Every year, over $1.5 billion of wine is exported from New Zealand; nearly one third of all horticulture exports in 2016 (http://www.freshfacts.co.nz/). The potential for catastrophic crop loss due to plant diseases is of significant concern to the New Zealand wine industry. Leafroll virus (causal agent for the grapevine leafroll disease) is a significant problem for wine grape growers; it is incurable and in some cases presents no symptoms at all. In cases where symptoms are present, early detection is key to remove the vine as soon as possible to avoid the spread of the virus.
A promising method involves the use of multispectral cameras and machine learning/computer vision in the early detection of certain plant diseases. Multispectral images are analysed to find patterns associated with certain diseases. A major advantage of this method is that human error is eliminated, giving objective results.
Our goal is to aid the early detection of leafroll virus on grapevines. The nature of the camera used introduces parallax error, hence pre-processing of the images is required.
This project is sponsored by Plant & Food Research, a Crown Research Institute of New Zealand.
Undergraduate
• Implementing a solution to eliminate parallax error in close range multispectral images.
• Implementing image segmentation on parallax-corrected images to distinguish between “bunch”, “leaf”, or “other”.
• Detection of leafroll virus in the leaf section of pre-processed images (if time allows).
Knowledge in signal processing and Image processing
Lab allocations have not been finalised