LiDAR you can work with

How Omniris turns a LiDAR survey into a labelled 3D point cloud, terrain models and clearance measurements, lets anyone you choose open it in a web browser, and keeps it in the same secure evidence vault as your photos.

Summary

A LiDAR scanner measures a site with laser pulses. The result is a point cloud: millions, sometimes billions, of measured points, each with a precise position and height. It usually arrives as one or more LAS or LAZ files (the standard LiDAR file formats; LAZ is the compressed form). On their own these files are hard to open, hard to share and hard to measure from. Omniris makes them usable, and keeps them beside the photos, maps and 3D models of the same site.

Why a LiDAR file is hard to use

LiDAR is the most direct measurement a drone or aircraft can make of a site, and the hardest to hand to anyone else:

• It is too big to open. A corridor survey is often several gigabytes. Most people who need the answer have no software that reads it, and desktop tools must load the whole file before showing anything.
• The points are not sorted. A raw survey is just points. Ground, trees, buildings and wires have to be told apart before anything can be measured.
• Units and height references vary. One provider delivers metres, another feet. Heights may be measured from different starting surfaces, and the file does not always say which.
• It lives apart from everything else. The photos, maps, 3D model and inspection notes for the same site are somewhere else, with their own access rules and no shared record of who handled what.

Omniris treats a LiDAR survey as one more capture of a site: processed when you ask, viewable by anyone you share it with, and held under the same custody record as everything else.

From upload to results

Upload the files into a folder, choose Process LiDAR on that folder, and confirm. Before anything is charged, the dialog reads the start of each file and shows what it found: the number of points, the coordinate system (the map grid the positions are given in), and the price.

flowchart LR
  A[LAS / LAZ files
in a vault folder] --> B[Checked:
points, coordinate system]
  B --> C[Converted to metres]
  C --> D[Ground: the provider's,
or found from the points]
  D --> E[Sorted:
vegetation, buildings,
wires, towers]
  E --> F[Point cloud, terrain,
surface, canopy, contours]
  F --> G[Results folder
beside the source]
  

Your source files are never changed. Processing the same folder again replaces the earlier results and removes any earlier result the new run did not produce, so the folder never mixes two runs. If someone has unapplied corrections on the earlier point cloud, you are warned before they are lost.

What you get

The three height maps are GeoTIFFs, the standard format for map-aligned images, in a cloud-friendly layout. They open on the Omniris map, can be exported from Omniris, and open in any mapping (GIS) software. The volume, change-comparison and profile tools work on height maps from photo surveys; they do not take LiDAR height maps.

A viewer that streams

A point cloud opens in the same 3D viewer as your models: on the web, in the desktop application, and on a public share link.

The viewer never downloads the whole file. The file is arranged so that a rough picture of the whole survey comes first and finer detail follows. The viewer reads only the part in view, at the detail your screen can show, and reads more as you zoom in. How much work the viewer does depends on what is on screen, not on how big the file is.

Each option appears only when the point cloud holds the data for it. The ruler measures between points, in feet or metres. It is offered only when the point cloud's units are known to be metres, which is always true of processed results. If the scale is not known, the ruler is switched off and the reason shown, rather than giving a number that might be wrong.

Encrypted point clouds can be viewed by anyone permitted to decrypt them. Every opening is recorded in the file's custody history.

Classification

Classification means giving every point a label that says what it is. Omniris uses the standard classes of the LiDAR industry, so the labels mean the same thing in any other LiDAR software:

Wires and towers

Wire detection looks for thin, straight runs of points hanging in the air, well above the ground, and labels them as wires (conductors). The tall, slender structures the wires attach to are labelled as towers, including ones standing in forest. The clearance check (below) measures from the wires found this way.

Correcting the classification

No automatic classification is right everywhere. A flat roof is taken for ground, a stretch of cable for a branch, a parked truck for a building. What matters then is how quickly a person can put it right, and whether the correction can be trusted afterwards. Omniris includes an editor for that, inside the same 3D viewer.

You can limit a selection to points that currently have certain labels, or to a range of heights, and build it up by adding and subtracting shapes. Then give the selected points a standard class, or a class of your own.

What you see is what is saved

The browser never holds the whole survey at once, so an edit cannot be stored as a list of points. Instead it is stored as a rule: this area, these filters, this class. The viewer applies the rules to the points on screen. When you apply your edits, a processing job applies the same rules to every point in the survey. The two follow one specification, so what you saw on screen is what is saved.

Every job checks each new version against its source: the new copy must have the same number of points, in the same place, at the same precision, as the one it was made from. A finished job reports exactly how many points moved from each class to each other class.

