Pill Segmentation for Quality Control of Automated Pouch Packaging

Riesebos, Robert (2019) Pill Segmentation for Quality Control of Automated Pouch Packaging. Bachelor's Thesis, Computing Science.

Preview	Text bCS_2019_RiesebosR.pdf Download (30MB) | Preview
	Text Toestemming.pdf Restricted to Registered users only Download (118kB)

Abstract

Automating the process of packing pills into pouches has become an increasingly growing industry. An important step in this process is quality control. The quality is controlled using machines that take and analyse images of each pill pouch. At the basis of this analysis is pill segmentation, the goal of which is to accurately segment pills from the background. Difficulties in segmenting the pills arise from transparent pills, fragmented pills and irregular, non-uniform backgrounds. This thesis proposes two methods that deal with these difficulties using a ‘separation of concerns’, meaning that method consists of branches that focus on specific properties of pills. After evaluating both methods on a data set containing 554 pouches, both methods achieved a Jaccard index of above 0.96 and a Sørensen-Dice coefficient of about 0.98. Furthermore both methods are sufficiently fast to be deployed on state of the art machines.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Sobiecki, A.
Degree programme:	Computing Science
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	09 Jul 2019
Last Modified:	11 Jul 2019 06:59
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/20038

Actions (login required)

View Item