Removing Low Noise in LiDAR Points with Median Ground Surface

Recently a user of LAStools asked a question in our user forum about how to classify LiDAR data that contains lots of low noise. A sample screen shot of the user’s failed attempt to correctly classify the noise using lasnoise and the ground with lasground is shown below: red points are noise, brown points are ground, and grey points are unclassified. In this article we show how to remove this low noise using a temporary ground surface that we construct from a subset of points at a certain elevation percentile. You can follow along by downloading the data and the sequence of command lines used.

example of miss-classified low noise points: ground points (brown) below ground

Download the LiDAR data set that was apparently flown with a RIEGL “crossfire” Q1560. You can also download the command line sequence here. We first run lasinfo with option ‘-compute_density’ (or ‘-cd’ for short) to get a rough idea about the last return density which is quite high with an average of over 31 last returns per square meter. We then use lasthin to classify one last return per square meter with the temporary classification code 8, namely the one whose elevation is closest to the 20th percentile per 1 meter by 1 meter grid cell. We then repeat this command line for the 30th, 40th, 50th percentile modifying the command line accordingly. You will need LAStools version 190819 (or newer) as options ‘-ignore_first_of_many’ and ‘-ignore_intermediate’ were just added to lasthin this weekend.

lasthin -i crossfire.laz ^
        -ignore_first_of_many -ignore_intermediate ^
        -step 1 ^
        -percentile 20 15 ^
        -classify_as 8 ^
        -odix _p20 -olaz

Below you see the resulting subset of points marked with the temporary classification code 8 for the four different percentiles 20th, 30th, 40th, and 50th triangulated into a surface and hill-shaded.

Next we reclassify only those points marked with the temporary classification code 8 into ground (2) and unclassified (1) points using lasground by ignoring all points that still have the original classification code 0.

lasground -i crossfire_p20.laz ^
          -ignore_class 0 ^
          -wilderness ^
          -odix g -olaz

Below you see the resulting ground points computed from the subsets of points at four different percentiles 20th, 30th, 40th, and 50th triangulated into a surface and hill-shaded.

Both the ground classification of the 40th and the 50th percentile look reasonable. Only a few down spikes remain in the 40th percentile surface and a few additional bumps appear in the 50th percentile surface. Next we use lasheight with those two reasonable-looking ground surfaces to classify all points that are 20 centimeter below the triangulated ground surface into the noise classification code 7.

lasheight -i crossfire_p40g.laz ^
          -classify_below -0.2 7 ^
          -do_not_store_in_user_data ^
          -odix h -olaz

Now that the low noise points were removed (or rather classified as noise) we start the actual ground classification process. In this example we want to create a 50 cm DTM, hence it is more than sufficient to find one ground point per 25 cm cell. Therefore we first move all lowest non-noise last return per 25 cm cell to the temporary classification code 8.

Side note: One might also consider to modify the following workflow to run the ground classification on more than just the last returns by omitting ‘-ignore_first_of_many’ and ‘-ignore_intermediate’ from the lasthin call and by adding ‘-all_returns’ to the lasground call. Why? Because for all laser shots that resulted in a low noise point, this noise point will usually be the last return, so that the true ground hit could be the second to last return.

lasthin -i crossfire_p40gh.laz ^
        -ignore_first_of_many -ignore_intermediate ^
        -ignore_class 7 ^
        -step 0.25 ^
        -lowest ^
        -classify_as 8 ^
        -odix _low25 -olaz

The final ground classification is obtained by running lasground only on the points with temporary classification code 8 by ignoring all others, namely the noise points (7) and the unclassified points (0 and 1).

lasground -i crossfire_p40gh_low25.laz ^
          -ignore_class 0 1 7 ^
          -wilderness ^
          -odix g -olaz

We then use las2dem to create the 50 cm DTM from the points classified as ground. We store this DTM raster to the LAZ format which has shown to be the most efficient format for storing elevation or height rasters. We have started calling this format RasterLAZ. It is supported by all LAStools and the new DEMzip tool. One advantage is that we can feed RasterLAZ directly back into LAStools, for example as done below, for a second call to las2dem that computes a hill-shaded DTM.

las2dem -i crossfire_p40gh_low25g.laz ^
        -keep_class 2 ^
        -step 0.5 ^
        -ocut 9 -odix _dtm50 -olaz

las2dem -i crossfire_p40_dtm50.laz ^
        -step 0.5 ^
        -hillshade ^
        -odix _hill -opng

Below the resulting hill-shaded DTMs computed for the 40th and the 50th elevation percentile – as well as for the 45th elevation percentile that we’ve added for comparison.

Below we finally take a closer look at an example 1 meter profile line through the LiDAR classified by the 45th percentile workflow. There is a small stretch of ground points that was incorrectly classified as noise points (find the mouse cursor) so it might be worthwhile to change parameters slightly to make the noise classification less aggressive.

Side note follow-up: The return coloring shows there are indeed some ‘intermediate’ as well some ‘first of many returns’ just where we expect the bare terrain to be. However, there are not so many that the results can be expected to drastically change by including them into the ground finding process.

