This paper introduces a novel, ultra-compact particle size analyzer using a CMOS image sensor and machine learning. The device uses a small form factor angular spatial filter to collect light scattered by particles at discrete angles. A light-emitting diode (LED) and CMOS image sensor array acquire angularly resolved scattering images, which are then used by a machine learning model to predict the volume median diameter of the particles. Testing with glass beads (13-125 µm) showed mean absolute percentage errors of 5.09% (without concentration as input) and 2.5% (with concentration as input). The device's compact size and low cost make it suitable for applications outside of standard laboratories, such as online industrial process monitoring.

Rubaiya Hussain, Mehmet Alican Noyan, Getinet Woyessa, Rodrigo R. Retamal Marín, Pedro Antonio Martinez, Faiz M. Mahdi, Vittoria Finazzi, Thomas A. Hazlehurst, Timothy N. Hunter, Tomeu Coll, Michael Stintz, Frans Muller, Georgios Chalkias, Valerio Pruneri