A soft computing approach to road classification

Current learning approaches to computer vision have mainly focussed on low- level image processing and object recognition, while tending to ignore high -level processing such as understanding. Here we propose an approach to obj ect recognition that facilitates the transition from recognition to underst anding. The proposed approach embraces the synergistic spirit of soft compu ting, exploiting the global search powers of genetic programming to determi ne fuzzy probabilistic models. It begins by segmenting the images into regi ons using standard image processing approaches, which are subsequently clas sified using a discovered fuzzy Cartesian granule feature classifier. Under standing is made possible through the transparent and succinct nature of th e discovered models. The recognition of roads in images is taken as an illu strative problem in the vision domain. The discovered fuzzy models while pr oviding high levels of accuracy (97%), also provide understanding of the pr oblem domain through the transparency of the learnt models. The learning st ep in the proposed approach is compared with other techniques such as decis ion trees, naive Bayes and neural networks using a variety of performance c riteria such as accuracy, understandability and efficiency.