Model-based object recognition to measure crystal size and shape distributions from in situ video images

Measurement, modelling, and closed-loop control of crystal shape distribution: Literature review and future perspectives

“…However, for on-line monitoring, optimisation and control of shape and size distributions in a crystalliser, it is essential to accurately and efficiently obtain 3D crystal information of the whole crystal population and the associated properties. Of the many techniques [1,[41][42][43][44][45][46][47][48][49][50][51] for the imaging and reconstruction of 3D objects, optical imaging and image analysis techniques proved to be the most promising ones for 2D shape measurements [1,17,19,23,24,26,29,30,32,34,37,39,[52][53][54], and the techniques that are likely the most suitable ones for obtaining 3D crystal shape, hence growth rate of each crystal face, have made some advances in recent years [1, 45-50, 55, 56]. This is a very active research area with huge challenges and opportunities to innovatively develop process instrument and/or process probes and the associated image processing and shape reconstruction software for the real-time monitoring, optimisation and control of particulate processes.…”

“…This system can provide flexible arrangement among camera, light and particles in order to obtain high quality images, however, it may take a long time to tune the optic arrangement for the best configuration because the camera must focus on a small volume of slurry near the inner wall with light guide pointing at the same volume with proper illumination intensity. In contrast, imaging probe [29,30,39,52,64] can overcome these disadvantages but the feature of flexible arrangement will lose and currently the system cannot be applied to small size reactors (say 0.1litre) due to the smallest size of the probe available. The probe may only be used for short periods due to the high temperature of slurry in reactors or continuously cooling of the probe is required.…”

“…Patience and Rawlings [18] investigated the use of the PVM for measuring the shape and size information of sodium chlorate crystals with support from Pfizer. Larsen et al [29,39] developed new algorithms including model-based object recognition for image analysis, and also highlighted the necessity of advances in image analysis to improve image segmentation when particle overlap is significant. Louhi-Kultanen and co-workers [31,32] studied batch cooling crystallisation of sulfathiazole using an ATR-FTIR spectroscopy and image analysis simultaneously.…”

“…Larsen et al [29,39] developed an image segmentation method for crystals with high aspect ratios which was also able to satisfy the demand of real-time use. However, for all these methods, there is still a significant challenge for high solid concentrations when crystals in an image frame are overlapped [59].…”

Particle shape characterisation and classification using automated microscopy and shape descriptors in batch manufacture of particulate solids

“…One of the properties for particulate products is shape, not only because it is now recognized that size alone as often defined as volume equivalent diameter is over simplified and sometimes misleading, but also because shape can affect product performance properties such as bioavailability of drug particles. Optical microscopy has been proved to be one of the most effective techniques to determine particle shape and size off-line or on-line (Brittain, 2001; Calderon De Anda, Larsen, Rawlings, & Ferrier, 2006;Larsen, Rawlings, & Ferrier, 2007;Wang, Calderon De Anda, & Roberts, 2007;Wang, Roberts, & Ma, 2007). The images can be analyzed using image analysis algorithms (Zhang & Lu, 2004) to obtain particle shape information, which is often defined by some descriptors that have physical meanings such as aspect ratio.…”