Resumen
This paper presents a vision-based indoor scene recognition method from aerial time-series images obtained using a micro air vehicle (MAV). The proposed method comprises two procedures: a codebook feature description procedure, and a recognition procedure using category maps. For the former procedure, codebooks are created automatically as visual words using self-organizing maps (SOMs) after extracting part-based local features using a part-based descriptor from time-series scene images. For the latter procedure, category maps are created using counter propagation networks (CPNs) with the extraction of category boundaries using a unified distance matrix (U-Matrix). Using category maps, topologies of image features are mapped into a low-dimensional space based on competitive and neighborhood learning. We obtained aerial time-series image datasets of five sets for two flight routes: a round flight route and a zigzag flight route. The experimentally obtained results with leave-one-out cross-validation (LOOCV) revealed respective mean recognition accuracies for the round flight datasets (RFDs) and zigzag flight datasets (ZFDs) of 71.7% and 65.5% for 10 zones. The category maps addressed the complexity of scenes because of segmented categories. Although extraction results of category boundaries using U-Matrix were partially discontinuous, we obtained comprehensive category boundaries that segment scenes into several categories.