The classification of bean landraces based on their local variety and coloration is of particular interest because each bean landrace has a wide variety of nutritional components that benefit health. The laboratory procedures evaluate the diversity of local bean varieties based on color comparison as the primary element. In this paper, the authors propose a procedure to identify the coloration of the seeds according to the information provided for the CIE L*a*b* space of each bean landrace, a representation of the estimated color as a 3D histogram. This method uses this histogram to fit a statistical model known as the Gaussian Mixture Model with the GI algorithm. This process captures the representative information for each bean landrace as a point in the CIE L*a*b* space that represents the corresponding landrace color and can be used to identify the landraces using the K-nn method obtaining good results in homogeneous landraces.