A draft, then a version

• Your edits are a draft. They are saved as you work and are still there when you come back, with undo and redo. One person edits a point cloud at a time; anyone else who opens it sees the saved edits and who is making them.
• Applying makes a new version. The edits are written into a copy of the point cloud, filed in its own folder beside the original. The original is never changed, so a correction can always be compared with, or traced back to, what the survey first said.
• Each version is evidence in its own right. It has its own custody history, records which point cloud it was made from and who applied the edits, and can be shared, encrypted and placed on legal hold.
• Your own classes. Name and colour classes the standard list lacks (solar panel, scaffold, crane) and keep them in a personal library or one shared across your organization. The names are stored inside the point cloud file, so a shared link shows them too.

Terrain and clearance from the corrected classes

The terrain model is built from the ground points, the surface model from everything except noise, and clearance from wires and vegetation. Correcting classes therefore makes those results out of date. Regenerate rebuilds them from the classes as they are now. Until you do, any clearance results on an edited version are marked as measured before the edit, so they never look current when they are not.

A point cloud that has not been fully classified, such as one from a photo survey (which marks only the ground) or one you uploaded, can be classified automatically first, at the normal processing price, and then corrected.

Clearance to overhead lines

This is an optional check for surveys of power-line corridors. It is off unless you tick it when starting the job. A survey of a building site or a stockpile is not checked for power lines, and its results never mention them.

When it is on, the job joins the laser hits on each detected wire into a continuous line. It then measures the distance from every nearby vegetation point to that line, not just to the nearest hit on the wire, so a gap between hits cannot hide a tree standing in it.

You set the distances for each job, in feet, to match your own standard. Results appear three ways:

• the points coloured by band in the 3D viewer;
• a list of places where vegetation is too close, each with the closest distance, the height of the vegetation and how far the patch runs along the line; selecting one takes the view straight to it;
• a results table giving the latitude and longitude of each place.

Three rules keep the results honest:

• As flown. Distances are straight-line distances at the moment of capture. Wire sag under load or heat, wind and vegetation growth are not modelled.
• Never "clear" by default. If no line is detected, the result is "not assessed", with the reason. Finding no wires says nothing about the vegetation.
• How well the line was seen is recorded. The total length of wire detected is shown with the results, and the job's processing record lists each wire's length, how densely it was hit and its largest gap, so you can judge how well the line was captured.

Coordinate systems and heights

Position is where LiDAR results most often go quietly wrong, so Omniris is explicit about it.

• Any common map grid. State plane in feet, UTM in metres, national grids, or plain latitude and longitude. A survey in latitude and longitude is placed on the UTM grid for its area.
• Results are always in metres. Feet are a display choice. Every result is stored in metres so that distances and areas taken from it are right.
• No guessing. If a file does not say which coordinate system it uses, Omniris does not guess. You are asked for its EPSG code (the standard number that identifies a coordinate system) before the job can start.
• The height reference is carried, not changed. Heights are always measured from some reference level, such as mean sea level. If the file says which, the results say the same. If it does not, you can state it; your answer is recorded as yours and never written into the survey file. If nobody knows, the results say "not known".
• Mixed sets are refused up front. Files in different coordinate systems cannot be combined meaningfully. The dialog says so, naming the file, before the job starts.

LiDAR and imagery together

A point cloud is most useful next to the other captures of the same place.

• On the model it came from. A 3D model built from photos has its own point cloud. One button in the viewer shows the points over the model, exactly where the model was built from them, so you can check the model against its source.
• On a model from another flight. A LiDAR survey can be shown over a 3D model from a different flight when the height references of both are known and can be related (see the limits below).
• On the map. The terrain, surface and canopy height maps open on the 2D map with your orthomosaics, site boundaries and pins.
• In conversation. Orion, the Omniris assistant, can answer questions about your LiDAR jobs: which surveys have been processed, how many points, what they cost, and which found vegetation inside the critical distance.

The same evidence vault

LiDAR results are not kept in a separate system. They are files in the vault, with everything that brings:

• Custody. Each result is recorded in the tamper-evident custody record as an output of the job that made it, and can be covered by a signed custody certificate.
• Legal holds and retention. Results under a legal hold or retention policy are never replaced. If a folder's earlier results are protected, a new run on it is refused before it starts, with the reason, and nothing is charged.
• Sharing. Share the results folder with a colleague or a client, or by public link with an optional password and expiry date. The recipient opens the point cloud in their browser.
• Encryption. Surveys in an encrypted folder can be processed, and their results are encrypted as they arrive. The working copy a job makes of your survey is deleted when the job ends, whether it succeeded, failed or was cancelled.
• Access. Every read of a point cloud is checked against the file's permissions, for signed-in users and share links alike. Revoking a share cuts access at once.

