VP Special Projects
Teledyne Optech
In 2014, Teledyne Optech launched the world’s first airborne multispectral lidar sensor, the Optech Titan. Bathy lidar sensors that use 532 nm (green) and 1064 nm for the airborne determination of water depths have existed for some time now. Some researchers analyzed data from these two wavelengths with other data collections at 1550 nm, but these three wavelengths were not collected at the same time. With the Optech Titan, the sensor operates simultaneously at all three wavelengths and collects data at up to 300 kHz in each of three channels: 1064 nm at nadir, 1550 nm looking 3.5° forward, and 532 nm looking 7° forward. The combined data acquisition rate of 900 kHz creates dense point clouds suitable for detailed analyses.
Different types of land cover react differently to each of the three wavelengths, and these differences can be used for land-type classification. The Normalized Difference Vegetation Index (NDVI) is a well-known parameter for characterizing the amount and health of vegetation, first invented in 1974 for digital imagery. It uses the difference between red and near-infrared (NIR) wavelengths. Twenty two years later, the Normalized Difference Water Index (NDWI) was created, which uses green and NIR wavelengths. The NDWI is used to distinguish water from vegetation and soil, and is mainly utilized on passive digital imagery.
Now, an active Normalized Difference Feature Index (NDFI) can be defined as the difference in intensity between any two of the three wavelengths from the Optech Titan lidar. These three NDFIs are less dependent upon cloud cover and ambient light conditions because the data is derived from an active sensor. Indeed, the lidar data can be acquired at dusk or nighttime, which is impossible for passive sensors. Much work has also been done to normalize the measured intensities to account for incident angles, ranges and other effects.
Standard classification techniques have been applied and adapted to the lidar data by many researchers, but most significantly lately by Ryerson University. They were able to achieve land cover classifications with over 90% accuracy using the Titan data. In addition to the NDFIs, other lidar parameters such as height and number of returns can be analyzed for further accuracy.
An important subset of land-cover classification is the delineation of land from water. This is critical for flood mapping and hydrological studies. In addition to being a 900-kHz terrain mapper, the Optech Titan also serves as a 300-kHz shallow-water bathy lidar with its 532-nm channel. Having automatic algorithms to distinguish water from land means that water depths can be derived quickly, without needing to outline the water bodies manually. It is also more accurate to use the IR wavelengths to define the water surface location because the green wavelength is well known to have a bias. The two IR wavelengths (especially the 1550 nm) are the key to accurate land/water discrimination. Ryerson University has also made great advances in this area, including the challenging regime of riverine systems. Comparative images are shown for Optech Titan data collected in Tobermory, Ontario. The land/water classification was automatic, based on lidar data alone.
Forestry applications are a natural extension of multispectral lidar data. Other researchers have been using the data to distinguish tree types and even species. More results are forthcoming in this area of research, which leads to very efficient surveys of forest health and viability.
