TL;DR: A 3D point cloud survey captures millions of precisely measured, georeferenced points across a site or structure to build a dense digital replica of real-world geometry. Using UAV photogrammetry, drone and terrestrial LiDAR, and survey-grade ground control, ISS delivers point clouds with absolute accuracies of 10–50 mm — the foundation for volumetrics, as-built models, clash detection, and deformation monitoring across Australian mining, processing, and construction sites.
Key takeaways
- A well-controlled UAV point cloud achieves 10–30 mm horizontal and 20–50 mm vertical absolute accuracy; terrestrial laser scanning reaches 2–6 mm at typical industrial ranges, with point densities from a few hundred to several thousand points per square metre.
- Point density (GSD or point spacing) and absolute accuracy are independent: a dense cloud is not necessarily an accurate one. Both must be specified and verified against independent check points before the data is fit for engineering use.
- The deliverable is rarely the raw cloud. Most ISS clients need a registered, classified, georeferenced cloud in LAS/LAZ or E57, plus derived products — DTMs, contours, volumes, or scan-to-CAD/BIM models.
- Mining and minerals processing are the dominant Australian users, for stockpile volumetrics, pit and dump models, plant as-builts, and shutdown clash detection — typically captured monthly for operations and per-event for shutdowns and turnarounds.
- Cost drivers are site area, required accuracy, point density, acquisition method (UAV vs. terrestrial vs. mobile), processing depth (raw cloud vs. modelled deliverable), and access or airspace constraints.
Table of contents
- What is 3D point clouds?
- Why point cloud quality matters
- The 3D point cloud survey process
- Capture methods and equipment
- Accuracy, density and standards
- When you need a 3D point cloud survey
- Deliverables and formats
- Cost factors
- How ISS delivers point cloud surveys
- Frequently asked questions
What is 3D point clouds?
A 3D point cloud is a dataset of discrete points, each defined by X, Y and Z coordinates — and usually colour (RGB) and intensity values — that together describe the surface geometry of an object, structure, or landscape. A single survey routinely produces tens of millions to several billion points. Viewed together, those points form a measurable digital twin of the real world that an engineer can rotate, section, and interrogate at the desktop.
A 3D point cloud survey is the process of capturing that data with survey-grade instruments, registering and georeferencing it to a known coordinate system, and verifying its accuracy so the cloud can be relied upon for measurement and design. The points are captured either by LiDAR — a laser scanner timing the return of emitted pulses to measure range directly — or by photogrammetry, where overlapping imagery is processed with Structure-from-Motion algorithms to triangulate point positions indirectly.
The problem it solves is the gap between the design intent on a drawing and the messy reality on site. Brownfield plants are never built exactly as drawn, and rarely maintained as drawn. A point cloud captures what is actually there — every pipe, beam, conveyor, and corroded flange — to within millimetres, so that retrofit design, volume reconciliation, and structural assessment proceed from measured fact rather than assumption.
Key point: Point density and absolute accuracy are different properties, and operators conflate them at their peril. A consumer drone can produce a visually dense, beautiful-looking cloud that is 300 mm out of position because it lacks proper ground control. Density describes detail; accuracy describes truth. A survey-grade point cloud needs both, and only the second can be verified with independent check points.
Why point cloud quality matters
The cost of a poor point cloud is not paid at capture — it is paid downstream, when the data feeds a decision. A stockpile cloud that is 5% high inflates a $20 million ore inventory by a million dollars on the balance sheet. An as-built cloud that is 40 mm out of position lets a fabricated pipe spool arrive on site for a shutdown and not fit, costing a fitting crew a shift and a critical-path window measured at $50,000–$150,000 per hour in lost production at a large processing plant.
The value of a good one is measured the same way. A correctly controlled cloud captured during a shutdown lets engineers model new tie-ins, run clash detection against the existing plant, and fabricate offsite with confidence — converting field rework into shop work and compressing the live work window. On a greenfield earthworks job, a fortnightly drone cloud gives the principal an independent, comprehensive cut-and-fill record that resolves payment claims on measured volume rather than estimate.
| Failure mode | Root cause | Downstream cost |
|---|---|---|
| Stockpile volume error | Weak ground control, poor edge definition | Misstated inventory, failed reconciliation |
| Spool / module won't fit | Cloud absolute accuracy not verified | Lost shutdown hours, emergency rework |
| Missed clash in retrofit | Occlusion, insufficient scan coverage | Redesign, schedule slip |
| False deformation alarm | Registration drift between epochs | Wasted investigation, lost confidence |
Warning signs that your point cloud data cannot be trusted: no independent check-point report, no stated coordinate system or datum, no measurement uncertainty figures, and a single-method capture (drone only) of a congested plant where occlusion is unavoidable. Each is a reason to question whether the cloud is fit for the decision it is supporting.
