Download FUSION

Overview of LIDAR research in the Pacific Northwest Research Station

Getting Started with FUSION

This document provides basic information to help you understand and explore the capabilities of the FUSION/LDV visualization system. It is not a comprehensive user’s guide.

What is LIDAR?

Overview of the FUSION/LDV visualization system

Example data


What is LIDAR? (top)

A LIDAR collection system uses a powerful laser sensor comprised of a transmitter and receiver, a geodetic-quality Global Positioning System (GPS) receiver and an Inertial Navigation System (INS) unit. The laser sensor is precision mounted to the underside of an aircraft. Once airborne, the sensor emits rapid pulses of infrared laser light, which are used to determine ranges to points on the terrain below.

Most LIDAR systems use a scanning mirror to generate a swath of light pulses. Swath width depends on the mirror's angle of oscillation, and ground-point density depends on factors such as aircraft speed and mirror oscillation rate. Ranges are determined by computing the amount of time it takes light to leave an airplane, travel to the ground and return to the sensor. A sensing unit's precise position and attitude, instantaneous mirror angle and the collected ranges are used to calculate 3-D positions of terrain points. As many as 10,000 positions or "mass points" can be captured every second.

LIDAR technology offers fast, real-time collection of 3-D points that are accurately geo-referenced. The parameters of flying height, swath angle, scanner rate, flight-strip side lap and aircraft velocity determine the point density as a system moves through the air, and these parameters are tailored to accommodate project requirements. An aircraft flies a regular pattern over a project area, for example, and a focused eye-safe infrared laser sends a variable number of pulses (10-100 KHz is common) to the ground in a fan array across the flight path.

Schematic of a typical LIDAR system.

When a pulse hits the ground, the beam's footprint varies between 0.5 and two feet, depending on flying height. The reflected light is collected by a receiver, and the interval between transmission and reception also is computed. A LIDAR system can discriminate among multiple returns from each pulse, simultaneously surveying the canopy top and terrain. Multiple returns also can be used to determine intermediate surfaces such as treetops and power lines. In a treed area, for example, the first return may locate the top of a tree, while the last return ideally locates the ground beneath the tree canopy. Multiple returns in between may represent branches, etc. All manmade and natural ground features are surveyed, including trees, buildings, cars, etc.

Overview of the FUSION/LDV visualization system (top)

The visualization system consists of two main programs, FUSION and LDV (LIDAR data viewer), implemented as Microsoft Windows applications coded in C++ using the Microsoft Foundation Classes. The primary interface, provided by FUSION, consists of a graphics display window and a control window. The FUSION display presents all project data using a 2D display typical of geographic information systems. It supports a variety of data types and formats but requires that all data be geo-referenced using the same projection system and units of measurement. LDV provides the 3D visualization environment, based on OpenGL, for the examination of spatially-explicit data subsets.

In FUSION, data layers are classified into six categories: images, raw data, points of interest, hotspots, trees, and surface models. Images can be any geo-referenced image but they are typically orthophotos, images developed using intensity or elevation values from LIDAR return data, or other images that depict spatially explicit analysis results. Raw data include LIDAR return data and simple XYZ point files. Points of interest (POI) can be any point, line, or polygon layer that provides useful visual information or sample point locations. Hotspots are spatially explicit markers linked to external references such as images, web sites, or pre-sampled data subsets. Tree files contain data, usually measured in the field, representing individual trees. Surface models must be in a gridded format that represents either a ground surface or other surfaces of interest such as the top of a forest canopy. FUSION uses the PLANS format for its surface models and provides utilities to convert a variety of formats into the PLANS format. The current FUSION implementation limits the user to a single image and a single surface model, however, multiple raw data, POI, and hotspot layers can be specified.

The FUSION interface provides users with an easily understood display of all project data. Users can specify display attributes for all data and can toggle the display of the five data types. The entire raw data layer can be rendered, however, it generally requires excessive time and is often left as a hidden layer.

FUSION allows users to quickly and easily select and display subsets of large LIDAR data. Users specify the subset size, shape, and the rules used to assign colors to individual raw data points and then select sample locations in the graphical display. LDV presents the subset for the user to examine. Subsets include not only raw data but also the related portion of the image and surface model for the area sampled. LDV provides the following data subset types:

·        Fixed-size square,

·        Fixed-size circle,

·        Variable-size square,

·        Variable-size circle,

·        Variable-width corridor.

Subset locations can be “snapped” to a specific sample location, defined by POI points. to generate subsets centered on or defined by specific locations. Elevation values for LIDAR returns can be normalized using a local surface model prior to display in LDV. This feature is especially useful when viewing data of forested regions in steep terrain as it is much easier to examine returns from vegetation after subtracting the ground elevation.

