Video-Monitoring grapevine pathology with hyperspectral imaging

2018 ASVO Seminar; Frontline pest and disease management for healthy vineyards
Monitoring grapevine pathology with hyperspectral imaging

Dr Bob Dambergs, The Australian Wine Research Institute (AWRI)

We have developed methods to detect Botrytis in grapes using hyperspectral imaging (HSI) with a camera that scans over the 400-1000 nm wavelength range (Vis-NIR). A normal photographic image is an RGB image, only containing information from red, green and blue wavelengths (650, 532 and 473 nm). A hyperspectral camera collects information from a large number of wavelengths and this information is stored as a spectrum in each pixel of a digital image. Hyperspectral cameras may collect visible light information (400-700 nm) but usually include some NIR wavelengths (700-2500 nm). One of the most commonly used wavelength ranges is 400-1000 nm. Pigments and organic molecules have distinct fingerprints in this wavelength range and this information can be used to prepare calibrations to identify material in regions within the spatial image. The calibrations are very robust and operate independent of grape variety, growing region and Botrytis strain. They can also discriminate Botrytis from Sour Rot. HSI works well on red and white grapes and was compared with the RotBot App. Although Rotbot could detect non-sporulating Botrytis in white grapes it could not detect sporulating Botrytis and could not detect Botrytis in red grapes. Other pathology such as sunburn, shrivel and invertebrate pests could also be detected with HSI, suggesting that this technology has potential for objective measurement of many grape quality parameters. Initial work was performed with large hyperspectral files but that are difficult to collect and process rapidly in a commercial situation, the next stage of this work will be to identify a small number of key wavelengths to be used in a multispectral camera.