Multiple attribute decision making for individual tree detection using high-resolution laser scanning |
| |
Authors: | Giovanni Forzieri Leonardo Guarnieri Enrique R Vivoni Fabio Castelli Federico Preti |
| |
Institution: | aDipartimento di Ingegneria Civile e Ambientale, University of Florence, Italy;bDepartment of Earth and Environmental Science, New Mexico Institute of Mining and Technology, Socorro, NM, USA;cDipartimento di Ingegneria Agraria e Forestale, University of Florence, Italy |
| |
Abstract: | A canopy height model (CHM) is a standard LiDAR-derived product for deriving relevant forest inventory information, including individual tree positions, crown boundaries and plant density. Several image-processing techniques for individual tree detection from LiDAR data have been extensively described in literature. Such methods show significant performance variability depending on the vegetation characteristics of the monitored forest. Moreover, over regions of high vegetation density, existing algorithms for individual tree detection do not perform well for overlapping crowns and multi-layered forests. This study presents a new time and cost-efficient procedure to automatically detect the best combination of the morphological analysis for reproducing the monitored forest by estimating tree positions, crown boundaries and plant density from LiDAR data. The method needs an initial calibration phase based on multi attribute decision making-simple additive weighting (MADM-SAW). The model is tested over three different vegetation patterns: two riparian ecosystems and a small watershed with sparse vegetation. The proposed approach allows exploring the dependences between CHM filtering and segmentation procedures and vegetation patterns. The MADM architecture is able to self calibrate, automatically finding the most accurate de-noising and segmentation processes over any forest type. The results show that the model performances are strongly related to the vegetation characteristics. Good results are achieved over areas with a ratio between the average plant spacing and the average crown diameter (TCI) greater than 0.59, and plant spacing larger than the remote sensing data spatial resolution. The proposed algorithm is thus shown a cost effective tool for forest monitoring using LiDAR data that is able to detect canopy parameters in complex broadleaves forests with high vegetation density and overlapping crowns and with consequent significant reduction of the field surveys, limiting them over only the calibration site. |
| |
Keywords: | Single-tree identification Forest monitoring Image segmentation LiDAR Remote sensing Decision making Simple additive weighting |
本文献已被 ScienceDirect 等数据库收录! |
|