Method For Detecting The Sizes And Agglomeration Of Needle-like Crystals During Crystallization Based On In-situ Imaging

Crystallization engineering is widely used in pharmaceutical,chemical,food and other industries.In order to obtain the desired crystal shape and ensure the quality of crystal products,it is necessary to monitor the crystal growth process for real-time control and optimization.In the past decade,more and more applications of image monitoring technology have been used for crystallization processes.This paper firstly introduces the research status of image detection technology for industrial crystallization processes,and then analyzes the advantages and difficulties of monitoring crystallization processes by using real-time imaging systems.An image preprocessing method suitable for in-situ image detection is proposed based on the shape features of needle-like crystals.It can overcome the negative influence arising from uneven illumination,particle motion and solution stirring,eliminate the noise information in the image to improve the image quality,and enhance the crystal shape area in the image.The proposed image preprocessing method mainly includes image denoising based on median filtering,image enhancement algorithm based on homomorphic filtering,and image segmentation based on the full name(OTSU)method,therefore realizing the separation of crystal shapes and the background image.Based on the above image preprocessing,a method is proposed for feature extraction and classification of target crystals.According to the texture features and defined convexity features of crystals,the classifiers are designed based on the crystal texture features and convexity features.Besides,a shape fitting method for needle-like crystals is proposed based on the minimum external moment,which can accurately count the length and width of crystals in the captured images.By constructing the gray level co-occurrence matrix to extract the texture features,on-focus crystals and off-focus crystals in the captured images can be identified.Regular crystals and overlapping crystals can be distinguished by defining the convexity features.Based on the overlapping crystals extracted by the above shape classifiers,an image re-segmentation method is proposed based on the corner features,which can effectively identify pseudo agglomerates in the in-situ captured images.The corner detection is performed by using a contour-based CPDA corner detection algorithm.The corner points are classified in terms of the right hand criterion.Then,according to the pre-defined corner points,a matching algorithm is developed to segment the agglomerated crystals,so as to separate individual crystals from the overlapped crystals.Subsequently,the pseudo agglomerates are recognized based on the texture features.Experiments are made for monitoring the cooling crystallization of ? form L-glutamic acid.The results well demonstrate the effectiveness and reliability of the proposed in-situ image analysis method in comparison with the existing image detection algorithms.