceilings, etc. The manual mapping process might consume 5 to 120 minutes, depending on the complexity and clearness of the source and amount of floors. This is just one example. Many engineers around the globe face the same or similar challenges. What if there was a way to identify each object through a common code? Wouldn’t this make a multi-level construction of multi-storeyed buildings much easier to view in the software? Like the Ontario, based office won’t be happy to participate in the emergence of digital assets to stay competitive in Industry 4.0?

Staying true to our DNA of innovation, transformation and customization with empathy at the core of our activities, we taught our machines to identify these objects and make Scan-to-BIM a level easier and convenient thus superior and smarter. At nCircle, we have developed a Machine Learning (ML) powered Scan-to-BIM framework, which automatically identifies and locates multiple objects in a point cloud and assigns different color code to different category of objects for easy visualization.  

The way we developed the Scan-to-BIM framework keeps you to follow the general workflow but reduces the cognitive load required for visualization. While loading the 3D point cloud model into the software of your interest as an intermediate stage we process it through the Scan-to-BIM framework which gives you the same model with the annotations in the “Viewing mode”. Our framework locates and identifies the various objects present in the point cloud model and categorizes them with different color codes.

For example, if the point cloud model was that of a house and you chose the living room; it will show you all the walls in blue color, the furniture like sofas and beds in Brown, the floor in Green and so forth. Want to understand it better with a demo? Here you go! Click on the object identification point cloud demo video below. 

Looks promising? Did it make Scan-to-BIM more simple? That’s exactly what we aim to do! This standard ML-powered Scan-to-BIM framework that can be customized and implemented as per your clients’ needs and workflows. 

Do you have any such other custom queries? contact us, We would love to solve it! At nCircle, our team is driven with the thought to empower passionate innovators in the AEC and Manufacturing industry to create impactful 3D engineering & construction solutions. Leveraging our domain expertise in CAD-BIM, we provide disruptive solutions that reduce time to market and meet business goals.