Lectures > Lecture 2

Lecture 2 by Anton Du Plessis, Stellenbosch University, South Africa / Dragonfly EMEAI, Comet Technologies, Canada

Title: Image segmentation: from thresholding to deep learning.

In this course we will discuss various image segmentation approaches. Segmentation is a necessary step prior to any 2D or 3D image analysis of CT data and the accuracy of any analysis depends on the segmentation performed. Various tools exist for this task and are available in popular commercial and open source image analysis programs. A wide variety of these will be discussed and demonstrated, explaining the working principles of each and the best ways to use them. Traditional segmentation tools combined with morphological operations and image filtering approaches will be explained and examples will be shown where combinations of these tools can be used to segment challenging features such as open vs closed pores, cracks, inclusions and multiple phases in a rock. Moving forward from this, AI tools will be discussed and demonstrated for machine learning and deep learning based segmentation.

Biography: Prof. Anton Du Plessis holds a PhD in Physics (2005) from Stellenbosch University. After working as researcher and research group leader at the Council for Scientific and Industrial Research (CSIR) for 6 years, he joined Stellenbosch University as head of the newly established CT facility, a core facility offering CT imaging services. In this role he developed CT imaging methods and applications for different application fields and pursued his research interests in X-ray tomography applications, additive manufacturing and engineering materials characterization. He has published over 150 journal papers and is on the editorial board of a number of journals. He is the editor in chief role at the journal Tomography of Materials and Structures and deputy editor of Additive Manufacturing Letters. He is currently Sales Director Dragonfly EMEAI at Comet Technologies Canada Inc (remotely) and affiliated as Associate Professor at Stellenbosch University (South Africa). He enjoys research collaboration and taking scientific imaging to the next level.