Point Cloud Processing
Point cloud processing transforms raw scanner output into usable spatial datasets through registration, cleaning, classification, and format conversion. Brisbane Point Cloud provides processing services for data captured by any scanner brand or platform.
Capabilities
What we deliver
- Multi-scan registration using target-based and cloud-to-cloud methods
- Noise filtering and outlier removal without losing fine detail
- Ground classification using progressive morphological filters
- Building, vegetation, and infrastructure class separation
- Format conversion between all major point cloud standards (E57, LAS, RCP, PTX, PTS)
- Coordinate transformation and datum conversion (MGA94 to MGA2020, local to geodetic)
- Point cloud decimation and tiling for large datasets
Equipment Used
Deliverables
- Registered and cleaned point cloud in specified format
- Classification report showing point counts per class
- Registration accuracy report (target residuals, cloud-to-cloud statistics)
- Tiled dataset for large-area projects (compatible with Potree, Cesium, or 3D Tiles)
- Quality assurance documentation with check-point analysis
Our Process
How we work
Data Ingest
We receive your raw scan data in any format from any scanner manufacturer. Data is inventoried and checked for completeness against the capture log.
Registration
Individual scans or strips are aligned into a unified coordinate system. We use target-based registration, cloud-to-cloud refinement (ICP), or trajectory-based methods depending on the data source. Registration accuracy is reported as RMS residuals.
Cleaning & Classification
Noise points, multipath artefacts, and scanner artifacts are removed. For terrain datasets, points are classified into ground, vegetation, building, and other categories using automated algorithms verified by manual inspection.
Export & QA
The processed dataset is exported in your required format and coordinate system. Independent check-point analysis verifies positional accuracy. A QA report accompanies the final deliverable.
FAQs
Common questions
Can you process data from scanners you did not operate?
Yes. We process point cloud data from any scanner brand or platform: Leica (RTC360, P-series, BLK), FARO (Focus, Orbis), Trimble (X7, X9, TX8), Riegl, Z+F, NavVis, GeoSLAM, Matterport, and UAV LiDAR systems including DJI, Velodyne, and Livox. We work with raw proprietary formats and can register data from multiple scanner types into a single unified point cloud. If your field team captured the data but lacks processing capability, we can take it from raw scans through to final deliverable.
What is point cloud registration and why does it matter?
Registration is the process of aligning multiple individual scans into a common coordinate system. Each scan is captured from a different position, so they initially exist in their own local reference frame. Registration computes the transformation (rotation and translation) that best aligns overlapping scan areas. Poor registration results in doubled surfaces, blurred edges, and positional errors in the final dataset. We achieve registration accuracy within 2-3mm using a combination of surveyed targets and iterative closest point (ICP) refinement. Every project is delivered with a registration accuracy report.
What point cloud formats do you work with?
We read and write all major formats: E57 (open standard, stores intensity and colour), LAS/LAZ (geospatial standard, supports classification), RCP/RCS (Autodesk ecosystem), PTX/PTS (ASCII legacy formats), PLY (mesh and point cloud), FLS (FARO proprietary), and ZFS/ZFPRJ (Z+F proprietary). For web-based delivery, we produce 3D Tiles (Cesium), Potree format, and Entwine Point Tiles (EPT). We can convert between any of these formats and handle coordinate system transformations during conversion.
How do you handle very large point cloud datasets?
Large datasets (billions of points, terabytes of data) require spatial indexing, tiling, and often multi-resolution representations. We tile data into manageable spatial blocks (typically 50m or 100m tiles) so that software can load only the required area. For delivery to web viewers, we build octree-based level-of-detail structures that stream progressively. For desktop GIS or CAD use, we provide spatially indexed LAS/LAZ files that load efficiently in software like CloudCompare, RealWorks, or Virtual Surveyor. Our processing workstations have 256GB RAM and GPU acceleration for handling datasets up to 50 billion points.
What quality checks do you perform?
Our QA process includes: registration accuracy check (target residuals must be below specified threshold, typically 3mm for terrestrial data), visual inspection for gaps in coverage, noise analysis (stray points per cubic metre), classification verification (random sampling of classified areas), positional accuracy verification against independent survey check points, and format validation (file opens correctly in specified software). Every project ships with a QA report documenting these checks and their pass/fail status. If any metric fails, we re-process before delivery.
Get a quote for point cloud processing
Tell us about your project and we will provide a fixed-price proposal within one business day.