Quantificação automatizada de variáveis dendrométricas em plantios experimentais de Eucalyptus sp. utilizando VANT

Abstract

The objective of this work is to evaluate the accuracy of the use of Unmanned Aerial Vehicle (UAV) based on the structure from motion to support and automate the quantification of dendrometric variables in experimental plantations of Eucalyptus sp. The study area is located in the Campus of Engineering and Agricultural Sciences of the Federal University of Alagoas, located in the municipality of Rio Largo – AL. The data survey was carried out in two experimental areas of forest stand of the hybrid Eucalyptus urograndis (E. grandis X E. urophylla), clones TP361, I144 and VE41. Planting took place in August 2017 in an area of approximately 2.5 hectares. Field data collection was carried out from portions of the Eucalyptus population. Thus, the total height, diameter at breast height (DBH) and crown diameter of all individuals in each plot were measured. To obtain the aerial images, the UAV from the company DJI, model Phantom 4 PRO, with an RGB camera was used. The planning and execution of the flight plan was done through the Dronedeploy application. The flight was carried out around 12h, in full sun condition, in order to minimize the shading effects of clouds and from one individual to another. The images acquired through UAV were processed using the SfM (Structure from motion) in the Agisoft Photoscan software and, in this way, point clouds were generated for each plot obtained in the processing. From the point cloud, digital surface model (DSM) and terrain (DTM) models were generated, followed by the generation of the orthomosaic. Through the difference between the DSM and the DTM, the CHM (Canopy Height Model), a digital crown height model, was generated in the Qgis software. Note that the insertion of control points is essential to reduce errors in image adjustment. So far, it is clear that the products generated from the processing of UAV images with the applied flight configurations, such as MDS, MDT, CHM and orthomosaic, have high resolution, adequate location accuracy and are capable to providing information to assist the steps of the forest inventory. The estimation of dendrometric variables from data from aerial images proved to be feasible and the metrics generated through the lidR and itcSegment package showed a positive correlation between the estimated values and the values measured in the field. It is possible to generate estimates of dendrometric variables such as DBH and heights using measurements calculated from passive sensors embedded in UAVs. These findings in terms of quality of forest attributes using UAV support a promising future in the field of forestry resources for precision forestry.