Assessment of supervised and unsupervised classification methods of High-Resolution images in dense urban environments for extracting land use (Case study of GeoEye-۱ images from Karaj region)

Today, spatial information is recognized a crucial element in information infrastructure and it has been utilized in various applications such as urban planning and sustainable development. Remote sensing images and data are the main sources for spatial information extraction. The rapid advancements in satellite and sensor technologies have facilitated the availability of high-resolution remote sensing images.These images have the capacity to enhance the extraction of spatial information in comparison to images with lower resolutions. Image classification techniques enable the extraction of spatial information from these images. Recent developments in machine learning methods have been successfully employed to classify high-resolution images, demonstrating promising outcomes in terms of accurate classification. These methods can be classified into two general categories: supervised and unsupervised. In this study, classification was performed using both supervised and unsupervised techniques. Classification was carried out using a high-resolution image from GeoEye-۱, which was acquired in the Karaj region. Unsupervised learning techniques like K-Means and IsoData, as well as supervised methods such as Minimum Distance, SVM, Mahalanobis Distance, Maximum Likelihood, Spectral Angle Mapper, and Spectral Information Divergence, were employed in this study. To assess the accuracy of the classification outcomes, the confusion matrix was used to calculate the overall accuracy and kappa coefficient. The results indicate that the SVM method outperformed other approaches, achieving an overall accuracy of ۹۷.۶۵ and a kappa coefficient of ۰.۹۶, making it the superior method for image classification.

کلیدواژه ها:

Remote sensing ، Resolution ، Classification ، Machine learning ، Land cover and land use