LASmoons: Nicolas Barth

Nicolas Barth (recipient of three LASmoons)
Department of Earth & Planetary Sciences
University of California, Riverside
UNITED STATES

Background:
The 850 km-long Alpine Fault (AF) is one of the world’s great laterally-slipping active faults (like California’s San Andreas Fault), which currently accommodates about 80% of the motion between the Australian and Pacific tectonic plates in the South Island of New Zealand (NZ). Well-dated sedimentary layers preserved in swamps and lakes adjacent to the AF currently provide one of the world’s most spatially and temporally complete record of large ground rupturing earthquakes (Howarth et al., 2018). Importantly these records reveal that major earthquakes occur with greater regularity on the AF than any other known fault, releasing a Magnitude (Mw) 7 to 8 earthquake on average every 249 ± 58 years and that the most recent earthquake was around Mw 8 in 1717 AD prior to European arrival. This computes to a conditional probability of 69% that the AF will rupture in the next 50 years. For a country that has recently had several notable earthquakes (e.g. 2010 Mw 7.1 Canterbury, 2016 Mw 7.8 Kaikoura) and has an economy heavily reliant on tourism, the next AF earthquake is the one NZ is trying to prepare for (note that a Mw 8 earthquake is about thirty times the energy release of a Mw 7).

The more data we can gather as scientists to constrain (1) the magnitude of the next AF earthquake, (2) the amount of lateral and vertical slip (offset roads, powerlines, etc.), (3) the coseismic effects (ground shaking, landslides, liquefaction), and (4) the duration it takes the landscape to recover (muddy rivers, increased sediment supply, prolonged landsliding), the more we can anticipate expected hazards and foster societal resilience.

Despite its name, the AF is almost completely obscured beneath a dense temperate rain-forest canopy, which has hindered fine-scale geomorphic studies. Relatively low quality airborne LiDAR (2 m-resolution bare-earth model) was first collected in 2010 for a 32 km-length of the central AF. Despite being the best studied portion of the AF, 82 % of the fault traces identified in the LiDAR were previously unmapped (Barth et al., 2012). The LiDAR reveals the width and style of ground deformation. Interpretation of the bare-earth landscape in combination with on the ground sampling, allows single earthquake displacements, uplift rates, recurrence of landslides, and post-earthquake sedimentation rates to be quantified. A new 2019 airborne LiDAR dataset collected along 230 km-length of the southern AF has great potential to improve our understanding of this relatively “well-behaved” fault system, what to expect from its next earthquake, and to give us insight into considerably more complex fault systems like the San Andreas.

(A) Aerial view of the South Island of New Zealand highlighting the boundary between the Pacific and Australian plates (white) and the Alpine Fault in particular (red). (B) View showing the extent of the 2019 airborne LiDAR survey to be processed by this lasmoons proposal. (C) Aerial imagery over Franz Josef, site of a 2010 airborne LiDAR survey. (D) 2010 Franz Josef LiDAR DTM hillshade (GNS Science). LiDAR has revolutionized our ability to map fault offsets and other earthquake ground deformation beneath this dense temperate rainforest.

Goal:
The LAStools software will be used to check the quality of the data (reclassing ground points and removing any low ground classed outliers if needed) and create a seamless digital terrain model (DTM) from the 1695 tiled LAS files provided. The DTM will be used to create derivative products including contours, slope map, aspect map, single direction B&W hillshades, multi-directional hillshades, and slope-colored hillshades to interpret the fault and landslide related landscape features hidden beneath the dense temperate rain-forest. The results will be used as seed data to seek national-level science funding to field verify interpretations and collect samples to determine ages of features (geochronology). The ultimate goal is to improve our understanding of the Alpine Fault prior to its next major earthquake and to communicate those findings effectively through publications in open access peer-reviewed journal articles and meetings with NZ regional councils.

Data:
+
airborne LiDAR survey collected in 2019 using a Riegl LSM-Q780 sensor by AAM New Zealand
+ provided data are as 1695 LAS files organized into 500 m x 500 m tiles and classified as ground and non-ground points (75 pts/m2 or ~0.8 ground-classed pts/m2; 320 GB total)

LAStools processing:
1) check the quality of the ALS data [lasinfo, lasoverlap, lasgrid]
2) [if needed] remove any low and high ground-classed outliers [lasnoise]
3) [if needed] reclassify ground and non-ground points [lasground]
4) create Digital Terrain Model (DTM) from ground points [blast2dem]

References:
Howarth, J.D., Cochran, U.A., Langridge, R.M., Clark, K.J., Fitzsimons, S.J., Berryman, K.R., Villamor, P., Strong, D.T. (2018) Past large earthquakes on the Alpine Fault: paleosismological progress and future directions. New Zealand Journal of Geology and Geophysics, v. 61, 309-328, doi: 10.1080/00288306.2018.1465658
Barth, N.C., Toy, V.G., Langridge, R.M., Norris, R.J. (2012) Scale dependence of oblique plate-boundary partitioning: new insights from LiDAR, central Alpine Fault, New Zealand. Lithosphere 4(5), 435-448, doi: 10.1130/L201.1

How Many Decimal Digits for Storing Longitude & Latitude?