The 3D point cloud survey process
ISS follows a consistent eight-step workflow regardless of capture method. A typical small site is captured in a day and delivered as a registered cloud within 2–3 business days; large plant or LiDAR jobs scale from there. The discipline is the same whether the instrument is a drone or a terrestrial scanner — the accuracy lives in the control and the verification, not the sensor alone.
Step 1: Scope and methodology
ISS defines the deliverable first and the method second. We confirm the required absolute accuracy, point density, coordinate system (MGA2020 zone or a local plant grid), the area and structures to be captured, occlusion risk, and the final output — raw cloud, DTM, or a modelled CAD/BIM product. This determines whether the job is UAV, terrestrial, mobile, or a hybrid.
Step 2: Control establishment
A stable survey control network is established or verified with a Leica or Trimble GNSS receiver and total station, tied to MGA2020 and AHD. For UAV work, ground control points (GCPs) are set out and surveyed across the site; for terrestrial scanning, registration targets or checkerboards are placed between scan positions. Control is the single largest determinant of absolute accuracy.
Step 3: Data capture
The cloud is acquired by drone flight, scanner setups, or mobile traverse. UAV missions fly planned grids at 80% forward and 70% side overlap; terrestrial scans are positioned for line-of-sight coverage of every required surface. Capture metadata — instrument, weather, operator, date, and overlap — is recorded for the quality report.
Step 4: Registration
Individual scans or image blocks are registered into a single coherent cloud. Terrestrial scans are aligned using targets and cloud-to-cloud matching, with registration residuals reported per setup. Photogrammetric blocks are bundle-adjusted in Pix4D, Agisoft Metashape, or RealityCapture, anchoring the model to the surveyed GCPs.
Step 5: Georeferencing
The registered cloud is transformed into the project coordinate system and height datum so every point carries a real-world position. This is what allows the cloud to be compared against design surfaces, prior epochs, and other datasets — and is the step most often skipped by low-cost providers.
Step 6: Cleaning and classification
Noise, moving objects (vehicles, personnel, dust), and stray returns are removed. Where required, points are classified — ground versus non-ground for a bare-earth DTM, or by feature for structural work. Classification quality directly determines the accuracy of any derived contour or volume.
Step 7: Verification
Independent check points — surveyed but withheld from the registration — are compared against the cloud to quantify achieved accuracy. ISS reports the RMS error in horizontal and vertical, not a marketing claim. A cloud without a check-point report has an unknown accuracy.
Step 8: Delivery and modelling
The verified cloud is delivered in the agreed format (LAS, LAZ, E57, RCP), accompanied by a quality report. Where commissioned, ISS extracts derived products: DTMs, contours, volume calculations, or a scan-to-CAD/BIM model built to the required level of detail.
Capture methods and equipment
No single instrument is right for every point cloud. The choice is driven by area, accuracy, density, vegetation, and access. ISS operates the major survey-grade platforms and selects — or combines — them per job.
UAV photogrammetry
A drone captures overlapping imagery that is processed into a dense, full-colour point cloud. It is the fastest method for open sites and the standard for stockpiles, pits, and earthworks. With RTK/PPK positioning and ground control, photogrammetry delivers 10–30 mm horizontal accuracy over hundreds of points per square metre, with rich RGB that aids interpretation.
UAV LiDAR
A laser scanner mounted on the UAV measures range directly and, critically, penetrates vegetation canopy through multiple returns to capture the ground surface beneath. Drone LiDAR is the right tool for vegetated terrain, corridor mapping, and any site where photogrammetry cannot see the ground. It operates in overcast and low light, with vertical accuracies of 20–50 mm.
Terrestrial laser scanning (TLS)
A tripod-mounted scanner — typically a Leica RTC360 or Trimble X-series — captures up to two million points per second at 2–6 mm range accuracy. TLS is the method of choice for congested plant: process structures, conveyors, pipe racks, kiln and mill interiors, and any environment where millimetre fidelity and dense coverage of vertical surfaces are needed. Multiple setups overcome occlusion.
Mobile laser scanning (MLS)
A vehicle- or backpack-mounted scanner captures clouds while moving, combining LiDAR with an inertial navigation system. MLS suits large, linear, or trafficable assets — haul roads, long conveyors, tunnels, and stockyards — where setting up static scans would be prohibitively slow.
| Method | Absolute accuracy | Point density | Best for |
|---|---|---|---|
| UAV photogrammetry | 10–30 mm | High (RGB) | Open sites, stockpiles, earthworks |
| UAV LiDAR | 20–50 mm | High | Vegetated terrain, corridors |
| Terrestrial scanning | 2–6 mm | Very high | Congested plant, structures, interiors |
| Mobile scanning | 15–50 mm | High | Roads, tunnels, linear assets |
Key point: On a real brownfield plant, the best answer is usually hybrid — a drone cloud for the open areas and roofs, terrestrial scans for the congested process structures, all merged into one georeferenced dataset. A provider that only owns drones will give you a drone-only answer, and the occluded plant interior will be the part you actually needed.
