Multitemporal fuzzy classification model based on class transition possibilities

This paper proposes a new method to model temporal knowledge and to combine it with spectral and spatial knowledge within an integrated fuzzy automatic image classification framework for land-use land-cover map update applications. The classification model explores not only the object features, but also information about its class at a previous date. The method expresses temporal class dependencies by means of a transition diagram, assigning a possibility value to each class transition. A Genetic Algorithm (GA) carries out the class transition possibilities estimation. Temporal and spectral/spatial classification results are combined by means of fuzzy aggregation. The improvement achieved by the use of multitemporal knowledge rather than a pure monotemporal approach was assessed in a real application using LANDSAT images from Midwest Brazil. The experiments showed that the use of temporal knowledge markedly improved the classification performance, in comparison to a conventional single-time classification. A further observation was that multitemporal knowledge may subsume the knowledge related to steady spatial attributes whose values do not significantly change over time.