Recently I came across this tweet containing the image below and it made me laugh … albeit not in the original way the tweet intended. The tweet was joking that “Anyone is able to open a GeoJSON file” and included the Microsoft Word screen shot seen below as a response to someone else tweeting that “Handing in a project as @GeoJSON. Let’s see if I get the usual “I can’t open this file” even though […]”. What was funny to me was seeing longitude and latitude coordinates stored with 15 decimal digits right of the decimal point. There are many memes about “German efficiency” but few about “German accuracy” 😁. Clearly it is time for another blog post about storage resolution and positional accuracy. The last blog post came on the heels of the national open elevation release of England with insane vertical resolution.

Longitude and latitude coordinates are stored with 15 decimal digits right of the decimal points.

By default LAStools will use 7 decimal digits to store longitude and latitude coordinates to a LAS or LAZ file. But what do 15 decimal digits mean for longitude and latitude coordinates? How “accurate” are the corresponding coordinates when converted to projected  coordinates? I took the second coordinate pair [ 10.049567988755534, 53.462766759577057 ] shown in the screen shot above and converted it from longitude and latitude to the easting and northing values of the WGS 84 / UTM 32N projection that has EPSG code 32632. Before conversion we quantize these numbers to have 5 through 15 decimal digits and then record the absolute difference to the coordinate pair that uses the most digits.

Number o decimal digits for longitude and latitude coordinate and absolute difference in projected position.

The table above shows that – at least for this particular longitude and latitude coordinate pair located in Germany – that 7 digits are sufficient to store coordinates with centimeter [cm] accuracy and that 8 digits are enough to store coordinates with millimeter [mm] accuracy. Any additional digit right of the decimal point will only be necessary when we need micrometer [um] or nanometer [nm] accuracy, which is very unlikely to be the case in most geospatial applications.

This means we could remove the 7 or 8 right most digits of each number from the screenshot that was tweeted and make this GeoJSON file even smaller, faster, and easier to store, transmit, open, and read. After this post was tweeted there was a follow-up tweet suggesting to have a look at this site for a more detailed analysis of what accuracy each digit in a longitude and latitude coordinate can store.

Clean DTM from Agisoft Photogrammetric Points of Urban Scene

We occasionally get permission to distribute a nice data sets and blog about how to best process it with LAStools because this gets around having to pay our “outrageous” consulting fees. (-: This time we received a nice photogrammetric point cloud of the Tafawa Balewa Square in Lagos Island, Lagos, Nigeria. This area is part of the central business district of Lagos and characterized by high-rise buildings. The Tafawa Balewa Square was constructed in 1972 over the site of a defunct track for horse racing and is bounded by Awolowo road, Cable street, Force road, Catholic Mission street and the 26-story Independence House. We want to create a nice Digital Terrain Model from the dense-matching point cloud that was generated with Photoscan by AgiSoft and – as always with photogrammetry – we have to take special care of low noise points. The final result is shown below. All processing commands used are here.

After downloading the data it is useful to familiarize yourself with the file, which can be done with lasview, lasinfo, and lasgrid using the command lines shown below. According to the lasinfo report there are around 47 million points points with RGB colors in the file and their average density is around 100 points per square meter.

lasview -i 0_raw\TafawaBalewa.laz

lasinfo -i 0_raw\TafawaBalewa.laz ^
        -cd -histo intensity 256 ^
        -histo z 1 ^
        -odir 1_quality -odix _info -otxt

lasgrid -i 0_raw\TafawaBalewa.laz ^
        -step 1 ^
        -density ^
        -false -set_min_max 50 150 ^
        -odir 1_quality -odix _d_50_150 -opng

The average point density value of 100 from the lasinfo report suggests that 50 as the minimum and 150 as the maximum are good false color ramp values for a map showing how the point density per square meter is distributed.

Color-coded point density: blue equals 50 or less and red means 150 or more points per square meter.

We use lastile to create a buffered tiling for the 47 million points. We use a tile size of 200 meters and request a large buffer of 50 meters around every tile because there are large buildings in the survey areas. We also flag buffer points as withheld so they can be easily be dropped later.

lastile -i 0_raw\TafawaBalewa.laz ^
        -tile_size 200 -buffer 50 -flag_as_withheld ^
        -odir 2_tiles_raw -o tafawa.laz

If you inspect the resulting tiles – such as ‘tafawa_544000_712600.laz’ as shown here – with lasview you will see the kind of low noise that is shown below and that may cause a ground classification algorithm. While our lasground software is able to deal with a certain amount of low noise – if there are too many it will likely latch onto them. Therefore we will first generate a subset of points that has as few as possible of such low noise points.

Typical low noise in dense-matching photogrammetry points in urban scene.

Next we use a sequence of three LAStools modules, namely lasthinlasground, and lasheight to classify this photogrammtric point cloud into ground and non-ground points. All processing commands used are here. First we use lasthin to give the point the classification code 8 that is closest to the 50th percentile in elevation within every 50 centimeter by 50 centimeter cell (but only if the cells containing at least 20 points).

lasthin -i 2_tiles_raw\tafawa*.laz ^
        -step 0.5 ^
        -percentile 50 20 ^
        -classify_as 8 ^
        -odir 3_tiles_median_50cm -olaz ^
        -cores 3

