Noninvasive Stereoscopic Backlight Imaging Design for In Situ Measurement of Particle Size Distribution During Continuous Crystallization via COBC

To realize in situ measurement of particle size distribution (PSD) during continuous crystallization via a continuous oscillatory baffled crystallizer (COBC), a novel noninvasive stereoscopic backlight imaging system is proposed in this article together with a deep-learning-based stereoscopic image analysis method. The stereoscopic imaging device consists of two telecentric lens-based cameras and backlight illuminators, while an optical platform is designed to acquire high-quality stereo images of flowing particles in the tubular crystallizer, which can be conveniently moved along the tubes of COBC for in situ measurement in different zones. Using a deep-learning-based image segmentation network for processing the snapshotted image pairs, a matching algorithm is established to sort out the interest-point correspondences regarding particle length and width from each particle mask pair. Based on the matched interest points, an optimal 3-D point reconstruction method is presented to measure each particle size with regard to its 3-D posture in COBC. The designed stereoscopic backlight imaging system and 3-D reconstruction method are validated by inserting a microscale checkerboard into the COBC for in situ calibration and measurement. Experimental results on measuring PSD during the continuous crystallization process of $\beta $ -type L-glutamic acid well demonstrate the effectiveness and merit of the proposed measurement method.