CD Skripsi
Klasifikasi Kematangan Tbs Kelapa Sawit Menggunakan Metode Pencitraan Multispektral
Sorting and grading FFB (fresh fruit brunch) are important tasks in order to obtain
high quality CPO (crude palm oil). Multispectral imaging is an alternative to
hyperspectral imaging to be implemented in a high speed sorting machine due to
less bandwidths used. Hence, less processing time is needed. This study aims to
use a multispectral imaging system for oil palm FFB maturity classification.
Multispectral imaging system uses 8 bandpass filters mounted in a filterwheel
consist wavelengths of 520, 680, 710, 740, 770, 800, 830, and 880 nm
respectively. The image acquisition and data processing were controlled using
python based program. The results showed a significant difference on the relative
reflectance intensity of hyperspectral images on three maturity levels: unripe, ripe,
overripe. Reflectance peak of hyperspectral image is higher for ripe and overripe
FFB compared to unripe FFB. FFBs firmness were measured as destructive
method to confirm the ripeness level. The relation between the firmness and
ripeness level are inversely proportional. The Spectral response of the
multispectral images were obtained same as on the hyperspectral images. Relative
reflectance intensity of multispectral images has a correlation with firmness at the
infrared spectrum. Using PCA (principal component analysis), all multispectral
data were processed for data visualization. Multispectral data which initially
consisted 8 wavelength variables were transformed into 2 variables (PC1 and
PC2) which can explain 90.41% of the overall data. The score plot of these two
components showed that the data has been grouped based on ripeness level.
Keyword: Hyperspectral imaging, multispectral imaging, filterwheel, band pass
filter, sorting, python, oil palm FFB, reflectance intensity, firmness, PCA
Tidak tersedia versi lain