Detecção de árvores individuais e cubagem de Pinus elliottii utilizando LiDAR UAV e MLS

Abstract

This study aimed to evaluate, improve, and validate computational methodological approaches based on LiDAR (Light Detection and Ranging) point clouds embedded in mobile and aerial devices to stimate dendrometric parameters in Pinus elliottii stands. The dissertation is divided into two chapters. The first evaluated the performance of Individual Tree Detection (ITD) algorithms using UAV (Unmanned Aerial Vehicle) LiDAR data, contrasting a methodology based on point cloud Cross-Sectioning (STAD) with clustering via the DBSCAN (Density Based Spatial Clustering of Applications with Noise) algorithm against traditional canopy based counting methods. The STAD method achieved the lowest RMSE (8.43% in circular plots and 11.11% in rectangular plots) by focusing on the stem section below the crowns. The second chapter addressed non-destructive stem profiling using MLS (Mobile Laser Scanning) LiDAR. The traditional destructive stem profiling method using a caliper was compared with two methods of diameter collection from the LiDAR point cloud: the first, based on Cross Sectioning and Contours by Angular Classes (STAC) in the R language, evaluating different metrics and angular classes; and the second, processed in the CloudCompare software using the 3Dfin plugin. The STAC method parameterization (72 angular classes and minimum point metric) obtained the highest accuracy (RMSE of 5.94% and BIAS of 2.5%). It is concluded that the STAD method overcomes the limitations of classic canopy-centered algorithms, while the STAC routine proves operational viability and greater adherence to the contour of each stem section compared to the traditional method and algorithms consolidated in the literature.