This project builds upon existing work by current and previous PhD students and will work closely with a current PhD student. We have developed a large database of pure honey which has been captured using the hyperspectral imaging system and developed some machine learning models around this to classify the type of honey.
This project requires students to capture samples of adulterated honey, which is honey that has had something else added to it (sugar in our case) and then applying machine learning techniques (mainly neural networks using python and TensorFlow) to the dataset to predict if it has been adulterated. This is important to develop because adulteration is a common type of honey fraud, as pure honey is very expensive to produce. Honey fraud is important to prevent as honey is very valuable to the New Zealand economy.
This is the first extensive work in the area of predicting honey adulteration, but there is a clear process to work through based on how we captured our existing dataset. Developing machine learning algorithms will require some research, as it will likely involve designing a new neural network architecture or technique.
The project will involve a mix of hands-on work in collecting data from honey samples, and advanced machine learning techniques to build the predictor.
Interested students must see the supervisor before bidding for the project.
Committed hard-working students. Machine learning experience and image processing background are advantegeous. Interested students must see the supervisor before bidding for the project.
Lab allocations have not been finalised