This work proposes a computational algorithm to differentiate five types of legumes and wheat using a support vector machine classifier (SVM). This tool is intended to provide a computational solution for automation applications that seek to differentiate or classify some types of legumes and cereal, in this case wheat. This automation is required for quality control, grain selection, sample statistics in laboratory analysis, or pedagogical use. This work’s main contribution is the selection of geometric and photometric descriptors that feed the SVM. The proposed algorithm first obtains an RGB, 24-bit color model image. Then, the image goes through a segmentation process consisting of thresholding, connectivity and labelling. Finally, the SVM classifies it according to descriptors that differentiate legumes and wheat. The validation process was carried out by visual inspection. An average result of 97% correctly identified legumes or wheat shows a very satisfactory performance of the proposed algorithm.