Application Technique of Hyperspectral Imaging for Estimation of Grass Chemical Compositions

Suzuki Yumiko, Tanaka Katsuyuki, Kato Wataru, Okamoto Hiroshi, Kataoka Takashi, Shimada Hiroshi, Sugiura Toshihiro, Shima Eikichi

doi:10.14941/grass.54.328

The chemical compositions (CP, ADF and NDF) of mowed grass were estimated by hyperspectral imaging. ImSpector V10 (Spectral Imaging Ltd.) was used as a hyperspectral imaging sensor, which is an integrated combination of an imaging spectrograph and a matrix camera. The spectral range of this sensor is the wavelengths between 360 and 1010nm, and the spectral resolution is 10nm. The sensor was mounted on the roof of a vehicle (AEBI, TT33), which has a drum mower attached in the front. The hyperspectral images of a meadow field were acquired by moving the vehicle. The estimation models of grass chemical compositions were obtained by spectral data sampled from hyperspectral images and by actual grass chemical compositions. The multiple regression analysis with stepwise selection of variables was used for obtaining estimation models. The cross validation (leave-one-out method) was applied to the validation of the estimation models, and the EI (evaluation index) method was applied to the confirmation of the practical accuracy of the estimation models. In conclusion, the cross validation and the EI method showed efficacy of all estimation models.

Application Technique of Hyperspectral Imaging for Estimation of Grass Chemical Compositions

Bibliographic Information

Search this article

Abstract

Journal

Citations (1)*help

References(25)*help

Related Projects

Keywords

Details 詳細情報について

Export

Report a problem