Accuracy, density and standards
ISS reports point cloud quality against two independent metrics — absolute accuracy and point spacing — and verifies the first with withheld check points rather than software-reported confidence. UAV surveys are georeferenced to MGA2020 and AHD; terrestrial and plant work is tied to a local plant grid where the client maintains one.
| Parameter | UAV photogrammetry | UAV LiDAR | Terrestrial scanning |
|---|---|---|---|
| Horizontal accuracy | ±10–30 mm | ±20–50 mm | ±2–6 mm |
| Vertical accuracy | ±20–50 mm | ±20–50 mm | ±2–6 mm |
| Typical point density | 200–1,000 pts/m² | 100–500 pts/m² | 1,000–10,000 pts/m² |
| Coordinate framework | MGA2020 / AHD | MGA2020 / AHD | MGA2020 or plant grid |
Accuracy is governed by ground control, sensor calibration, geometry, and processing — not by the headline specification of the instrument alone. ISS follows the principle, common to engineering survey practice, that ground control should be two to three times more accurate than the target survey accuracy, and that achieved accuracy is meaningless without independent verification. Every point cloud deliverable carries a quality report stating the coordinate system, datum, check-point RMS error, registration residuals, and point density. Survey instruments are calibrated to manufacturer schedules, and control is GNSS-derived and traceable to the national datum.
When you need a 3D point cloud survey
A 3D point cloud survey is warranted whenever a decision depends on accurate, comprehensive measurement of as-is geometry that is too complex, large, or hazardous to capture point by point.
Mining and minerals processing
The dominant Australian use case. Stockpile and pit volumetrics for monthly reconciliation and inventory; waste-dump and tailings models for compliance and capacity; plant as-builts for retrofit and expansion; and shutdown clash detection where a cloud of the existing plant lets new equipment be designed and pre-fabricated to fit. Operations capture clouds monthly; shutdowns and turnarounds are captured per event.
Construction and civil
Earthworks cut-and-fill measurement, progress documentation, control and conformance checking, and as-constructed records for handover. A point cloud gives the principal an independent, comprehensive record that resolves payment and quality disputes on measured fact.
Brownfield engineering and retrofit
Scan-to-CAD and scan-to-BIM modelling of existing plant for tie-in design, equipment replacement, pipe routing, and structural assessment. Designing a modification against a verified cloud — rather than decades-old, drifted drawings — is the single most effective way to eliminate field rework.
⚠️ Watch out: Operators commonly assume an old set of as-built drawings is good enough for a retrofit. On a plant that has been modified, patched, and corroded over twenty years, the drawings describe a building that no longer exists. The first clash discovered in the field during a live shutdown costs more than the entire point cloud survey would have.
Deliverables and formats
The point cloud itself is the starting point, not usually the end product. ISS delivers the verified cloud plus whatever derived products the project requires.
| Deliverable | Description | Typical format |
|---|---|---|
| Registered point cloud | Cleaned, georeferenced master cloud | LAS, LAZ, E57, RCP |
| Quality report | Coordinate system, datum, check-point RMS, density | |
| Digital terrain model | Bare-earth surface for design and volumes | LandXML, GeoTIFF |
| Contours | Elevation lines at specified interval | DWG, DGN |
| Volume report | Calculated volumes with methodology | PDF + surfaces |
| Scan-to-CAD model | 2D/3D as-built drawings from the cloud | DWG, DGN |
| Scan-to-BIM model | Intelligent 3D model at agreed level of detail | RVT, IFC |
| Textured mesh | Visual 3D model for presentation | OBJ, FBX |
Format and level of detail are agreed at scoping. A monthly stockpile job needs a volume report and surfaces; a retrofit needs a modelled deliverable in the client's CAD or BIM platform. ISS delivers in the format that drops straight into the client's downstream workflow.
Cost factors
Point cloud survey pricing is project-specific. ISS provides a fixed-price quote after a brief scoping discussion. The main cost drivers are summarised below.
| Factor | Effect on cost | Typical range |
|---|---|---|
| Site area / structure complexity | More area and congestion means more capture and processing | Baseline to +100% |
| Required accuracy | Higher accuracy needs more control and verification | +10–30% |
| Capture method | TLS and hybrid jobs cost more than open-site UAV | UAV cloud $2,000–$7,000; TLS plant $6,000–$25,000+ |
| Processing depth | Modelled scan-to-CAD/BIM costs more than a raw cloud | +50–200% |
| Access / airspace | Confined space, height access, controlled airspace, live plant | +10–40% |
| Travel and accommodation | Remote mine sites outside major centres | At cost |
ROI context: A point cloud captured during a shutdown costs a fraction of a single hour of unplanned downtime at a large processing plant — and the clash detection it enables routinely prevents the field rework that consumes that downtime. For mining inventory, a cloud that tightens stockpile accuracy from 5% to 2% on a $20 million pile removes $600,000 of balance-sheet uncertainty for the price of a day's survey. The payback is almost always measured against a single avoided event.
