Scan-to-Map, Scan-to-BIM with Point Clouds, Engineered by AI.

Nova Spatium provides deep learning solutions and tools of 3D point clouds, from efficient labeling, semantic segmentation to final map/BIM products.

3D Maskster* is an AI-assisted point cloud labeling tool that enables users to perform efficient annotation on data points with self-defined classes.
*US Patent Application Pending
- 3D Maskster performs fast unsupervised segmentation on the point clouds or user-captured 2D scene, facilitating users undergoing pre-labeling and re-labeling.
- Users can perform labeling on the point cloud segments or the 2D segments instead of going after point by point. Our tool also provides segment merge, label fusion, and data cleaning before users export the resulting point clouds for training.

Nova Spatium has its in-house AI model to perform semantic segmentation of common road corridor objects from mobile mapping system (MMS) point clouds. The segmented point clouds can aid in extracting on-road surface objects and create digital map (GIS/CAD) products.
- Our AI solution toward MMS point cloud can help classify objects in common road corridor projects, including lamp posts, traffic signs, road markings, trees, utility lines, moving objects, etc. Once the MMS point clouds are segmented, our data processing workflow converts these objects into corresponding GIS features or CAD drawings.
- Alongside with these features, our workflow also adopt a large language model (LLM) and image recognition techniques to automatically classify road signs and extract textual features on the signs. These can help enrich the geospatial database of road mapping projects and minimize human-induced error.

Nova Spatium has developed an automatic workflow to convert LiDAR or photogrammetric based point clouds collected on underground utility pipework collected during open trench into corresponding building information models (BIMs).
- We have a well-trained deep learning model to classify point clouds of pipework with objects including pipes, elbows, joints, tees, landslide barriers, etc. Our workflow can reconstruct the pipework through cylindrical fitting and assign corresponding BIM objects of appurtenances under BIM platform, such as Autodesk Civil 3D.
- We also offer a side-product of quality checking to assess the spatial location of pipework. The tool can automatically recognize control targets placed on the pipework so that the extracted coordinates can be checked against with the measurements obtained by total station. All these help improve the efficiency of as-built 3D reconstruction.