Developing an Interpretation System for High-Resolution Remotely Sensed Images Based on Hybrid Decision-Making Process in a Multi-scale Manner

Abstract

High spatial resolution (HSR) images are valuable data sources for urban applications. HSR images by high intra-class and low inter-class variabilities lead to a reduction in the statistical separability of the different land-cover classes in the spectral domain; therefore, conventional classification methods using merely spectral and textural information have proven to be inadequate for the HSR data. In this regard, interpretation systems by employing spatial and conceptual compatibility can be useful to closer a machining process to human analysis. In this research, a new framework has been introduced to classify remotely sensed imagery using a combination of ontological rules and geographic object-based image analysis. This paper to some degree attempts to untangle a few of the gaps in this field, especially by incorporating multi-scale analysis into the proposed framework, which is very advantageous in such an application. The present study used interactive segmentation and interpretation segmentation process with respect to the geometry of the image classes. In the proposed method, the segmentation process has been performed to avoid under-segmentation problems at several levels of the scale. The levels of scale are entered in the process of scoring and interpretation of the decision (not just applied at the level of results). This brings the process of labeling and interpretation closer to structural and natural reality. Additionally, a hybrid decision-making process (knowledge based and support vector machine) has been considered to achieve better results. The knowledge-based method is implemented to model the ontological relationships with the aim of labeling and controlling the decision-making process. To evaluate the efficiency of the method, the results of this research were assessed and compared with those of other methods in an urban area. The results showed that the proposed technique improved the overall accuracy and kappa coefficient by 9% and 11.5%, respectively.