In-Situ Agglomeration Measurement during Kdp Crystallization Based on Double-View Image Analysis

A double-view image analysis method is proposed to detect particle agglomeration during a crystallization process, based on using a non-invasive imaging system composed of two cameras installed at different angles outside the crystallizer. A fast image segmentation approach is firstly adopted for image preprocessing in order to reduce the influence from particle motion, uneven illumination background, and solution turbulence. Then a preliminary sieving algorithm is given to screen out candidate agglomerates from preprocessed images. By introducing new texture descriptors for pattern recognition, a fast feature matching algorithm is developed to recognize pseudo agglomerates. Subsequently, an efficient particle number counting algorithm is established to assess the agglomeration degree, by which both the unagglomerated particles and the primary particles involved in agglomerates can be accurately counted, respectively. Experimental tests on monitoring the potassium dihydrogen phosphate (KDP) crystallization process are shown to demonstrate the effectiveness and advantage of the proposed method.