Next we use lasground to ground-classify only the points that have classification code 8 (i.e. by ignoring those with classification codes 0) and set their classification code to ground (2) or non-ground (1). Because of the large buildings in this urban scene we use ‘-metro’ which uses a large step size of 50 meters for the pre-processing. This also sets the internally used bulge parameter to 5.0 which you can see if you run the tool in verbose ‘-v’ mode. In three different trial runs we determined that forcing the bulge parameter to be 0.5 instead gave better results. The bulge and the spike parameters can be useful to vary in order to improve ground classification results (also see the README file).

lasground -i 3_tiles_median_50cm\tafawa*.laz ^
          -ignore_class 0 ^
          -metro -bulge 0.5 ^
          -odir 4_tiles_ground_50cm -olaz ^
          -cores 3

The resulting ground points are a subset with a resolution of 50 centimeter that is good enough to create a DTM with meter resolution, which we do with las2dem command line shown below. We really like storing DTM elevation rasters to the LAZ point format because it is a more compressed way of storing elevation rasters compared to ASC, BIL, TIF, or IMG. It also makes the raster output a natural input to subsequent LAStools processing steps.

las2dem -i 4_tiles_ground_50cm\tafawa*.laz ^
        -keep_class 2 ^
        -step 1 -kill 100 ^
        -use_tile_bb ^
        -odir 5_tiles_dtm_1m -olaz ^
        -cores 3

Finally we use blast2dem to create a seamless hill-shaded version of our 1 meter DTM from on-the-fly merged elevation rasters. This is the DTM pictured at the beginning of this article.

blast2dem -i 5_tiles_dtm_1m\tafawa*.laz -merged ^
          -step 1 ^
          -hillshade ^
          -o dtm_1m.png

The corresponding DSM pictured at the beginning of this article was generated with the two command lines below by first keeping only the 95th percentile highest elevation of every 50 cm by 50 cm cell with lasthin (which remove spurious high noise points) and then by triangulating the surviving points with blast2dem into a seamless TIN that is also hill-shaded and rasterized with 1 meter resolution. Running the 64 bit version of lasthin (note the ‘-cpu64‘ switch) allows us to work on the entire data set (rather than its tiles version) at once, where the standard 32 bit version may run out of memory.

lasthin -i 0_raw\TafawaBalewa.laz ^
        -cpu64 ^
        -step 0.5 ^
        -percentile 95 20 ^
        -o 0_raw\TafawaBalewa_p95_50.laz

blast2dem -i 0_raw\TafawaBalewa_p95_50.laz ^
          -step 1 ^
          -hillshade ^
          -o dsm_1m.png

In order to generate the final DTM at higher resolution we use lasheight to pull all points into the ground class that lie within a 5 cm distance vertically below or a 15 cm distance vertically above the triangulated surface of ground points computed in the previous step. You could experiment with other values here to be less or more conservative about pulling detail into the ground class.

lasheight -i 4_tiles_ground_50cm\tafawa*.laz ^
          -classify_between -0.05 0.15 2 ^
          -odir 6_tiles_ground -olaz ^
          -cores 3

We repeat the same processing step as before las2dem to create the raster DTM tiles, but this time with a resolution of 25 cm.

las2dem -i 6_tiles_ground\tafawa*.laz ^
        -keep_class 2 ^
        -step 0.25 -kill 100 ^
        -use_tile_bb ^
        -odir 7_tiles_dtm_25cm -olaz ^
        -cores 3

And we again use blast2dem to create a seamless hill-shaded version of the DTM from on-the-fly merged elevation rasters, but this time with a resolution of 25 cm. This is the DTM shown below. All processing commands used are here.

blast2dem -i 7_tiles_dtm_25cm\tafawa*.laz -merged ^
          -step 0.25 ^
          -hillshade ^
          -o dtm_25cm.png

Hill-shade of final DTM with resolution of 25 cm.

Using Open LiDAR to Remove Low Noise from Photogrammetric UAV Point Clouds

We collected drone imagery of the restored “Kance” tavern during the lunch stop of the UAV Tartu summer school field trip (actually the organizer Marko Kohv did that, as I was busy introducing students to SUP boarding). With Agisoft PhotoScan we then processed the images into point clouds below the deck of the historical “Jõmmu” barge on the way home (actually Marko did that, because I was busy enjoying the view of the wetlands in the afternoon sun). The resulting data set with 7,855,699 points is shown below and can be downloaded here.

7,855,699 points produced with Agisoft Photoscan

Generating points using photogrammetric techniques in scenes containing water bodies tends to be problematic as dense blotches of noise points above and below the water surface are common as you can see in the picture below. Especially the low points are troublesome as they adversely affect ground classification which results in poor Digital Elevation Models (DTMs).

Clusters of low noise points nearly 2 meters below the actual surface in water areas.

In a previous article we have described a LAStools workflow that can remove excessive low noise. In this article here we use external information about the topography of the area to clean our photogrammetry points. How convenient that the Estonian Land Board has just released their entire LiDAR archives as open data.

Following these instructions here you can download the available open LiDAR for this area, which has the map sheet index 475681. Alternatively you can download the four currently available data sets here flown in spring 2010, in summer 2013, in spring 2014, and in summer 2017. In the following we will use the one flown in spring 2014.

