The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLII-3
https://doi.org/10.5194/isprs-archives-XLII-3-2047-2018
https://doi.org/10.5194/isprs-archives-XLII-3-2047-2018
30 Apr 2018
 | 30 Apr 2018

WATER MAPPING USING MULTISPECTRAL AIRBORNE LIDAR DATA

W. Y. Yan, A. Shaker, and P. E. LaRocque

Keywords: Land Cover Classification, Multispectral LiDAR, Optech Titan, Point Cloud Processing, SLIER

Abstract. This study investigates the use of the world’s first multispectral airborne LiDAR sensor, Optech Titan, manufactured by Teledyne Optech to serve the purpose of automatic land-water classification with a particular focus on near shore region and river environment. Although there exist recent studies utilizing airborne LiDAR data for shoreline detection and water surface mapping, the majority of them only perform experimental testing on clipped data subset or rely on data fusion with aerial/satellite image. In addition, most of the existing approaches require manual intervention or existing tidal/datum data for sample collection of training data. To tackle the drawbacks of previous approaches, we propose and develop an automatic data processing workflow for land-water classification using multispectral airborne LiDAR data. Depending on the nature of the study scene, two methods are proposed for automatic training data selection. The first method utilizes the elevation/intensity histogram fitted with Gaussian mixture model (GMM) to preliminarily split the land and water bodies. The second method mainly relies on the use of a newly developed scan line elevation intensity ratio (SLIER) to estimate the water surface data points. Regardless of the training methods being used, feature spaces can be constructed using the multispectral LiDAR intensity, elevation and other features derived from these parameters. The comprehensive workflow was tested with two datasets collected for different near shore region and river environment, where the overall accuracy yielded better than 96 %.