Detection of explosives has always been a priority for homeland security. Jointly, terahertz spectroscopy and imaging are emerging and promising candidates as contactless and safe systems. In this work, we treated data resulting from hyperspectral imaging obtained by THz-time domain spectroscopy, with chemometric tools. We found efficient identification and sorting of targeted explosives in the case of pure and mixture samples. In this aim, we applied to images Principal Component Analysis (PCA) to discriminate between RDX, PETN and mixtures of the two materials, using the absorbance as the key-parameter. Then we applied Partial Least Squares-Discriminant Analysis (PLS-DA) to each pixel of the hyperspectral images to sort the explosives into different classes. The results clearly show successful identification and categorization of the explosives under study.