We can view both data sets simultaneously in lasview. By adding ‘-faf’ to the command-line we can switch back and forth between the two data sets by pressing ‘0’ and ‘1’.

lasview -i Kantsi.laz ^
        -i 475681_2014_tava.laz ^
        -points 10000000 ^
        -faf

We find cut the 1 km by 1 km LiDAR tile down to a 250 m by 250 m tile that nicely surrounds our photogrammetric point set using the following las2las command-line:

las2las -i 475681_2014_tava.laz ^
        -inside_tile 681485 6475375 250 ^
        -o LiDAR_Kantsi.laz

lasview -i Kantsi.laz ^
        -i LiDAR_Kantsi.laz ^
        -points 10000000 ^
        -faf

Scrutinizing the two data sets we quickly find that there is a miss-alignment between the dense imagery-derived and the comparatively sparse LiDAR point clouds. With lasview we investigate the difference between the two point clouds by hovering over a point from one point cloud and pressing <i> and then hovering over a somewhat corresponding point from the other point cloud and pressing <SHIFT>+<i>. We measure displacements of around 2 meters vertically and of around 3 to 3.5 meter in total.

Before we can use the LiDAR points to remove the low noise from the photogrammetric points we must align them properly. For simple translation errors this can be done with a new feature that was recently added to lasview. Make sure to download the latest version (190404 or newer) of LAStools to follow the steps shown in the image sequence below.

las2las -i Kantsi.laz ^
        -translate_xyz 0.89 -1.90 2.51 ^
        -o Kantsi_shifted.laz

lasview -i Kantsi_shifted.laz ^
        -i LiDAR_Kantsi.laz ^
        -points 10000000 ^
        -faf

The result looks good in the sense that both sides of the photogrammetric roof are reasonably well aligned with the LiDAR. But there is still a shift along the roof so we repeat the same thing once more as shown in the next image sequence:

We use a suitable displacement vector and apply it to the photogrammetry points, shifting them again:

las2las -i Kantsi_shifted.laz ^
        -translate_xyz -1.98 -0.95 0.01 ^
        -o Kantsi_shifted_again.laz

lasview -i Kantsi_shifted_again.laz ^
        -i LiDAR_Kantsi.laz ^
        -points 10000000 ^
        -faf

The result is still not perfect as there is also some rotational error and you may find another software such as Cloud Compare more suited to align the two point clouds, but for this exercise the alignment shall suffice. Below you see the match between the photogrammetry points and the LiDAR TIN before and after shifting the photogrammetry points with the two (interactively determined) displacement vectors.

The final steps of this exercise use las2dem and the already ground-classified LiDAR compute a 1 meter DTM, which we then use as input to lasheight. We classify the photogrammetry points using their height above this set of ground points with 1 meter spacing: points that are 40 centimeter or more below the LiDAR DTM are classified as noise (7), points that are between 40 below to 1 meter above the LiDAR DTM are classified to a temporary class (here we choose 8) that has those points that could potentially be ground points. This will help, for example, with subsequent ground classification as large parts of the photogrammetry points – namely those on top of buildings and in higher vegetation – can be ignored from the start by a ground classification algorithm such as lasground.

las2dem -i LiDAR_Kantsi.laz ^
        -keep_class 2 ^
        -kill 1000 ^
        -o LiDAR_Kantsi_dtm_1m.bil
lasheight -i Kantsi_shifted_again.laz ^
          -ground_points LiDAR_Kantsi_dtm_1m.bil ^
          -classify_below -0.4 7 ^
          -classify_between -0.4 1.0 8 ^
          -o Kantsi_cleaned.laz

Below the results we have achieved after “roughly” aligning the two point clouds with some new lasview tricks and then using the LiDAR elevations to classify the photogrammetry points into “low noise”, “potential ground”, and “all else”.

We thank the Estonian Land Board for providing open data with a permissive license. Special thanks also go to the organizers of the UAV Summer School in Tartu, Estonia and the European Regional Development Fund for funding this event. Especially fun was the fabulous excursion to the Emajõe-Suursoo Nature Reserve and through to Lake Peipus aboard, overboard and aboveboard the historical barge “Jõmmu”. If you look carefully you can also find the barge in the photogrammetry point cloud. The photogrammetry data used here was acquired during our lunch stop.

Fun aboard and overboard the historical barge “Jõmmu”.

No Sugarcoating: Sweet LiDAR from RiCOPTER carrying VUX-1UAV over Sugarcane

Recently we saw an interesting LiDAR data set talked about on social media by Chad Netto from Chustz Surveying in New Roads, Louisiana and asked for a copy. It is a LiDAR scan of a sugarcane plantation in Assumption Parish, Louisiana carried out with the VUX-1UAV by RIEGL mounted onto a RiCOPTER and guided by an Applanix IMU and a Trimble base station. That is probably one of the sweetest (but also one of the most expensive) UAV LiDAR system you can buy today. I received this LiDAR file and this trajectory file. In the following we talk a detailed look at this data set.

