The image processing technique for particle sizing is simple and, in principle can handle particles with various shapes since it is based on direct visualization. There are two major research subjects concerned with this technique: identification of particles (i.e., boundary detection and pattern recognition) and determination of in-focus criteria. Among them, the present study focused on the pattern recognition algorithm to process particles with various shapes (circular or elliptic shapes, heavily overlapped particles). For this, the Hough transform algorithm and the boundary curvature detection algorithm were proposed.
Hough transform is an algorithm to detect parametric curves such as straight lines or circles which can be represented by several parameters. By using this algorithm along with the false circle elimination process, true particles were identified from the parameter space.
Conceptually, the boundary curvature detection algorithm has an advantage over the others because it can identify the particle size and shape simultaneously, and can separate the overlapped particles more effectively. The boundary curvature was estimated from the change of the slopes of two neighboring segments at the corresponding location. Average curvatures were used in sizing circular particles, and the elliptic shapes were identified through the Fourier transform of curvature.
The developed algorithms were assessed by using artificially prepared images of particles and compared with the algorithm of the convex-hull method. The result showed that the Hough transform algorithm is useful for counting and sizing the heavily overlapped spherical particles; however this algorithm needs much longer processing time and is unable to recognize the elliptic objects. On the other hand, the boundary curvature detection algorithm can handle the particle images with various shapes (i.e., circular, elliptic and overlapped shapes) effectively. In overall, the boundary curvature detection algorithm turned out to be the best among the algorithms tested in the present study in terms of the recognition efficiency and the measurement accuracy.