How ISS delivers point cloud surveys
ISS is an independent industrial surveying firm working across Australian mining, minerals processing, and construction. We are CASA-certified for commercial UAV operations under a current Remote Operator Certificate with licensed pilots and registered, insured aircraft — so the regulatory burden of drone capture sits with us, not the client.
Our advantage on point cloud work is method-agnostic capability. Because we own and operate UAV photogrammetry, drone and terrestrial LiDAR, and mobile scanning, we scope each job to the right tool — or combination of tools — rather than forcing the site to fit a single instrument. Every cloud is georeferenced to MGA2020 and AHD or the client's plant grid, verified with independent check points, and delivered with a quality report that states the achieved accuracy rather than claiming it. Where the deliverable is a model, our processing team builds scan-to-CAD and scan-to-BIM products to the agreed level of detail, ready to drop into the client's design environment. We work to shutdown and turnaround schedules across remote sites, and deliver registered clouds within 2–3 business days of capture for standard jobs.
Frequently asked questions
How accurate is a 3D point cloud survey?
It depends on the capture method and the control. A well-controlled UAV point cloud achieves 10–30 mm horizontal and 20–50 mm vertical absolute accuracy. Terrestrial laser scanning reaches 2–6 mm at typical industrial ranges. ISS verifies achieved accuracy with independent check points withheld from processing and reports the RMS error — accuracy that is claimed but not verified should be treated with caution.
What is the difference between point cloud density and accuracy?
Density (points per square metre or point spacing) describes how much detail the cloud captures. Accuracy describes how close each point is to its true real-world position. They are independent: a consumer drone can produce a dense cloud that is hundreds of millimetres out of position because it lacks ground control. A survey-grade deliverable needs both, and only accuracy can be independently verified.
Which is better for point clouds — drone or laser scanning?
Neither, universally. UAV photogrammetry and LiDAR are fastest for open sites, stockpiles, and earthworks. Terrestrial laser scanning is superior for congested plant, structures, and interiors where millimetre fidelity and dense vertical coverage are needed. Most brownfield plant jobs are best served by a hybrid of drone and terrestrial capture merged into one georeferenced cloud.
What formats does ISS deliver point cloud data in?
The registered cloud is delivered in LAS, LAZ, E57, or RCP. Derived products include LandXML and GeoTIFF surfaces, DWG/DGN contours and CAD models, RVT/IFC BIM models, and PDF volume and quality reports. ISS delivers in whatever format drops straight into the client's downstream CAD, BIM, or mine-planning workflow.
Can a point cloud be captured while the plant is operating?
Often, yes. UAV and terrestrial scanning are non-contact and can frequently be performed around live plant under the site's permit-to-work and exclusion-zone controls. However, occlusion from operating equipment, dust, and personnel reduces coverage and accuracy. For congested process structures and clash-critical retrofit work, capture during a shutdown — with the plant accessible and clear — delivers a substantially better cloud.
How long does a point cloud survey take?
Field time is usually hours, not days. A small open site is flown in under an hour; a congested plant may need a day or more of terrestrial setups. Processing, registration, and verification typically take 2–3 business days for a standard cloud. Modelled scan-to-CAD or scan-to-BIM deliverables add time proportional to the level of detail required.
What to do next
A 3D point cloud is only as useful as it is accurate, and accuracy is a function of control and verification — not the price of the drone. The data you need to design, reconcile, or model from measured reality is achievable within days of a site visit.
- Define your deliverable — Do you need a raw cloud, a volume report, or a modelled scan-to-CAD/BIM product, and in what coordinate system and format?
- Identify the constraint — Open site, congested plant, vegetation, live operations, or a shutdown window? This determines the right capture method.
- Call us on 0407 057 015 — Discuss your site with a surveyor who will recommend the appropriate method, confirm achievable accuracy, and provide a fixed-price quotation.
Industrial Spatial Solutions delivers 3D point cloud surveys across Australia using UAV photogrammetry, drone and terrestrial LiDAR, and mobile scanning — every cloud georeferenced, independently verified, and delivered in your required format. To request a quote, call 0407 057 015 or visit industrialspatial.com.
Related reading: UAV and aerial surveys guide, 3D laser scanning for industrial plants, Volumetric surveying for stockpiles and earthworks