APPLICATION OF NEAR INFRARED HYPERSPECTRAL IMAGING SPECTROSCOPY FOR THE ANALYSIS OF COCOA BEANS

INTRODUCTION
Cocoa (Theobroma cacao) is a valuable crop that played an important role in many old South American cultures. Native of the Central and South American rainforests, cocoa trees are now cultivated in many tropical areas around the world. Cocoa beans are the fermented seeds of the fruits and are the most important ingredient of all chocolate products. They are about the size of an almond and, after roasting, the taste is rich, astringent, creamy and full of cacao's complex flavor notes that range from berry to coffee. There are three varieties of cocoa – Forastero, Criollo, and Trinitario —, the last one being a hybrid of the others. The flavor of the beans depends on the variety, cultivation methods, soil and climate. These parameters could be largely summarized through the geographical origin of bean production. Furthermore, the post-harvest treatments, performed by the cocoa farmers on the seeds are determinant to achieve the expression of all its aromatic potential. These treatments mainly include fermentation, drying, winnowing,. Different studies using vibrational spectroscopy have proved the performance of these techniques for the discrimination of cocoa beans according to fermentation level or variety among other.
The aim of this work is to prove that Near Infrared Hyperspectral Imaging can be used as fast and reliable technique for the analysis of whole cacao beans from the Amazon basin of Brazil in order to study the possibility of discriminating according to their geographical origin and their fermentation time. Other parameters directly linked to the fermentation stage, as the year of harvest, the layer thickness during fermentation and the drying process have been taken into account.
EXPERIMENTAL
This preliminary work has been carried out on more than 2000 cocoa beans sorted from 147 samples collected by the Federal University of Para (UFPA), Brazil. These cocoa beans consisted on samples with different origins and producers, two harvest years and different thickness of the layer during fermentation as well as different fermentation time and drying processes.
Hyperspectral images were collected at CRA-W using a NIR hyperspectral line scan or push-broom imaging system, combined with a conveyor belt. Each image consists of 320-pixel lines acquired at the 1,100-2,400 nm wavelength range.
RESULTS AND DISCUSSION
A complete experimental design has been developed including the large variability of the samples. After optimization of the acquisition parameters using the Hyperspectral Imaging, all the cacao beans have been analyzed in duplicate. Different PCA analyses have been performed. The beans were differentiated in a first step, according to their geographical origin. This is probably related to the specific post-harvest handling practices in a certain locality, climate or varietal difference. Then the beans were grouped according to the different post-harvest treatment suffered, i.e. fermentation, fermentation and/or drying in the shadow or in the sun and the level of fermentation. However, based on the results of the PCA, no significant difference was found between the samples of the two year available, probably due to the limited number of samples available.