LAS Point Cloud Format
Points clouds are a category of GIS data.
The most common format for exchanging points clouds is LAS. Maintained by
ASPRS , the format is currently at
(it can be hard to find on the ASPRS website). There are no other
comparable formats that I am aware of; and if there were other formats, they
would contain essentially the same information.
LAS projects can be very large (many GB are very common), and
benefit greatly from compression and tiling.
- Specialized compressors that understand
the structure of the LAS format and the common data characteristics can
significantly outperform general purpose compression algorithms.
- LAZ format uses optimized
compression and is open source.
- zLAS is a proprietary ESRI format which has not gained wide acceptance
by government mapping agencies which tend to prefer open formats.
- Tiles allow two significant advantages
- After reading the file header, software knows immediately if the
tile is on the current map. Because storage within a LAS file can
have any order, it must be processsed sequentially, a reflection at a
time. There are schemes to index the files, which can
improve this, but they use the initial determination if there is any
relevant data in the tiles.
- Tiles can be processed in parallel for many operations, speeding up
the display or analysis by spreading it among several processors or
cores. The software must be explicitly written for
The LAS file contains a number of fields for each point that can be useful
for analysis and for display:
- x,y coordinates, most often in the UTM projection
- z, the elevation
- the intensity of the returned pulse. This looks like a grayscale
image, with the caveat that includes a single frequency of light, which may
most often be NIR. There is no standard for the range or scaling of
these values, and they may not be corrected for range; despite these
caveats, the intensity display can be very informative.
- an RGB value, merged from a camera flown with the laser scanner
- the return number, and the total number of returns from the pulse. The LAS format allows multiple returns per transmitted pulse, but typically the
vast majority will be the first and only return. When there are multiple returns from a
pulse, only the last could be on the ground, but even the last return could be
in the vegetation or features like power lines. The older LAS 1.2 and 1.3
could handle 5 returns per pulse, while the new 1.4 can handle up to 15.
- the scan angle, which indicates how far from nadir the scanner was
pointed; this is typically up to about 20-25 degrees, positive and negative
depending on which side of nadir. Away from nadir, smooth water
surfaces will act like mirrors and there will be no returned energy.
- Overlap points, where two flight lines covered the same area.
- The LAS format allows distribution of the full waveform data, but that
option is rarely used in practice. The resulting file sizes greatly
increase the data storage, and few ultimate end users want or need that
The LAS file has a header with metadata about the file.
- It is supposed to have the datum/projection information embedded,
but it may be missing.
- It has the elevation range, but usually has elevations that are both
excessively high and excessively low. When used to color the
elevations, this leads to a very compressed color scale. There is a
sense in the lidar community that you should never delete returns, but just
code them as noise (the older LAS versions had a classification code for low
noise, and 1.4 added high noise), but this does not help in determining the
true elevatioin range. I think the never delete a point may make sense
for the data producers, but not the end users.
Lidar data issues
- Missing data
- Data holidays: flight path did not allow data collection. For
the fastest data collection, you want to minimize overlap, but then if
the navigation is off, or the topography changes (the closer to the
ground the sensor flies, the narrow the swath), you miss points.
It can be very expensive to go back to cover the voids, so a little
extra overlap might be good insurance.
- Very low reflectivity (e.g. dark asphalt)
- Specular reflections from surfaces like water where the laser
hits the surface at an angle
- High points: caused by reflections from transient objects like birds
which are above the ground and not a part of the landscape
- Low points: multipath reflections which bounce off buildings will
increase the TWTT and place the point incorrrectly; this is similiar to the
problems of GPS in dense urban jungles, where
the GPS positions are often displaced significantly, and straight travel
paths show a great deal of noise.
Lidar Point Cloud Data Sources
last revision 11/16/2017