This paper presents a smart algorithm that segments the different parts of a plant image. The processed image is actually produced of two bands, visible and fluorescence, generated by a plants health imaging system that we designed at the Canadian Space Agency (CSA). The main design criteria are precision and the speed of classification. The proposed algorithm is based on vector extraction - a feature vector is extracted for each class (leaf, stem, root and background) with the help of a pair of images containing the gray levels in regular image and fluorescent, which yields a two-dimensional vector. Using Pattern Recognition techniques, the algorithm scans all the image pixels and assigns them to their respective class. The obtained field results shows that with the characterization method developed and the use of clustering algorithms, the goal of segmentation of the plant was accomplished with as little as 8.75% error rate and a classification time of only 35 sec.