First outdoor tests of the mobile hyperspectral LiDAR, Dec 2012
Applications: Chlorophyll estimation using hyperspectral LiDAR
The performance of the prototype 8-channel full waveform hyperspectral LiDAR was validated in chlorophyll estimation. The chlorophyll estimation was performed by calculating various vegetation indices to ten Scots pine shoots and oat samples with varying health. Reference data was acquired by accurate laboratory concentration analysis. The results indicate that the hyperspectral LiDAR instrument holds a good capability to estimate vegetation biophysical parameters such as the chlorophyll concentration.
Regression plot between the laboratory analyzed chlorophyll concentrations of ten Scots pine shoots and the Modified Chlorophyll Absorption Index using reflectance at 750 and 705 nm (MCARI[750, 705]).
New results 2012:
The latest results of the hyperspectral LiDAR development were presented at the XXII Congress of the ISPRS in Melbourne, the presentation can be downloaded here.
New results 2011: development of a 8-channel hyperspectral range finder
A prototype 8-channel full waveform active hyperspectral LiDAR is ready for research use. First measurements have been taken from a 2 meter high Norway spruce, showing possibility to determine various vegetation indexes including NDVI, position of red edge and water concentration index. Instrument development and data analysis are still under continuous research providing constantly new results and more refined measurement equipment.
Prototype of a 8-channel full waveform active hyperspectral LiDAR.
First demonstrations on the remote sensing of vegetation
Hypersperctral LiDAR point cloud.
Spectral indices of a Norway spruce target.
Previous steps in development 2009-2010
The idea of hyperspectral laser scanning is to use a supercontinuum source, i.e. ’white light lidar’. This is an entirely new concept in laser scanning and offers a possibility for a new level of technology. Active hyperspectral imaging is acquired simultaneously with topographic information, so instead of receiving one intensity value for every laser point, a full spectrum is available.
Above is the basic setup of hyperspectral measurement: supercontinuum laser source sends a pulse to the target, and returning laser radiation is collected to a spectrometer. Below is the result of first tests with hyperlaser scanner: a branch of a spruce was scanned with hyperlaser to obtain spectral responses of the tree. On the left is a photograph, in the middle the measured spectra presented in RGB colors, and on the right the normalized difference vegetation index (NDVI) computed using the spectrum.
Applications: Tree species classification from hyperspectral LiDAR data
Below are the example pictures of each of the tree species. Tree species are presented on different columns. LIDAR intensity images of each species are on the top row and images formed from hyperspectral data are on the bottom row. White dots in subimages are reference marks attached on the background canvas. The large white rectangle in spruce images is the Spectralon™ panel. The colour channels in the hyperspectral images were interpolated over a given band for each of the three channels: Red = 625–675 nm, Green = 525–575 nm, and Blue = 460–480 nm. The channels were chosen for visualization purpose only, so there is offset in the blue channel.
Mean, maximum, and minimum reflectance spectra over averaged tree data is shown below. All tree species had a significant variance in their averaged reflectance spectra, which is represented with the thin dashed and dash-dotted minimum and maximum lines.
Dec 2009/Jan 2010: a two-channel hyperspectral range finder prototype
The prototype rangefinder system is capable of processing multiple hits, from which it can extract multiple targets from a single echo.
Two-channel hyperspectral LiDAR echoes from two targets: Spectralon reference panel (last echo) placed behind the Norway spruce (Picea abies) sample (first echo). The intensities are represented by the echo amplitude, and the distance can be sampled from the difference in the time of flight sampled in the x-axis.