First we run lasinfo to get an idea of the contents of the data set. We create various histograms as those can often help understand an unfamiliar data set:

lasinfo -i sugarcane\181026_163424.laz ^
        -cd ^
        -histo gps_time 5 ^
        -histo intensity 64 ^
        -histo point_source 1 ^
        -histo z 5 ^
        -odix _info -otxt

You can download the resulting report here. For the 84,751,955 points we notice that

  1. both horizontal and vertical coordinates are stored in US survey feet
  2. with scale factors of 0.00025 this means a resolution of 76 micrometer
  3. there is no explicit flight line information (all point source IDs are zero)
  4. gaps in the GPS time stamp histogram are suggesting multiple lines

First we use las2las to lower the insanely high resolution from 0.00025 US survey feet to something more reasonable for an airborne UAV scan, namely to 0.01 or 1 hundredths of a US survey foot or centi-US-survey-feet:

las2las -i sugarcane\181026_163424.laz ^
        -rescale 0.01 0.01 0.01 ^
        -odix _cft -olaz

I have already done this for you. The file that is online is already in “centi-US-survey-feet” because it reduced the file size from the original 678 MB file that we got from Netto to the 518 MB file that is online, meaning that you had 160 MB less data to download.

Next we use lassplit to recover the original flight lines as follows:

lassplit -i sugarcane\181026_163424.laz ^
         -recover_flightlines ^
         -odir sugarcane\0_recovered_strips ^
         -o assumption.laz

This results in 5 strips. We then use lassort to bring the strips back into their original acquisition order by sorting first based on the GPS time stamp (which brings all returns of one pulse together) and second on the return number (which sorts them in ascending order). We do this on 3 cores in parallel with this command:

lassort -i sugarcane\0_recovered_strips\*.laz ^
        -gps_time ^
        -return_number ^
        -odir sugarcane\1_sorted_strips -olaz ^
        -cores 3

We also create a spatial index for each of these strips using lasindex so that any area-of-interest query that we do later will be significantly accelerated. See the README file for the meaning of the parameters:

lasindex -i sugarcane\1_sorted_strips\*.laz ^
         -tile_size 10 -maximum -100 ^
         -cores 3

Then we check for flight line alignment using lasoverlap by comparing – per 2 feet by 2 feet area – the lowest elevation value of points from different strips wherever there is overlap. Cells with an absolute vertical difference of less than a quarter of a foot are mapped to white. Cells with vertical differences of more (or less) than a quarter foot are mapped to an increasingly red (or blue) color that is saturated red (or blue) when one full foot is reached.

lasoverlap -i sugarcane\1_sorted_strips\*.laz ^
           -files_are_flightlines ^
           -step 2.0 ^
           -min_diff 0.25 -max_diff 1.0 ^
           -o sugarcane\2_quality\overlap.png

The resulting image looks dramatic at first glance. But we have to remember that this is sugarcane. The penetration of the laser can vary greatly depending on the direction from which the beam hits the densely standing stalks. Large differences between flight lines can be expected where sugarcane stands tall. We need to focus our visual quality checks on the few open areas, namely the access roads and harvested areas.

Color-mapped highest vertical difference in lowest point per 2 feet by 2 feet area between overlapping flight lines.

We use las2las via its native GUI to cut out several suspicious-looking open areas with overly red or overly blue shading. By loading the resulting image into the GUI these areas-of-interest are easy to target and cut out.

las2las -i sugarcane\1_sorted_strips\*.laz -gui

Overlaying the difference image in the GUI of las2las to cut out suspicious areas for closer inspection.

We cut out four square 100 by 100 meter tiles in open areas that show a suspiciously strong pattern of red or blue colors for closer inspection. The command lines for these four square areas are given below and you can download them here:

  1. assumption_3364350_534950_100.laz
  2. assumption_3365600_535750_100.laz
  3. assumption_3364900_535500_100.laz
  4. assumption_3365500_535600_100.laz
las2las -i sugarcane\1_sorted_strips\*.laz ^
        -merged -faf ^
        -inside_tile 3364350 534950 100 ^
        -o sugarcane\assumption_3364350_534950_100.laz

las2las -i sugarcane\1_sorted_strips\*.laz ^
        -merged -faf ^
        -inside_tile 3365600 535750 100 ^
        -o sugarcane\assumption_3365600_535750_100.laz

las2las -i sugarcane\1_sorted_strips\*.laz ^
        -merged -faf ^
        -inside_tile 3364900 535500 100 ^
        -o sugarcane\assumption_3364900_535500_100.laz

las2las -i sugarcane\1_sorted_strips\*.laz ^
        -merged -faf ^
        -inside_tile 3365500 535600 100 ^
        -o sugarcane\assumption_3365500_535600_100.laz

In the image sequence below we scrutinize these differences which will lead us to notice two things:

  1. There are vertical miss-alignments of around one foot. These big difference can especially be observed between flight lines that cover an area with a very high point density and those that cover the very same area with a very low point density.
  2. There are horizontal miss-alignments of around one foot. Again these differences seem somewhat correlated to the density that these flight lines cover a particular area with.

