This work proposes a Hidden Markov Model (HMM)-based methodology to classify agricultural crops, exploring information of temporal image sequences from TM and ETM+/Landsat sensors. HMMs are used to relate the varying spectral response along the crop cycle with plant phenology for different crop classes. The method recognizes different agricultural crops by analyzing their spectral profiles over a temporal sequence of medium resolution satellite images ( approximation 30m). In our approach the temporal behaviour of each crop class is modelled by a specific HMM. A segment- based classification is performed using the average spectral values of the pixels in each image segment across an image sequence, which is subsequently submitted to the HMMs of each crop class. The image segment is assigned to the crop class, whose corresponding HMM delivers the highest probability of emitting the observed sequence of spectral values. Experiments were conducted upon a set of 12 co-registered and radiometrically corrected LANDSAT images. The images cover an area of the State of São Paulo, Brazil with about 124.100ha, between the years 2002 and 2004. The following classes were considered: sugarcane, soybean, corn, pasture and riparian forest. Performance assessment was carried out upon a data set classified visually by two analysts and validated by extensive field work. The performance of the proposed multitemporal classification method was compared to that of a monotemporal maximum likelihood classifier, and the results indicated a remarkable superiority of the HMM-based method, which achieved an average of no less than 91% accuracy in the identification of the correct crop, for sequences of data containing a single crop class.