Martin Isenburg received his MSc in 1999 from UBC in Vancouver, Canada, and his PhD in 2004 from UNC in Chapel Hill, USA — both in Computer Science. Currently he is an independent scientist, lecturer, and consultant. Martin has created a popular LiDAR processing software called LAStools that is widely used across industry, government agencies, research labs, and educational institutions. LAStools is the flagship product of rapidlasso GmbH, the company he founded in 2012. Martin’s ultimate goal is to combine high-tech remote sensing and organic urban farming in a “laser farm” that promotes green projects as hip and fun activities for the iPad generation.
Q: Thanks, Martin, for taking the time to have a chat with GeoHipster! Tell us about your ideas on “Front yard chickens”. Chickens are so awesome. We’ve heard that your chickens were about to be equipped with lasers. How did that go?
A: Happy to talk geospatial chickens. See, in the backyard chickens are a fun way to be green. But put three (not twenty!) in the front yard, and they create green communities. You meet your neighbors (dragged to your fence by their kids), and soon you are bartering eggs for kale because Lori and Dan across the street now have “front yard veggies”. And why lasers? Not for arming or roasting the chickens (common mistake), but for filming them in real-time 3D. Sort of like Radiohead’s video “House of Cards”, but better. A “happy feed” of urban farm bliss to lure folks beyond this neighborhood into green fun. Troubles over “cluster-bombing” the homeland with garden-fresh zucchinis forced me to put this project on hold.
Q: Can you give us an overview of LiDAR and how it works?
A: Fire a really short burst of laser light, and measure the exact duration until its reflection comes back. That allows you to compute the distance to whatever object was hit. Record the exact position from where and the exact direction in which you fired the laser, and you can calculate the exact position of these hits. Repeat this several hundred thousand times a second with an airborne LiDAR whose laser beam sweeps out the terrain, and you get enough information to model ground, buildings, and vegetation in 3D.
Q: How has LiDAR data storage evolved over the years?
A: The LiDAR points were first stored and exchanged as plain text files: x, y, z, intensity. But ASCII becomes inefficient as a storage format as point numbers go up. Several industry players got together and created a simple binary data exchange format — the LAS format — that was eventually donated to the ASPRS. LAS has become a huge success, and everybody supports it. Nowadays the specification is maintained by the LAS Working Group (LWG) of the ASPRS. That sounds fancy, but is really just a dozen or so email addresses that get cc-ed when an issue is discussed.
Q: You explained that LiDAR data is huge. How much data are we talking about?
A: One LiDAR return — how the hits are called — is typically 28 bytes. An airborne survey with 4 shots per square meter may average 8 LiDAR returns per square meter. For a small area of 100 square kilometers this is over 20 GB of raw LAS. Subsequent processing steps often create multiple copies of this data. Nowadays countries either have, or are going for nation-wide LiDAR coverage. So many terabytes of LiDAR are already out there, and petabytes are going to come.
Q: You developed the compressed LAZ format. Can you give us some background?
A: I spent years of graduate work on compressing polygon meshes, but few people have such data. When I stumbled upon folders of LAS files, I figured having a compressor for these point clouds may actually be useful. I wrote the LASzip prototype mainly to supplement an academic paper, but people found it on my web pages and used it. In 2010 I was asked to release LASzip with an open license to defeat a proprietary format that federal agencies feared would make compressed LiDAR costly. Eventually the US Army Corps of Engineers (USACE) sponsored the open-sourcing of LASzip.
Q: What is the development process that you use for making changes to the compressed LiDAR format?
A: I am very careful with changes, and try to be as transparent as possible about them. First I seek community input on new features via the “LAS room” and the “LAStools” forums. Once discussed, the new feature is implemented as a prototype for testing. If the prototype proves itself over time, these features are moved into the new release. But maybe the time has come to make LASzip an official standard with a committee overseeing future changes.
Q: The release of LAS 1.4 means new point types. This is a disruption of the LAS format in general. What opportunities does this present for LASzip?
A: I have held back extending LASzip to LAS 1.4. Like you say, the new point types in LAS 1.4 are a “natural break” in the format that offers the opportunity to improve the compressor without creating incompatibilities. An open “call for input” was issued to get feedback on features the community wants to see in the next generation of LASzip.
However, LASzip can already compress LAS 1.4 content. NOAA stepped up to sponsor the “LAS 1.4 compatibility mode” where new point types are re-coded into old ones by storing their new attributes as “extra bytes”. Added bonus: many older software packages can read re-coded LAS 1.4 content without upgrade.
Q: You’ve released a simpler interface to LASzip in 2013. How did that turn out?
A: When Esri came knocking, saying the LASzip code was too complicated, I asked them to sponsor the effort for a simpler DLL. But then I decided to create this DLL without further delay. LAZ was the de-facto LiDAR compression standard, and I wanted to remove any possible hurdle for its adoption. It is disappointing that Esri has still not added LAZ support to ArcGIS. The new DLL was essentially written for them.
Q: Recently Esri announced a variation of the open LAS standard called “Optimized LAS”. Can you describe the changes to how LAS files are supported?
A: The name is rather misleading. “Optimized LAS” is a closed format that compresses the content of a LAS file into a proprietary file. The resulting ‘*.zlas’ files are very similar to the ‘*.laz’ files produced by LASzip, which is why the new Esri format is also known as the “LAZ clone”.
Q: So from your view, how do they stack up?
A: The technical differences between LASzip and “Optimized LAS” are minimal. In terms of compression and speed, the two are almost identical. In terms of features, Esri includes spatial indexing information into the *.zlas files whereas we had been storing it as separate *.lax files. It took just a couple of hours to “upgrade” LASzip to match the feature set of “Optimized LAS” by adding one option for spatial sorting and another option for integrating the spatial index into the file. The argument that Esri could not use LASzip due to missing features is obviously a dud. The “LAZ clone” was created to tie LiDAR folks long-term to the ArcGIS platform.
Esri likes to point out that their format contains point statistic summaries. This is so trivial that any developer could add this in an afternoon. Such summaries are a good idea. I encourage Esri to propose a new “Variable Length Record” for that purpose as an addendum to the LAS specification. This is why they are part of the ASPRS LAS Working Group.
Q: Some readers may have seen this post back in 2014, believing the LASzip controversy was resolved. The post was your April Fools’ Day prank. Why did you do it?
A: I modified LASzip just a few days before April 1st 2014 to feature-match the “LAZ clone”. The triviality of these modifications made it obvious that further technical reasoning with Esri was moot. My last hope was to show Esri management how much applause they would garner from working with the community. So I wrote a prank press release, stating that Esri and rapidlasso were developing a joint compressor. Almost everything in this press release was true, except that Esri had not agreed yet to such cooperation. The response was incredible as the collected comments show…
Q: What are the ramifications of dueling data formats? What’s the point the entire GIS community at large should take home?
A: The instant loser is the user who will have to convert data back and forth. The instant winners are companies that provide data converters. Hey wait, that includes me! 🙂 The long-term loser is the GIS community that will find more and more LiDAR locked to a single platform. The long-term winner is the provider of this platform (or so they hope).
Q: You just mentioned processing LiDAR in a web browser. Is that a new thing?
Q: The term geohipster is bestowed affectionately. With your urban farming and your front yard chickens, I feel like using Steven Feldman’s geohippy term. Do you feel more like a geohipster or a geohippy?
A: The original idea behind downtownfarm — the mash-up of chickens and lasers — was to get away from the granola-hippy image of urban farming and make green more trendy and cool. How about geoyuppy?