E3De creates photorealistic 3D representations of LiDAR data. Using E3De, power lines are automatically extracted from LiDAR data with one click. Power line data is then stored as an independent, portable dataset for further analysis in a GIS. Point clouds can not only be represented with RGB values provided in the raw LAS file, but accurate digital ortho-images are also derived with a single click. Using the results from E3De, it is easy to determine where future trouble spots will occur by locating the trees that grow directly under and adjacent to the power lines.
Fig. 2
E3De creates photorealistic 3D representations of LiDAR data. Using E3De, power lines are automatically extracted from LiDAR data with one click. Power line data is then stored as an independent, portable dataset for further analysis in a GIS. Point clouds can not only be represented with RGB values provided in the raw LAS file, but accurate digital ortho-images are also derived with a single click. Using the results from E3De, it is easy to determine where future trouble spots will occur by locating the trees that grow directly under and adjacent to the power lines.
Fig. 3
ERDAS APOLLO Web Client displaying LiDAR and imagery data from the Swiss municipality Romanshorn. The LiDAR data was collected by a Leica ALS50 along with color imagery from a Leica ADS40. Other delivery options of data are by download or delivery via web services into other web client applications. Courtesy of ERDAS.
Fig. 4
Visualize point clouds across or along track profile view using IMAGINE Point Cloud Tool. Courtesy of ERDAS.
Fig. 5
New Building Detection using IMAGINE Objective with multi-date LiDAR data to show the detection of new building activity to determine legality and tax status. The LiDAR is shown as a sun-angle shaded representation at two dates and the IMAGINE Objective results are shown as a vector overlay on the raw height data. Courtesy of ERDAS.
Fig. 6
Terrestrial point cloud in Phoenix - visualize airborne or terrestrial point clouds in 3D. Courtesy of ERDAS.
Like raster and vector data many years ago, 3D clouds of billions of LiDAR points — which can be colored and very realistic — are an exciting new data type. However, they pose many software challenges, and software vendors are working hard to develop new software to manage, process, visualize, and extract features from point clouds.
The first challenge posed by LiDAR to computer software and hardware is the huge size of the files generated by ever more powerful scanners. Often, the limiting factor is the operating system. For example, the 32-bit operating system can handle only files of up to a few gigabytes, and most systems are not designed for such large datasets. Though software providers have made progress in how they handle the display of point clouds, experts agree that this remains a challenge.
More recently, a new challenge emerged: managing the thousands of files of point cloud data generated by the explosion in the use of terrestrial LiDAR scanners, making these files available to users across an organization, and integrating them across different workflows. The final two challenges are fusing LiDAR point clouds with data from other remote sensing devices and further automating measurements and feature extraction.
Visualizing Lidar Data
Software vendors have taken different approaches to enabling their software to visualize LiDAR data.
People want to visualize LiDAR data as points and they want the ability to do it in 3D, says Brad Skelton, CTO of ERDAS. If you simply load all the points and display them, he explains, you quickly run out of memory. Enabling users to perform the tasks they want to perform with these very large datasets requires indexing the files and providing an efficient paging mechanism.
“The conventional solution,” Skelton says, “is to tile LiDAR datasets into smaller pieces that will fit into memory and perform your processing in a piece-wise manner. We are trying not to do piece-wise processing, but instead provide the user with a single, large dataset and continuous experience and still have everything be as fast as it is with the entire dataset in memory at once.” The company’s competitive advantage, he argues, is going to be its ability to deal with a large dataset, as well as offering both a server product and a suite of desktop products that can deal with LiDAR data. Next year, it will release a new version of its software that will fully support LiDAR as a data type.
