On the manufacturing condition estimation of spray nozzles via image processing and epicycle representation

Resumo

Defects on irrigation spray nozzles, occurred during the manufacturing process, can affect such distribu-
tion pattern leading to an uneven product application and increasing wastage. With a specific test bench, side-view

images of hollow-cone nozzle sprays were collected as an attempt to assess the manufacturing condition of very

brand-new nozzles. Classical image processing tools was applied on this side-view images, using a filter to re-
duce any spurious noise and the combined use of power-law transformation together with the contrast stretching

technique to enhance image contrast, highlighting the presence of grooves in sprays, which is an indication of
uneven distribution. An algorithm was develop to represent the spray as a contour with a few representative points,
reduced from the large number of pixels and so compressing heavily the image size. These points were then used
as input to estimate the spray format via epicycle representation carried out using a computation algorithm based
on the discrete Fourier transform (DFT) leading to coefficients that are sensitive to the spray jet format. Samples
of nozzles presenting no manufacturing defects was used to define threshold values for these coefficients, so that,
any change in this regard can be an indication of manufacturing problems.