LDV strives to make effective use of color, lighting, glyph shape, motion, and stereoscopic rendering to help users understand and evaluate LIDAR data. Color is used to convey one or more attributes of the LIDAR data or attributes derived from other data layers. For example, individual returns can be colored using values sampled from an orthophoto of the project area to produce semi-photorealistic visual simulations. LDV uses a variety of shading and lighting methods to enhance its renderings. LDV provides point glyphs that range from single pixels, to simple geometric objects, to complex superquadric objects. LDV operates in monoscopic, stereoscopic, and anaglyph display modes. To enhance the 3D effect on monoscopic display systems, LDV provides a simple rotation feature that moves the data subset continuously through a simple pattern (usually circular). We have dubbed this technique “wiggle vision”, and feel it provides a much better sense of the 3D spatial arrangement of points than provided by a static, fixed display. To further help with understanding LIDAR data, LDV can also map orthographic images onto a horizontal plane that can be positioned vertically within the cloud of raw data. Surface models including “bare-earth” and canopy models are rendered as a shaded 3D surface and can be textured-mapped using an orthographic image.

FUSION and LDV have several features that facilitate direct measurement of LIDAR data. FUSION provides a “plot mode” that defines a buffer around the sample area and includes data from the buffer in a data subset. This option, available only with fixed-size plots, makes it easy to create LIDAR data subsets that correspond to field plots. In the context of this paper, “plot mode” lets us measure tree attributes for trees whose stem is within the plot area using all returns for the tree including those outside the plot area but within the plot buffer. The size of the plot buffer is usually set to include the crown of the largest trees expected for a site. When in “plot mode”, FUSION includes a description of the fixed-area portion of the subset so LDV can display the plot boundary as a wire frame cylinder or cube.

LDV provides several functions to help users place the measurement marker and make measurements within the data cloud. The following “snap functions” are available to help position the measurement marker:

·  Set marker to the elevation of the lowest point in the current measurement area (don’t move XY position of marker)

·  Set marker to the elevation of the highest point in the current measurement area (don’t move XY position of marker)

·  Set marker to the elevation of the point closest to the marker (don’t move XY position of marker)

·  Move marker to the lowest point in the current measurement area

·  Move marker to the highest point in the current measurement area

·  Move marker to point closest to the marker

·  Set marker to the elevation of the surface model (usually the ground surface)

The measurement marker in LDV can be elliptical or circular to compensate for tree crowns that are not perfectly round. The measurement area can be rotated to better align with an individual tree crown. Once an individual tree has been isolated and measured, the points within the measurement area can be “turned off” to indicate that they have been considered during the measurement process. This ability makes it much easier to isolate individual trees in stands with dense canopies.

Example data (top)

The example project contains 4.8 million data points covering about 295 acres (120 ha). The study area is part of a project designed to study the effect of several forest management alternatives. The site is located in western Washington near Olympia. The Washington Department of Natural Resources, Resource Mapping Section produced the digital orthophoto for the project using a softcopy photogrammetry system (Socket Set). The source imagery was 1:12,000 color aerial photography. Orthorectification was accomplished using a pseudo-canopy surface model developed using autocorrelation techniques. The final image used a 1-foot (0.3m) pixel.

A small footprint, discrete return LIDAR system was used to map the study area in the spring of 1999. The contractor used a Saab TopEye LIDAR system mounted on a helicopter to collect data over the study site. Table 1 presents the flight parameters and instrument settings for the data acquisition. Each return includes the pulse number, return number for the pulse (up to four returns were recorded per pulse), X, Y, elevation, off-nadir angle and return intensity.


Table 1.Flight parameters and scanning system settings.


Flying height

200 m

Flying speed

25 m/sec

Scanning swath width

70 m

Forward tilt

8 degrees

Laser pulse density

4 pulses/m2

Laser pulse rate

7,000 points/sec

Maximum returns per pulse


Footprint diameter

40 cm

Several “point-of-interest” and “hotspot” files have been included in the project. The green lines are treatment area boundaries, the yellow dots are survey points used to assess the accuracy of the LIDAR-derived terrain model, and the blue information markers are hotspots linked to photos of the forest conditions for treatment areas.

The terrain model was created using ground return points identified by the LIDAR provider. The model is a 5- by 5-foot grid (1.524- by 1.524-meter). The terrain model has been compared to a network of ground survey points and found to have an RMSE of 22cm.

Using FUSION/LDV (top)

To start using the FUSION/LDV system, launch FUSION and load the example project named “demo_4800K.dvz” using the File…Open menu option. The default sample options are set to use a stroked box. LIDAR returns will be colored according to their height above ground.

To extract and display a sample of data, stroke a rectangle using the mouse. The status display in the lower left corner of the FUSION window shows the size of the stroked area. Try for a sample that is about 250 feet by 250 feet. This should yield a sample of about 22,000 points. After a short time, the data subset should show up in LDV. The first time, you select a sample in FUSION, the LIDAR data files will be indexed so the first sample may take a minute or two to display.

To manipulate the LIDAR data in LDV, it is easiest to image that the data is contained in a glass ball. To rotate the data, use the mouse (with the left button held down) to roll the ball and thus manipulate the data. As you move the data, LDV may display only a subset of the points depending on the size of the sample.

Options in LDV are accessed using the right mouse button to activate a menu of options. There are lots of options and the best way to understand them is to try them. Additional functions are assigned to keystrokes. These functions are described when you click the “About LDV” button located in the lower left corner of the LDV window.  Keyboard shortcuts are also listed on the right mouse menu.