Banca de QUALIFICAÇÃO: JHIORRANNI FREITAS SOUZA

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
STUDENT : JHIORRANNI FREITAS SOUZA
DATE: 29/11/2023
TIME: 09:00
LOCAL: Sala 01 do PPGF
TITLE:

Use of an unmanned aerial vehicle (UAV) to survey weeds in forest plantations for the restoration of the Atlantic Forest

KEY WORDS:

image processing: remote sensing; weed competition; forest management

PAGES: 42
BIG AREA: Ciências Agrárias
AREA: Agronomia
SUBÁREA: Fitotecnia
SPECIALTY: Matologia
SUMMARY:

High-resolution spectral information can contribute to the identification of weeds and aid in the management of weed competition, in an instantaneous and non-destructive way, in plantations aimed at forest restoration. In this research, the objective is to develop digital models for weed identification with the use of Unmanned Aerial Vehicle (UAV), in a forest restoration area in the Atlantic Forest biome. The experiment will be conducted in an area of the Guapiaçu Ecological Reserve (REGUA), located in the municipality of Cachoeiras de Macacu - RJ. The experimental area will consist of four blocks (replicates) arranged in a randomized design, and six treatments: T1. Herbicide-free control; T2. Indaziflam directed jet; T3. Indaziflam "over the top"; T4. indaziflam + haloxyfop + mineral oil "over the top"; T5. haloxyfop + mineral oil "over the top", and T6.  Glyphosate directed jet, totaling 24 experimental units, which will be composed of five rows, with 20 seedlings of different native species, with planting spacing of 2.5m between rows and 2m between seedlings. The treatments will be applied with the aid of an electric knapsack sprayer, equipped with a bar of 4 spray tips model AVI 110.02, spaced 0.5 m apart, operating at a pressure of 2.5 Bar, at an application rate of 150 L ha-1. At 30, 60, 90 and 120 days after application (DAA) of the treatments, plant poisoning evaluations of native species will be carried out according to the EWRC scale (1984). Captures of the images (RGB) will be carried out with Mavic 3E model UAV every 30 days at flight heights of 20m, 40m and 60m, for a period of 12 months. Phytosociological evaluations will be carried out at each moment of flight to survey the weeds. Photogrammetric processing will be obtained through Agisoft Metashape, for the elaboration of orthomosaics with MDT, MDS, CHM image layers and spectral indices. ArcGIS Pro software will be used to build the Deep Learning deep learning model, which will be trained to differentiate native species and weed groups (monocots and dicots). Deep learning techniques will be applied using Convolutional Neural Networks (NCR). The accuracy of the model will be estimated by evaluating the ability to identify native species and weed groups from Deep Learning learning, through four accuracy metrics: Average Precision (AP), F1 score, mean Mean Precision (mAP) and Precision x Recall curve.

COMMITTEE MEMBERS:
Presidente - 1905333 - AROLDO FERREIRA LOPES MACHADO
Externo ao Programa - 2145654 - BRUNO ARAUJO FURTADO DE MENDONCA - UFRRJExterna à Instituição - NATALIA DA SILVA GARCIA - UFRRJ