Banca de QUALIFICAÇÃO: SIMONE SILVEIRA RODRIGUES

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : SIMONE SILVEIRA RODRIGUES
DATE: 04/03/2022
TIME: 08:00
LOCAL: Remoto
TITLE:

Detection of diseases and pests in a sorghum crop using digital images

KEY WORDS:

 Precision agriculture, remote sensing, object-oriented classification, UAV

PAGES: 35
BIG AREA: Ciências Agrárias
AREA: Engenharia Agrícola
SUBÁREA: Máquinas e Implementos Agrícolas
SUMMARY:

Precision Agriculture (PA) can be defined as the set of techniques and procedures of various technologies that enable the farmer to improve the performance of crops through the detailed study of the variability existing in cultures. In large areas of plantation, it is possible that there is a greater variability existing in the cultures in it, and, consequently, the control of the management of the crop becomes more difficult. Unmanned Aerial Vehicles (UAVs), together with remote sensing techniques, have been widely used in AP, as they are portable equipment and allow better monitoring of the crop during all stages of planting. This type of study allows the farmer to correct interferences so that they do not negatively affect the final productivity of the crop. In this way, the present work has as main objective the detection of diseases and pests in a sorghum crop through the analysis of RGB images, and as specific objectives to obtain the validation of the classifier used to classify the information obtained by the UAV, differentiate healthy plants and unhealthy using color descriptors, to compare the classification generated by the UAV images and by the color descriptors using radiometric information obtained by the spectroradiometer and evaluate the values of crop productivity. Classification allows extracting information from images, so that it is possible to recognize patterns and objects that correspond to the classes of interest. In the present research project we intend to work with RGB images obtained through UAV, which have high spatial resolution and low spectral response and, in this case, the most suitable method to classify the images is the object-oriented classification, because different pixel-to-pixel sorting, reduces the amount of noise and, consequently, also reduces the loss of information. In parallel with the images taken, it is also intended to distinguish healthy and unhealthy leaves by means of evaluation by color descriptors and the collection of spectral information of the crop using a spectrorrariometer to compare the results obtained. It is expected to obtain, throughout the season, the mosaic of images collected every 15 days from the first stage of planting the crop, and that from these, it will be possible to carry out the biweekly correction of the variations detected in the crop. At the end, with all the mosaics in hand, it is intended to generate a yield map that allows comparing the yield obtained and that expected for the harvest.

BANKING MEMBERS:
Presidente - 1050615 - MURILO MACHADO DE BARROS
Interno - 2161955 - ANDERSON GOMIDE COSTA
Interna - 2466219 - CAMILA FERREIRA DE PINHO