Improvement of pattern recognition algorithm for drop size measurement

Abstract

In the present work, the pattern recognition algorithm for drop size measurement has been improved by focusing on the processing of the partially detected or overlapped drop images and the oval-shaped drop images. The improved algorithm was assessed by using an artificially prepared image frame, where the overlapped and the oval-shaped particles are mined with the normal spherical ones (with their hue size distributions known apriori). The results show that both the number of particles recognized and the measurement accuracy are improved significantly.