Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
A simplified model for cloud height determination with fisheye cameras stereo computer vision
Submission Author:
Darío Gerardo Fantini , DF
Co-Authors:
Darío Gerardo Fantini, Matheus Carvalho, Mario Benjamim Baptista de Siqueira
Presenter: Mario Benjamim Baptista de Siqueira
doi://10.26678/ABCM.COBEM2023.COB2023-1894
Abstract
Short-term solar forecasting, among others, is necessary so that different solar-based power generation systems can have a greater share in the energy matrix without causing high impacts due to the variability of the solar resource. Forecasting can be obtained through sky images by detecting and analysing clouds mainly responsible for the high variability of solar irradiation in the short term. The accuracy of image-based solar irradiance forecasting systems depends on several processing steps, two of which are vital: height detection and cloud tracking. These tasks are complex because cloud conditions are often distributed in several layers that move simultaneously in different directions. Therefore, ceilometers are commonly used to perform this task but are expensive to acquire. Thus, in this work we propose a model that uses image triangulation through stereo vision with two fisheye cameras to measure the height of the cloud base, providing a lower cost and easy-to-implement system for applications that do not require or not justify the use of a ceilometer. This system is based on two Vivotek FE9381-EHV cameras and a wireless router for IP CCTV from Intelbras, currently installed in the Energy and Environment Laboratory of the University of Brasilia. The stereo vision model used to process the images and determine the heights does not require the image distortion removal process, typical of fisheye lenses, for pinhole images, nor the rectification of the original images, being these the major contributions compared to other existing models. For the validation of our model, we have used a drone model DJI Mavic Air 2 that provided the flight altitude. With it in stationary flight at various points and desired altitudes, we captured several images of the sky with our cameras. With this, we verified the heights obtained by our model and compared them with those reported by the drone. The first results showed an error of less than 10% (concerning the value reported by the drone) at heights up to 100 metres.
Keywords
cloud height, computer vision fisheye stereo, Short-term solar forecasting, stereo fisheye camera model
