Evaluating High-resolution Imagery and LiDar for Mapping Structures in the Wildland-urban Interface PDF Download

Author:
Publisher:
ISBN:
Category :
Languages : en
Pages : 26

Get Book

Book Description
This project tested remote sensing tools for locating man-made structures in the wildland-urban interface (WUI). Creating fire-suppression plans and responding to wildland fires become much easier when information on the location of critical structures is available. High-resolution digital imagery and lidar data were tested in two study areas. Vegetation ranged from sagebrush and scrub in Oregon to dense forest canopy in Montana. The tools were semi-automated, meaning that users had to interact with the data sets to interpret the structures. The digital imagery used in the study was 1-meter natural-color photography from the National Aerial Imagery Program (NAIP). The multireturn lidar data had 4 returns and a density of 1.7 points per square meter. The tests were completed in the following manner: 1) In the Oregon study area, Feature Analyst, a semi-automated feature-extraction software package developed by Visual Learning Systems (VLS) mapped structures using 1-meter natural-color imagery. The area is sagebrush-dominated urban wildland containing both dense subdivisions and dispersed buildings. 2) In the Montana study area, a lidar feature-extraction software package, VLSs LIDAR Analyst, mapped structures in a dense forest containing dispersed cabins. An independent accuracy assessment was completed for both study sites by comparing the results with manual image interpretation. Although the results for feature extraction using NAIP imagery were promising, roughly 50 percent of the structures were missed (omission error) and an additional 50 percent were wrongly delineated (commission error) in the Oregon study. In the heavily forested area in Montana, LIDAR Analysts analysis of the multireturn lidar data was mediocre because structures were confused with understory forest canopy. Based on these case studies, it appears that feature extraction using high-resolution imagery or multireturn lidar data is only partially effective. We recommend using heads-up image interpretation and manual digitizing for more accurate and timely results.