Yield prediction from digital image analysis: A technique with potential for vineyard assessments prior to harvest

• Autores: Gregory M. Dunn, Stephen R. Martin
• Localización: Australian journal of grape and wine research, ISSN 1322-7130, Vol. 10, Nº 3, 2004, págs. 196-198
• Idioma: inglés
• Texto completo no disponible (Saber más ...)
• Resumen

  • Digital photographs were taken of four 1 m X 1 m portions of canopy of Cabernet Sauvignon grapevines, as they were being progressively de-fruited close to harvest. The program EasyAccess version 6.3 was used to select 'fruit' pixels by visually setting red, green and blue threshold values and tolerances for the first image and applying these to all other images. The program was then used to automatically count 'fruit' pixels and the total number of pixels for each image. Even though two hours separated the first and last photographs, the ratio of 'fruit' pixels to total image pixels explained 85% of the variation in yield (kg per linear m of fruiting wire) for all 16 vine X de-fruiting combinations. This improved to between 94 and 99% for individual portions of canopy. Implications from our present digital image analysis for future development of both automated and spatially aware methods to predict vineyard yield are discussed.