< Terug naar vorige pagina

Publicatie

Hyperspectral waveband selection for detecting floral pear buds

Boekbijdrage - Boekhoofdstuk Conferentiebijdrage

Whereas fruit thinning is traditionally mostly done by hand, recently several prototype thinning machines have been developed. A disadvantage these machines share is that they are not selective. Real-time knowledge of the distribution of floral buds can greatly improve efficiency. For this reason, a multi-spectral vision sensor is being developed to detect the floral buds of the pear cultivar, Conference. This paper describes the selection of a combination of wavebands with high discriminating power between buds and branches. Hyper-spectral scans in the range 400-1000 nm were taken from tree samples. The spectra obtained were divided into bands to build a variable pool from which a stepwise algorithm selected a combination of six wavebands. Using the selected bands, a misclassification rate of just 4.55 % of the (pixel) observations was achieved.
Boek: Precision Agriculture 2011 - Papers Presented at the 8th European Conference on Precision Agriculture 2011, ECPA 2011
Pagina's: 90 - 98
ISBN:9788090483057
Jaar van publicatie:2011