The companion paper, Evidence you can prove, describes the custody record, certificates, holds and retention in full.

Price

Processing is priced by the size of the survey: one credit per million points, rounded up, with a minimum of one credit. The dialog shows the price for the files you selected before you start.

• You pay for what was read. The price shown is set aside when you start, based on the point count each file states about itself, and settled to the number of points the job actually read.
• A wrong file header cannot run away with the bill. A file that turns out to hold more than twice the points it stated is stopped before the expensive work, and refunded.
• Encrypted files are priced when opened. The point count of an encrypted file is not known until the job decrypts it, so its points are charged when the job finishes, at the same rate.
• Failure costs nothing. A job that fails, is refused or is cancelled is refunded in full.
• Viewing is included. Opening, measuring and sharing point clouds carries no processing charge.
• Correcting is not charged for at present. Applying classification edits and regenerating terrain products carry no processing charge. Each version is a full copy of the point cloud and counts toward your storage.

Where it runs

Viewing point clouds needs no extra processing power and is part of the 3D viewer everywhere Omniris runs, alongside LiDAR processing and the classification editor.

Processing needs no graphics card. On-premises sites run it with the same processing software they already have for 3D models, so there is nothing more to install. Once started, a job is carried through by Omniris itself, not by the browser that started it: you can close the page, and a running job survives a restart of the service.

What the results claim, and what they don't

A measurement is only as good as the honesty of its limits.

• Classification is automatic. It follows fixed rules, gives the same answer every time, and can be wrong.
• Applying edits is not instant. The editor shows the effect of an edit at once, but saving a version rewrites the whole point cloud on a processing machine: allow about ten minutes even for a small correction. One person edits a point cloud at a time.
• A shape drawn on screen selects all the way through the point cloud, including points hidden behind what you see. Use the class filter or the section view to control how deep it reaches. The count of selected points shown while editing covers only the points loaded on screen; the exact count is reported when the edits are applied.
• A 3D model's point cloud has no coordinate system, so it can be classified by hand but not automatically, and terrain models cannot be rebuilt from it.
• Clearance is as flown and is not a compliance determination. "Not assessed" means no line was detected, not that there is no line.
• Heights in the results are never moved to another reference. A survey is shown on a 3D model from a different flight only when both height references are known and can be related, as heights above the GPS ellipsoid and above mean sea level (EGM96) can. Heights above NAVD88, the common North American reference, cannot yet be related to others, so such a survey is viewed and processed on its own.
• Job size limits are cautious estimates. Processing holds the whole survey in memory, and the limits above are estimates of how much one job can hold.
• Encrypted point clouds have a size limit of 1 GB for viewing and processing. Larger encrypted point clouds are refused, with the reason.
• The report is on screen. A job's Results view shows what was read, the references, the class counts and the clearance results. It is not issued as a signed document; the custody certificate covers the result files themselves.
• Mobile. In the mobile app's viewer, a point cloud can be shown over a 3D model, with fewer points drawn. Starting LiDAR jobs and correcting classification are done on the web or in the desktop application; shared links are for viewing only.
• Availability. The complete path, from upload through processing and classification editing to shared viewing, is in production.

Where it matters

Standards

Inputs and results use open, published formats, so your data is never locked into Omniris.

Glossary

Canopy height model

A height map of whatever stands on the ground: the surface model minus the terrain model.

Classification

The label on each point saying what it is: ground, vegetation, building, wire, tower.

Cloud Optimized Point Cloud (COPC)

A LAZ file arranged so that a viewer can read a rough version of the whole survey first and finer detail when needed, without reading the whole file.

Conductor

A wire of an overhead power line.

Coordinate system

The map grid a survey's positions are given in, such as a state plane or UTM zone. Each has a standard number, its EPSG code.

GeoTIFF

An image file that also records exactly where on Earth it lies; used here for height maps.

Height above ground

How far a point is above the ground directly beneath it, as opposed to its elevation.

LAS / LAZ

The standard file formats for LiDAR points. LAZ is LAS, compressed.

Point cloud

A survey stored as millions of individually measured points, each with a position and height.

Surface model (DSM)

The elevation of the highest thing at each spot: ground, vegetation or structure.

Terrain model (DTM)

The elevation of the bare ground, with vegetation and structures removed.

Vertical reference

What heights are measured from: for example the ellipsoid used by satellite positioning (GPS), or a model of mean sea level. The two can differ by dozens of feet (tens of metres).