LIDAR (Light Detection And Ranging) is an optical means of measuring reflected light from distant objects to determine range, and from this information, to determine position. Coupled with accurate position and orientation systems, LIDAR systems can take accurate 3D measurements of objects and surfaces using high sampling densities. Applications include engineering, remote sensing, forestry, geomatics and more. Today, LIDAR systems can be divided into two major types: terrestrial and airborne. Terrestrial systems are operated at ground level scanning in a vertical direction while rotating about a vertical axis. Airborne systems are mounted in an aircraft looking down, scanning in the direction perpendicular to the flight.
Terrestrial LIDAR systems are typically mounted at a single location and collect points by rotating in both the horizontal and vertical axis. Repeated collects from multiple locations are then aligned and fused into a single large point cloud. Data collected are then used in applications such as engineering to map as-built structures (like the interior of factories) or even in archeological applications to make precise non-destructive measurement of artifacts. Typically, the data are collected, often resulting in billions of points, representing a more or less hemispherical collection of points.
Airborne LIDAR systems are mounted in an aircraft, scanning in one direction perpendicular to the flight of the aircraft in a push broom fashion. The data can be thought of as a long strip representing the surface of the underlying terrain. The application of these data is more oriented toward direct measurement of the earth's surface used in conjunction with GIS, remote sensing or photogrammetric applications, such as land cover mapping, hydrology analysis, site selection, etc. This article focuses on the processing and management of airborne LIDAR data in such geospatial applications.
A number of different tools exist, both commercial and open source, for processing airborne LIDAR data. The majority of these applications process LIDAR data stored in the LASer (LAS) file format. One of the first tools that anyone working with point data needs is some type of viewer. Because LIDAR data consist of collections of 3D points, these viewers are most often 3D tools that usually work with the assumption that all of the data can be read into memory at once. Since LIDAR data collections can be in the billions of points, it is easy to have LAS files which cannot be viewed because of memory constraints. This is also a common problem for many processing tools, which also make the assumption that all of the data can be first read into memory. This has led to a convention of cutting LIDAR data sets into tiles of a maximum size and then processing large datasets a tile at a time.
LAS Data File Format
LAS is a file format specification for the interchange of 3D point data, which is maintained by committees within the American Society for Photogrammetry and Remote Sensing (ASPRS) as directed by its board. The format is freely available for download from its website. It stores an x, y, z triplet per point. In addition, information such as the actual intensity of the return value, and user defined values, such as point classification, can be stored per point. The LAS format is widely adopted and used throughout the industry. One problem with the LAS format is that it is not spatially indexed and it does not have a provision for generalizations, both of which are problems when trying to work with very large data sets. Spatial indexing provides a means of locating all the points within a given area quickly without scanning the whole file and generalization allows for a representative subset of the points to be used for visualization at small scales.
Point Cloud vs. Contour vs. Grid
Airborne LIDAR data are most often used as a representation for a terrain surface. However, these data may not be useful in the form of discrete points, as many of the existing processing applications expect to be given a terrain as either a set of contours or as a raster grid, for example, a digital elevation model.
The form to use is often a topic of debate, with pros and cons for each. Using discrete points, one can vary the density with the complexity of the terrain, providing for smaller data sets. However, to work with such data, an application must be designed to create and interpolate triangulated irregular networks (TINs). If all the data are not read into memory, the TINs must be created in segments, introducing further complexity. If the data are converted into grids, they can be stored in common raster formats. As a raster file, the data are more readily accessible by many applications (USGS DEMs are a very common raster surface model). Another benefit of grids is that they are easier by their very nature to break into workable segments and can be readily served through several existing protocols
Managing LIDAR Datasets
Managing and serving LIDAR data is becoming a growing problem. Many organizations are accumulating large collections of LIDAR data, but they have a problem keeping track of what they have and providing it to their end users in an effective manner.
There are even some national LIDAR datasets now that consist of high-resolution LIDAR data of a country's terrain. Today, Denmark, Finland, Sweden and Switzerland have such datasets already complete or in progress. The U.S. is considering creating such a dataset, as described on the USGS website at http://lidar.cr.usgs.gov/. Many states in the U.S. are currently doing the same.
ERDAS's APOLLO manages, processes and delivers massive amounts of imagery and gridded datasets to end users. It can automatically discover imagery, harvesting its metadata and then deliver the data through widely used protocols such as Web Map Service (WMS), Web Coverage Service (WCS) or even Enhanced Wavelet Compression Protocol (ECWP). Because ERDAS APOLLO uses an aggregate data model, it serves up its datasets as individual granules or as composite datasets. An aggregate data model organizes related datasets into a hierarchy like folders, with the higher level aggregates acting like a virtual mosaic of the contained datasets. This, coupled with the "clip-zip-and-ship" capability makes data available to a very large number of existing applications.
ERDAS APOLLO capabilities are also available for LAS data via a plug-in. During the process of autonomous data discovery called crawling, the LAS files are examined to extract metadata (called harvesting) and a multi-resolution gridded version of the dataset is created and stored transparently to the user. Information such as a spatial reference system is read directly from the LAS file. Information such as the ground sample distance is automatically estimated by examining the data. All of these metadata are then stored in the APOLLO catalog. Once this is done, the LAS files can then be searched using metadata queries, and served using any of the raster protocols. This allows the LAS data to be used in any existing workflow already prepared for dealing with gridded terrains. For example, the LAS data can be viewed as shaded reliefs via the WMS service, and then included in any WMS aware Web service.
The images below show two different LAS datasets of Mount St. Helens viewed as shaded reliefs. The image on the left is prior to the eruption that took place during November 2004 and the image on the right is afterward, making the displacement from the lava dome visible.
Because the ERDAS APOLLO data model can automatically mosaic the granules of an aggregate into a single virtual dataset, multiple LAS datasets can be merged on-the-fly into a single large terrain layer. If an application requires the data to be present on the local workstation, then the user can choose to have those data (or a subset of them) delivered through the "clip-zip-and-ship" provisioning mechanism.
Another benefit of introducing the data as a raster is that some image processing tools can be used to extract information from the data. In this example, a spatial model has been used to detect and quantify change within the area of the two data sets from above.
LIDAR data, in the form of point clouds, are entering the mainstream as the "third" primary type of geospatial data, complementing the traditional raster and vector models. However, the practical adoption of these data is limited since many of the traditional tools are not yet point cloud ready. The density and semi regularity of LIDAR data tend to make these data readily gridded, which is a means of managing, delivering and consuming these data within a wide range of existing tools.