Land-Use and Land-Cover Mapping Using a Gradable Classiﬁcation Method
Conventional spectral-based classiﬁcation methods have signiﬁcant limitations in the digital classiﬁcation of urban land-use and land-cover classes from high-resolution remotely sensed data because of the lack of consideration given to the spatial properties of images. To recognize the complex distribution of urban features in high-resolution image data, texture information consisting of a group of pixels should be considered. Lacunarity is an index used to characterize diﬀerent texture appearances. It is often reported that the land-use and land-cover in urban areas can be eﬀectively classiﬁed using the lacunarity index with high-resolution images. However, the applicability of the maximum-likelihood approach for hybrid analysis has not been reported. A more eﬀective approach that employs the original spectral data and lacunarity index can be expected to improve the accuracy of the classiﬁcation. A new classiﬁcation procedure referred to as “gradable classiﬁcation method” is proposed in this study. This method improves the classiﬁcation accuracy in incremental steps. The proposed classiﬁcation approach integrates several classiﬁcation maps created from original images and lacunarity maps, which consist of lacnarity values, to create a new classiﬁcation map. The results of this study conﬁrm the suitability of the gradable classiﬁcation approach, which produced a higher overall accuracy (68%) and kappa coeﬀicient (0.64) than those (65% and 0.60, respectively) obtained with the maximum-likelihood approach.
aerial photograph,cover classiﬁcation,gradable classiﬁcation,lacunarity,land,likelihood classiﬁcation,maximum,use and land