I often hear local governments who are flying Lidar to establish the terrain component of their aerial ortho projects. These local governments request a copy of the ‘raw’ Lidar data. What does ‘raw’ mean?
Typically, what the local government really wants is not the ‘raw’ Lidar data, but all Lidar returns further down the processing line. Commercial tools to process ‘raw’ Lidar imagery are not easy to use, nor are they currently cost effective. ERDAS, the inventor of commercial remote sensing, has historically been the company to introduce cost-effective remote sensing tools into the market-place. (I see Lidar as a remote sensing tool that heretofore has been used most successfully by photogrammetrists.)
It is understandable that local governments want more than detailed terrain model from their Lidar collection. Heretofore, Lidar collections have centered on elevation, but there is much more value available in Lidar data than only an elevation model. It is my belief, to use Lidar only as an elevation source, is like using imagery only as a backdrop to a GIS vector dataset; valuable, but very wasteful of the tax payer’s money.
Local governments need more value from their Lidar collections than just elevation, but that value is not easily captured in the ‘raw’ data. What Lidar data do local governments need to request? They need Lidar data that has been bore-sighted. Bore-sighting removes the small imprecision found in the GPS and IMU. These errors can be quite significant errors at a flying height of 5,000 feet (1500m).
After bore-sighting, the data are edited to define ground points and other features. This is a needed step to prepare the point cloud for terrain purposes. Yet, the Editing process can introduce very valuable information. Here is where vendor can classify the points into categories. The philosophy that points should be flagged with a classification, rather than be deleted from the dataset, emerged during the development of the LAS data format standard.
Of the several classification categories that can be defined in a LAS file, the categories local governments may wish to take advantage are: Ground; Low Vegetation; Medium Vegetation; High Vegetation; Building; and Water.
It is my opinion; the local government should obtain the intermediate product at this stage. All returns, bore-sighted and classified, no points added or removed. And, a full QA/QC report. If it were me, I’d ask for the intensity information as well.
The final need, the LAS tiles should each have fully defined horizontal and vertical coordinate systems (as ASPRS suggests). So many times local governments accept LAS tiles that only have the horizontal datum defined. Without knowing the vertical datum, you do not really know the height. Also, putting the vertical datum inside the LAS file rather than on a yellow sticky note on the DVD is far more appropriate. (And what happens when the yellow sticky note looses its sticky?)
Local governments ask, what can I use to work with my Lidar data? I said earlier, “Commercial tools to process ‘raw’ Lidar imagery are not easy to use, nor are they currently cost effective. ERDAS, the inventor of commercial remote sensing, has historically been the company to introduce cost-effective remote sensing tools into the market-place.” Now, ERDAS is presenting tools to convert your Lidar data to raster for further processing.
But why raster?
- Raster conversion of the point cloud can shrink LAS file to a IMG file ¼ the file size of the LAS (you only store x & y values once in the raster UL corner). But that is not the main reason, as I suggest keeping the LAS files in backup for the day when point processing engines are cost effective.
- The files being converted into raster allows a visual QA/QC of the data.
- Open Lidar collections up to the ERDAS community in a very cost effective manner (no additional costs). Once the data are in raster, the Spatial Modeler, Model Maker, Classification, Expert Classification, and Objective are available to the customer to process the data.
6 comments:
Wonderful post as always Paul and great to see these new tools in IMAGINE 2011. I agree that many, if not most, local governments fail to capitalize on their LiDAR data. I also agree that raster is the optimal format for making LiDAR data accessible to the the vast majority of the GIS community. It can literally takes weeks to build a Terrain in ArcGIS for a medium-sized county. I believe it is very important to for one to visualize the point cloud prior to converting to raster forat. Going straight to raster without understanding the point cloud attributes, particularly how a contractor may have classified "bad" points can lead to less than desirable surface models.
I question the need for local governments to have a point cloud classification that goes beyond "ground." Classifying points as "buildings" and the various vegetation categories adds quite a bit of cost. This cost would be justifiable if it was easy to construct 2D or 3D features from a classified point clouds, but this capability is not within reach of the majority of local governments. Furthermore, feature extraction is greatly improved, in my opinion, if imagery is used on combination with LiDAR.
Jarlath, I agree with your assessment of the cost of the data providers classifying the Lidar data. I have run into some datasets where the data providers classified some of the water, buildings and trees as part of their process to pull out the ground points. This by-product effort has value that can be used.
Yet as you point out, local governments can combine the Lidar with their ortho-imagery for feature extraction.
The uniqueness of working with LIDAR data is the infinite possibilities of information that can be extracted from each data set. Working as a major vendor on large government LiDAR contracts over the past 10 years, I have started to see a cultural shift. What started as initial requests for DEM applications of the technology has morphed into request for completely brand new applications. The only limit being the imagination and need of the end user.
LiDAR technology has become main stream. What was once voodoo science is now a regularly demanded product in most mapping contract proposals. In particular government end users are more aware now of what information they can extract from LiDAR data sets. The issue lies in the ability to cost effectively extract this information from the data sets. On top of this constraint government end user are still struggling with the access and management of the information they are provided.
Current industry standards have relied on mathematical classification algorithms to extract information from the raw point clouds. These algorithms are only robust enough to take feature extraction to a specific percentage of accuracy depending on the information extracted. To create a finished product requires manual editing of the data. The time to clean up the imperfections is the most time consuming part of the LiDAR processing.
In most government contracts the budget described will not allow for all the necessary information to be properly classified and edited. This is why most contracts simply ask for a finished ground DTM as the final product. Furthermore the government end user may not have the capabilities to manage all the data provided if more information is requested.
As the sensor technology improves and greater repetition rates are achievable we compound the problem by again providing more information than can be properly digested by the end user. Plus we increase the time necessary to extract the information.
Suppliers should focus more on the tools used to work with the data and help create innovative was to manage the information. By developing better tools we will free up our local governments to focus on their mapping problems instead of creating new ones.
While I'll leave it to others to debate LAS utility to governments, it's worth pointing out the growing number of counties across the US that do have LIDAR coverage. They are listed on Wikipedia as part of the National LiDAR Dataset article, at this link
Thanks Anon, but Wikipedia lives down its reputation, again. There is no Santa Rosa County in South Carolina. But, my family owns property in Richland County South Carolina, and the county has Lidar that they sell, but to photogrammetrists and surveyors only. http://www.richlandmaps.com/index.html#data
Nice point. Getting only the raster info is like getting only photo prints from your wedding photographer but not the Digital pictures. (-:
Bt demanding the LAS files as well and providing access to it, local governments can harness the imagination and creativity of the people to derive interesting information beyond elevation from the data. With libLAS, LAStools, and other free packages anyone can post-process the LAS files for their purposes.
Post a Comment