We have many tools to make 3D measurements from imagery that can be applied to make 3D measurements from point cloud data
— Zhang of BAE Systems
Bentley began by embedding point cloud management capability into MicroStation, its CAD 3D editing platform, and ProjectWise Navigator, its tool for collaboration and design review. Because most of the company’s industry applications share the same technologies as MicroStation, most of the desktop products that came after MicroStation V8i (SELECTseries 2), released last year, including Bentley Map, OpenPlant, and Bentley Architecture, are able to visualize point clouds. “We bring point clouds inside our engineering tools, whether it is generated by aerial LiDAR or by static or mobile terrestrial scanners,” says Benoit Fredericque, a product manager for Bentley. “For us, it is all about our capability to handle huge point clouds properly.”
Esri’s ArcGIS desktop and server applications, all of which share the same underlying technology, can ingest, process, and display LiDAR data. The trick, explains Clayton Crawford, one of the company’s product engineers, is coming up with an efficient I/O stream for moving that data around and accessing it efficiently. “For some applications, working on thinned data is OK, while other applications require full-resolution processing.”
“Our next release, ArcGIS 10.1, which is already in beta, adds significant capabilities for LiDAR, including native support for LAS,” says Crawford, referring to the industry standard format for LiDAR data. “This means LiDAR data can be managed, viewed, updated, and shared, all while remaining in its native format. We’ll be introducing a new data type called a LAS dataset that makes it easy to access your data quickly. You just tell it what LAS files belong to a project and it manages the collection as one dataset for you.”
“Our Mosaic dataset, which is used to manage and share imagery,” he adds, “is being enhanced to read LAS directly. It will perform fast, on-demand rasterization of LAS points and can handle many LAS files from multiple projects. Through its support of image and elevation services, the Mosaic dataset can also be used to publish LiDAR either in source LAS form or as derivative rasters. It provides a very effective means of managing large collections of LiDAR in a way that’s similar to managing lots of imagery.”
ITT VIS’ flagship image processing and analysis software product, ENVI, can extract elevation and intensity values from LiDAR point clouds. IDL, which is an extensible platform for ENVI and a programming language, gives users API access to LiDAR point cloud data and also has a broad range of tools to visualize LiDAR data and process it volumetrically. The company’s latest product, E3De, scheduled for release October 14, is completely specialized for LiDAR data. ITT VIS is initially releasing it as a full stand-alone application, but it will ultimately also serve, optionally, as a module to ENVI, says Peter McIntosh, the company’s Manager of Industry Solutions.
Managing Lidar Data
Bentley is now working on the management of the point cloud on the server side, for which it plans to use Bentley ProjectWise. Meanwhile, on the desktop, it is adding advanced capabilities related to point clouds inside Bentley Descartes, which has been used for more than a decade as a raster editing product that enhances MicroStation workflows. “We are also developing applications to make point cloud data available on iPads and on the Web,” says Fredericque. Bentley plans to release in early 2012 ProjectWise V8i (SELECTseries 4), which will provide new capabilities for managing point cloud data.
One of the first ERDAS products to deal with LiDAR data was ERDAS APOLLO, which is the company’s enterprise data manager. “A couple of years ago,” says Skelton, “we added to ERDAS APOLLO the ability to catalog and manage LiDAR datasets, so that it can discover your LiDAR datasets, extract various pieces of geospatial metadata from them, put them in the catalog, and give you a means of executing spatial and attribute queries to find the datasets that are pertinent to the areas on which you are working.”
Fusing Data From Lidar and Other Sensors
“We are focusing our energies on integrating LiDAR data with the other modalities,” says McIntosh, “to create a holistic and integrated dataset that tells us the volumetric properties that you get from LiDAR, the surface properties that you get from SAR (synthetic aperture radar), and the material properties that you get from optical and hyperspectral. That overall data fusion is the target and the goal for the future. Looking forward, I see the holistic integration of all of the modalities into a unified resource as the biggest challenge with LiDAR.”
ITT VIS has worked in the past several years to integrate ENVI with the Esri product line, making image analysis tools available directly from the Arc environment. The road map for E3De, McIntosh says, will include the integration for the GIS user as well. “As with ENVI, it will also have full API programmatic access and extensibility.”
