AN ASSESSMENT OF LIDAR VISUALIZATION TECHNIQUES TO AID IN AUTOMATED FEATURE EXTRACTION

Various methods can be used to visualize lidar data and make the automated feature extraction process possible. The goal of this project is to compare several different visualizations (i.e red relief, advanced hillshade, local relief, topographic position index) derived from lidar data to determine if one method or a combination of methods can enhance the efficiency of feature extraction. Various commercial and shareware programs including ArcGIS, SAGA GIS, and QGIS were used to construct the lidar visualizations that enhance the features of interest. For this study the features of interest are landslides and slumps along rivers throughout Minnesota. These visualizations were incorporated into visual search experiments given to several volunteers to gauge which of the techniques might aid the human eye in picking out landslides given only simple instruction. We will report preliminary data from a visual search paradigm designed to quantify the efficacy of these visualizations for feature extraction using response time, visual search behavior, and search accuracy. The visualizations will also be applied to a computer-based semi-automated feature detection algorithm to gauge which technique can help the computer extract features with the most efficiency. The results of the human survey and the computer testing will ultimately be compared to determine which visualization or group of visualizations (if any) are best suited for feature detection and will possibly aid future work in refining automated feature detection methodology.