Semi supervised machine learning techniques applied to Hyperspectral Imaging

My research is within the field of image analysis and machine learning classification. The PhD focuses on machine learning techniques applied to hyperspectral images, images containing several hundred wavelengths, with a practical focus on vegetation assessment. We will focus on semi-supervised machine learning techniques which use limited labelled data to produce accurate classifications. Semi supervised techniques are very desirable due to the time and monetary cost of collecting labelled data.

The main objectives of the research are:
- How to utilise semi-supervised learning in hyperspectral image classification to overcome the lack of labelled data?
- How to utilise modern computer vision techniques to improve image classification techniques?

In this PhD we will be creating new novel semi-supervised classification frameworks for image classification. The PhD will explore methods such as generative methods and very recent work on graph-based learning. Time will be spent exploring and expanding the theoretical theory of graph-based learning.

Considerable time will be spent coding and implementing these frameworks as well as processing hyperspectral data. This will involve using C++/Python to create fast efficient code which could be used in the private sector as well as implementing modern machine learning packages such as TensorFlow etc.