A geometric-based method is proposed to recognize the overlapping particles of different polygonal shapes such as rectangular, regular and/or irregular prismatic particles in a gray tone image. The first step consists in extracting the salient corners, identified by their locations and orientations, of the overlapping particles. Although there are certain difficulties like the perspective geometric projection, out of focus, transparency and superposition of the studied particles. Then, a new clustering technique is applied to detect the shape by grouping its correspondent salient corners according to the geometric properties of each shape. A simulation process is carried out for evaluating the performance of the proposed method. Then, it is particularly applied on a real application of batch cooling crystallization of the ammonium oxalate in pure water. The experimental results show that the method is efficient to recognize the overlapping particles of different shapes and sizes.