BAE Systems is planning to add to future releases of its SOCET GXP the capability to register multiple strips of LIDAR data with each other and LIDAR data with imagery. It is also working to fuse imagery and terrestrial LIDAR data — which, points out Dr. Bingcai Zhang, an Engineering Fellow with BAE Systems, are now denser, less expensive, and more abundant than a few years ago due to advances in ground sensor hardware.
Not surprisingly, Esri is focusing on integrating LiDAR data into GIS, so that it is not being processed in a specialized application but rather in a more general purpose, one for dealing with all kinds of geospatial data. “So we have the integration of the processing, analysis, and display of LiDAR data along with other geospatial data and that brings a lot of power to the end user,” says Crawford.
Automatic Feature Extraction
To enable users to get the most out of LiDAR data, software must be able to extract geometries, features, and measurements automatically. For example, it must recognize such 3D objects as houses, trees, and, if the point cloud is sufficiently dense, cars. This is especially valuable for certain industries, such as transportation and building. “We have seen very interesting progress in this automation with aerial LiDAR, but in other areas of the infrastructure world, the level of automation is pretty low,” comments Fredericque.
SOCET GXP v4.0, which BAE Systems plans to release in early 2012, will include automatic feature extraction. To increase efficiency and reduce manual entry for end users, the underlying algorithm will automatically compute the dimensional attributes of a 3D object, for about 20 attributes. This capability builds on the company’s 16 years of experience in developing photogrammetric algorithms for SOCET SET and SOCET GXP. “We have many tools to make 3D measurements from imagery that can be applied to make 3D measurements from point cloud data,” says Zhang. “That is our legacy and our competitive advantage.”
ITT VIS is also adding tools that automatically extract information from files of various data types. “We already provide tools for the automatic extraction of building footprints, trees, power lines, etc.,” says McIntosh. “We are going to see more and more automated tools that span across industries. So, for example, someone in agriculture will say, ‘I want to see crop height for my corn in July,’ and it is going to be much more point-and-click, decision-ready information out of these. We are going to see that in forestry, urban planning, etc. — more and more advanced processing. LiDAR brings a whole 3D element into the remote sensing picture, where we can use things like hill shades — slope, aspect, elevation. It just gives us yet another element to discriminate and identify specific features or information that people want.”
LPS, ERDAS’ photogrammetry software, has a very high resolution terrain extraction component, eATE, which does optical stereo point extraction. It allows users to extract point cloud data at LiDAR densities and output the data as LAS files. Therefore, Skelton points out, in addition to ingesting, processing, and displaying LiDAR data, LPS can also produce LiDAR-like data.
Beyond Automatic Feature Extraction
Many people think of a point cloud only as a kind of intermediate form of data from which to generate geometry. While this is a good application area, Fredericque argues, a point cloud is also very valuable when used ‘as is.’ “For example,” he says, “you can use it as a kind of 3D base map or, in some contexts, as a 3D model. An industrial plant might not have a 3D model; it might just want to replace a piece of equipment. If you consider point clouds only as an intermediate data type, the extraction of all the features becomes a mandatory and demanding process. Another way to address that challenge is to use the point cloud ‘as is,’ as a background 3D model, and use tools to remove the points that correspond to the piece of equipment that you plan to remove. This way, you can manipulate the point cloud. You can isolate a piece of equipment and replace it with a new one. That kind of workflow and usage is not frequently mentioned or investigated, but I believe it makes a lot of sense.”
Clouds of billions of LiDAR points — often colored and very realistic — bring a new 3D element into the remote sensing picture and require new software tools. Software vendors, already challenged by the huge size of the point cloud files created by ever more powerful scanners, are developing new ways to manage the thousands of files generated by the explosion in the use of terrestrial LiDAR scanners, to make them available to users across an organization, and to integrate them into different workflows. They are also embedding point clouds into engineering tools and GIS and developing new ways to automate the extraction of geometries, features, and measurements. The ultimate goal is to fuse volumetric properties derived from LiDAR data, surface properties derived from radar, and material properties derived from optical and hyperspectral data to create a single, integrated dataset.
