摘要：Particle agglomeration commonly associated with crystallization processes has been increasingly studied for assessing the crystallization quality. An effective image analysis method is proposed for the agglomeration detection of needle-like particles during the crystallization process. The proposed method mainly consists of image pre-processing, primary sieving, re-segmentation, and particle classification. Firstly, the online captured images are efficiently pre-processed to reduce the influence from uneven illumination. Secondly, a primary sieving algorithm is established for discriminating candidate agglomerates, based on a shape feature. Thirdly, a re-segmentation algorithm is developed to extract the potential primary particles from the candidate agglomerates. In the end, a shape identification algorithm is given to recognize pseudo agglomerates from the candidate agglomerates based on two texture features. Thus, the agglomeration degree could be effectively assessed during the crystallization process. Experimental results on monitoring the cooling crystallization of β-form L-glutamic acid well demonstrate the effectiveness of the proposed image analysis method.
