In experimental fields, scientists assess the resistance to orange and brown rust of sugarcane exclusively by identifying and grading infection by visual estimation on the leaves. This is time-consuming and may deliver subjective evaluations, limiting phenotyping experiments. Thus, to facilitate the leaf disease identification process, the goal of this study was to test an image analysis approach to differentiate the two types of rust on sugarcane leaves. Radial Support Vector Machine (SVM) models showed high accuracy (>0.88) in identifying the two types of rust, classifying segments of RGB images of infected leaves generated with object-based image analysis (OBIA) segmentation. This provides a basis for the development of applications that identify the two types of rust automatically through RGB images of sugarcane leaves.