BRITO, K. B.; http://lattes.cnpq.br/6209621390993225; BRITO, Kal-El Basílio.
Résumé:
On inspection with ultraviolet sensitive camera, it is only supplied information on where
the corona discharges are concentrated and an estimative of intensity of these discharges
in high voltage equipment. For this purpose, it was developed a methodology of faculae
attribute extraction of ultraviolet images from 230 kV polymeric insulators. In order to
do this, videos captured by the ultraviolet detection camera were used as work material.
The methodology of the work consists of the development of several digital image
processing routines, initially, to adapt images and, later, to quantify the attributes of
interest extracted. The image adaptation routines seek to segment the shape of the
insulator from the background, to segment in an image the faculae that represent
discharges caused by corona effect, to integrate faculae from multiple frames in only one
image, to eliminate what is to be considered noise and to segment the shape of the
insulator into parts of interest. The attribute quantification routines are destined for
faculae area, perimeter, form factor and persistence, in addition to their distances of the
230 kV polymeric insulator parts of interest. As a result of the development of these
routines an algorithm for attribute extraction is presented. Finally, a sensitivity analysis
of the parameters used in the algorithm is made, with the intention to delimit its use
conditions. Two case studies of the application of the algorithm were performed, for each
of them 64 combinations of number of frames and noise limit were processed. Presenting
the surfaces of area, perimeter, shape factor and persistence of the faculae attributes in
function of the two cited parameters. In order to determine which number of frames and
noise limit would be suitable, it was adopted as metric to look for the lowest gradient
value of the attributes surfaces. In conclusion the methodology for extracting attributes
from the UV image faculae of insulators to improve the classification of insulators was
successfully developed and the contribution of the work is presented in extracting
attributes and determining suitable number of frames and noise limits for UV videos.
