SMALL ITEM PHOTOGRAPHING TRIAGE ROBOT (SIPHTR)

Abstract

Small object sorting machines play a crucial role in various industries, from pharmaceuticals, where they sort pills, to agriculture, where they sort seeds. Our machine, the Small Item Photographing Triage Robot (SIPhTR), serves as a proof of concept for a small-scale sorting machine utilizing computer vision. SIPhTR is specifically designed to sort small objects like beads based on size, shape, color, and imprinted characters. It can handle up to 12 different variations (one at a time) with an accuracy of nearly 90% and operates at an average speed of 0.2 Hz (equivalent to sorting one bead every 5 seconds). SIPhTR achieves this through its three subsystems: the bead individualizer, computer vision, and bead sorter. The bead individualizer prepares the beads for imaging, allowing the computer vision subsystem to capture images of each bead individually, which are then processed using a machine learning algorithm for sorting. Finally, the bead sorter physically directs each bead into its designated bin. SIPhTR's high accuracy and speed demonstrate its potential as a scalable solution for small object sorting across industries such as pharmaceuticals, agriculture, and manufacturing.