For the most part the miss-aligned points come from a flight line that has only sparse coverage in that area. In a flat terrain the return density per area goes down the farther we are from the drone as those areas are only reached with higher and higher scan angles. Hence an immediate idea that comes to mind is to limit the scan angle to those closer to nadir and lower the range from -81 to 84 degrees reported in the lasinfo report to something like -75 to 75, -70 to 70, or -65 to 65 degrees. We can check how this will improve the alignment with these lasoverlap command lines:

lasoverlap -i sugarcane\1_sorted_strips\*.laz ^
           -files_are_flightlines ^
           -keep_scan_angle -75 75 ^
           -step 2.0 ^
           -min_diff 0.25 -max_diff 1.0 ^
           -o sugarcane\2_quality\overlap75.png

lasoverlap -i sugarcane\1_sorted_strips\*.laz ^
           -files_are_flightlines ^
           -keep_scan_angle -70 70 ^
           -step 2.0 ^
           -min_diff 0.25 -max_diff 1.0 ^
           -o sugarcane\2_quality\overlap70.png

lasoverlap -i sugarcane\1_sorted_strips\*.laz ^
           -files_are_flightlines ^
           -keep_scan_angle -65 65 ^
           -step 2.0 ^
           -min_diff 0.25 -max_diff 1.0 ^
           -o sugarcane\2_quality\overlap65.png

lasoverlap -i sugarcane\1_sorted_strips\*.laz ^
           -files_are_flightlines ^
           -keep_scan_angle -60 60 ^
           -step 2.0 ^
           -min_diff 0.25 -max_diff 1.0 ^
           -o sugarcane\2_quality\overlap60.png

This simple technique significantly improves the difference image. Based on these images would suggest to only use returns with a scan angle between -70 and 70 degrees for any subsequent analysis. This seems to remove most of the discoloring in open areas without loosing too many points. Note that only using returns with a scan angle between -60 and 60 degrees means that some flight lines are no longer overlapping each other.

Note also that by limiting the scan angle we get suddenly get white areas even in incredible dense vegetation. The more horizontal a laser shoot is the more likely it will only hit higher up sugarcane plants and the less likely it will penetrate all the way to the ground. The white areas coincide with where laser pulses are close to nadir which is in the flight line overlap areas that directly below the drone’s trajectory.

Can we improve alignment further? Not with LAStools, so I turned to Andre Jalobeanu, a specialist on that particular issue, who I have known for many years. Andre has developed BayesStripAlign – a software by his company BayesMap that is quite complementary to LAStools and does exactly what the name suggests: it align strips. I gave Andre the five flight lines and he aligned them for me. Below the new difference images:

We cut out the very same four square areas from the realigned strips for closer inspection and you may investigate them on your own. You can download them here.

  1. assumption_3364350_534950_100_realigned.laz
  2. assumption_3365600_535750_100_realigned.laz
  3. assumption_3364900_535500_100_realigned.laz
  4. assumption_3365500_535600_100_realigned.laz

In the image sequence below we are just looking at the last of the four square areas and you can see that most of the miss-alignment we saw earlier between the flight lines was removed.

We would like to thank Chad Netto from Chustz Surveying to make this data set available to us and Andre Jalobeanu from BayesMap to align the flight lines for us.

Digital Pothole Removal: Clean Road Surface from Noisy Pix4D Point Cloud

How to generate a clean Digital Terrain Model (DTM) from point clouds that were generated with the image matching techniques implemented in various photogrammetry software packages like those from Pix4D, AgiSoft, nframes, DroneDeploy and others has become an ever more frequent inquiry. There are many other blog posts on the topic that you can peruse as well [1,2,3,4,5,6,7,8]. In the following we go step by step through the process of removing low noise from a high-density point cloud that was generated with Pix4D software. A composite of the resulting DTM and DSM is shown below.

Final DSM and DTM created with LAStools for a photogrammetric point cloud of a road generated by Pix4D.

After downloading the data it is useful to familiarize oneself with the number of points, the density of points and their geo-location. This can be done with lasview, lasinfo, and lasgrid using the command lines shown below. There are around 19 million points in the file and their density averages around 2300 points per square meter. Because the RGB values have a 16 bit range (as evident in the lasinfo report) we need to add the option ‘-scale_rgb_down’ to the command line when producing the RGB raster with lasgrid.

lasview -i 0_photogrammetry\densified_point_cloud.laz

lasinfo -i 0_photogrammetry\densified_point_cloud.laz ^
        -cd ^
        -o 1_quality\densified_point_cloud.txt

lasgrid -i 0_photogrammetry\densified_point_cloud.laz ^
        -scale_rgb_down ^
        -step 0.10 ^
        -rgb ^
        -fill 1 ^
        -o 1_quality\densified_point_cloud.png

The first step is to use lastile and create smaller and buffered tiles for these 19 million photogrammetry points. We use a tile size of 100 meters, request a buffer of 10 meters around every tile, and flag buffer points as withheld so they can be easily be dropped later. We also make sure that all classification codes are reset to 0.

lastile -i 0_photogrammetry\points.laz ^
        -set_classification 0 ^
        -tile_size 100 -buffer 10 -flag_as_withheld ^
        -o 2_tiles_raw\seoul.laz -olaz

We start with lassort as a pre-processing step that rearranges the points into a more coherent spatial order which often accelerates subsequent processing steps.

lassort -i 2_tiles_raw\seoul_*.laz ^
        -odir 3_tiles_temp0 -olaz ^
        -cores 4

Next we use a sequence of four modules, namely lasthin, lasnoiselasground, and lasheight with fine-tuned parameters to remove the low noise points that are typical for point clouds generated from imagery by photogrammetry software. A typical example for such noise points are shown in the image below generated with this call to lasview:

lasview -i 3_tiles_temp0\seoul_210400_542900.laz ^
        -inside 210406 542894 210421 542921 ^
        -points 20000000 ^
        -kamera 0 -95 90 0 -0.3 1.6 ^
        -point_size 4

Ground surface noise (exaggerated by pressing <]> in lasview which doubles the scale in z).

