Banca de DEFESA: HEVERTON BARBOSA DO NASCIMENTO

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : HEVERTON BARBOSA DO NASCIMENTO
DATE: 29/05/2023
TIME: 13:30
LOCAL: Sala virtual do google meet
TITLE:

Prediction of flooded area of water bodies via series models time frames generated by images.

KEY WORDS:

digital image processing, otsu method, time series, area

PAGES: 53
BIG AREA: Outra
AREA: Ciências Ambientais
SUMMARY:

Digital image processing methodologies can be applied to a wide range of
class of problems in computer vision. Among the applications, we can mention the
evaluation of the dynamics of water bodies from images. The monitoring and
prediction of volumes and flooded areas of water bodies are essential to issue
alerts from floods to water shortages. To predict the dynamics of the flooded area
of water bodies, this article proposes the use of time series models generated from
from multitemporal images. The water body investigated was Lake Kirwin (USA)
in the period from 2013 to 2022. The images of the lake were acquired in each period and
their flooded areas were estimated using the Otsu method. To describe
the temporal behavior of the Kirwin lake wetland, were investigated
autoregressive models integrated moving averages (ARIMA), autoregressive
integrated moving averages with seasonality (SARIMA), exponential smoothing
(ETS) and Holt-Winters. Among the models investigated to predict the dynamics of
flooded area of the water body, the Holt-Winters was the one that obtained the best
performance measures, with a forecast error of less than 6% in relation to the series
predicted. Its forecast suggests a scenario that for the 12 months following the series
under study there is a low possibility of an increase in the flooded area in the region
of Lake Kirwin.

BANKING MEMBERS:
Interno - CARLOS WAGNER OLIVEIRA
Presidente - FRANCISCO ALIXANDRE AVILA RODRIGUES
Interno - PAULO RENATO ALVES FIRMINO
Externa ao Programa - ROSILDA BENICIO DE SOUZA
Externo à Instituição - Ricardo Holanda Nobre