IEEE Project Abstract

In this paper, we presented a novel building recognition method based on sparse representation of spatial texture and color features. At present, most popular methods are based on gist featureswhich can only roughly reflect the spatial information of building images. The proposed method, incontrast, uses multi-scale neighborhood sensitive histograms of oriented gradient (MNSHOG) and colorauto-correlogram (CA) to extract texture and color features of building images. Both MNSHOG and CA takespatial information of building images into account while calculating texture and color features. Then, colorand texture features are combined to form joint features whose sparse representation can be dimensionallyreduced by an autoencoder. Finally, extreme learning machine is used to classify the combined featuresafter dimensionality reduction into different classes. Several experiments were conducted on the benchmarkSheffield building dataset. The mean recognition rate of our method is much higher than that of the existingmethods, which shows the effectiveness of our method.