Caractérisation de jets agricoles par traitement d'images

A digital image analysis algorithm is proposed as an alternative to established optical techniques such as Phase Doppler Particle Analyser (PDPA) or laser diffraction spectrometry (LDS) for the experimental determination of the size and velocity distributions of droplets produced by agricultural nozzles . The algorithm requires pairs of grayscale images acquired within a very short time for particle tracking in a backlighted arrangement. Spatial illumination heterogeneities are corrected by subtracting a mean background from each image. Droplets are then localized and sized. The algorithm accuracy is ensured by the rejection of out of focus particle using a focus parameter depending on gradient intensity at particle edges. Thresholds for in-focus particle selection were determined by studying the evolution of the focus parameter and the error on particle size measurement from images containing known size droplets at various distance of the object plane. Selected droplets are identified on both images of the pair for determining their velocities. Finally, drop size distribution is corrected to account for the uneven sampling probability caused by the volumetric method.