As always, the idea is to construct a surface that is close to the ground but always above the noise so that it can be used to declare all points beneath it as noise. Below is a processing pipeline whose parameters work well for this data and that you can fine tune for the point density and the noise profile of your own data.

First we use lasthin to give those points the classification code 8 that are closest to the 70th percentile in elevation within every 20 cm by 20 cm cell. As statistics like percentiles are only stable for a sufficient number of points we only do this for cells that contain 25 points or more. Given that we have an average of 2300 points per square meter this should easily be the case for all relevant cells.

lasthin -i 3_tiles_temp0\seoul_*.laz ^
        -step 0.20 ^
        -percentile 70 25 ^
        -classify_as 8 ^
        -odir 3_tiles_temp1 -olaz ^
        -cores 4

The we run lasnoise only points on the points with classification code 8 and reclassify all “overly isolated” points with code 9. The check for isolation uses cells of size 20 cm by 20 cm by 5 cm and reclassifies the points in the center cell when the surrounding neighborhood of 27 cells has only 3 or fewer points in total. Changing the parameters for ‘-step_xy 0.20 -step_z 0.05 -isolated 3’ will remove isolated points more or less aggressive.

lasnoise -i 3_tiles_temp1\seoul_*.laz ^
         -ignore_class 0 ^
         -step_xy 0.20 -step_z 0.05 -isolated 3 ^
         -classify_as 9 ^
         -odir 3_tiles_temp2 -olaz ^
         -cores 4

Next we use lasground to ground-classify only the surviving points (that still have classification code 8) by ignoring those with classification codes 0 or 9. This sets their classification code to either ground (2) or non-ground (1). The temporary surface defined by the resulting ground points will be used to classify low points as noise in the next step.

lasground -i 3_tiles_temp2\seoul_*.laz ^
          -ignore_class 0 9 ^
          -town -ultra_fine -bulge 0.1 ^
          -odir 3_tiles_temp3 -olaz ^
          -cores 4

Then we use lasheight to classify all points that are 2.5 cm or more below the triangulated surface of temporary ground points as points as noise (7) and all others as unclassified (1).

lasheight -i 3_tiles_temp3\seoul_*.laz ^
          -classify_below -0.025 7 ^
          -classify_above -0.225 1 ^
          -odir 4_tiles_denoised -olaz ^
          -cores 4

The progress of each step is illustrated visually in the two image sequences shown below.

Now that all noise points are classified we start a standard processing pipeline, but always ignore the low noise points that are now classified with classification code 7.

The processing steps below create a 10 cm DTM raster. We first use lasthin to classify the lowest non-noise point per 10 cm by 10 cm cell. Considering only those lowest points we use lasground with options ‘-town’, ‘-extra_fine’, ‘-bulge 0.05’, and ‘-spike 0.05’. Using las2dem the resulting ground points are interpolated into a TIN and rasterized into a 10 cm DTM cutting out only the center 100 meter by 100 meter tile. We store the DTM raster as a gridded LAZ for maximal compression and finally merge these gridded LAZ files to create a hillshaded raster in PNG format with blast2dem.

lasthin -i 4_tiles_denoised\seoul_*.laz ^
        -ignore_class 7 ^
        -step 0.10 ^
        -lowest ^
        -classify_as 8 ^
        -odir 5_tiles_thinned_lowest -olaz ^
        -cores 4

lasground -i 5_tiles_thinned_lowest\seoul_*.laz ^
          -ignore_class 1 7 ^
          -town -extra_fine ^
          -bulge 0.05 -spike 0.05 ^
          -odir 6_tiles_ground -olaz ^
          -cores 4

las2dem -i 6_tiles_ground\seoul_*.laz ^
        -keep_class 2 ^
        -step 0.10 ^
        -use_tile_bb ^
        -odir 7_tiles_dtm -olaz ^
        -cores 4

blast2dem -i 7_tiles_dtm\seoul_*.laz -merged ^
          -hillshade ^
          -step 0.10 ^
          -o dtm_hillshaded.png

The processing steps below create a 10 cm DSM raster. We first use lasthin to classify the highest non-noise point per 10 cm by 10 cm cell. With las2dem the highest points are interpolated into a TIN and rasterized into a 10 cm DSM cutting out only the center 100 meter by 100 meter tile. Again we store the raster as gridded LAZ for maximal compression and merge these files to create a hillshaded raster in PNG format with blast2dem.

lasthin -i 4_tiles_denoised\seoul_*.laz ^
        -ignore_class 7 ^
        -step 0.10 ^
        -highest ^
        -classify_as 8 ^
        -odir 8_tiles_thinned_highest -olaz ^
        -cores 4

las2dem -i 8_tiles_thinned_highest\seoul_*.laz ^
        -keep_class 8 ^
        -step 0.10 ^
        -use_tile_bb ^
        -odir 9_tiles_dsm -olaz ^
        -cores 4

blast2dem -i 9_tiles_dsm\seoul_*.laz -merged ^
          -hillshade ^
          -step 0.10 ^
          -o dsm_hillshaded.png

The final result is below. The entire script is linked here. Simply download it, modify it as needed, and try it on this data or on your own data.