Remote Sensing and GIS Applications in Environmental Sciences
Comparison of the efficiency of pixel-based support vector machine kernel functions and object-based fuzzy operators in monitoring the changes in urban growth and expansion of Tabriz.
Document Type : Original Article
Authors
1 ph.D student of remote sensing and geographic information system of Tabriz University
2 Department of Mapping, Technical Faculty, Tabriz University
3 Ph.D Student. Acecr Researcher: Development and Planning Research Institute, GIS & RS Center, Tabriz
Abstract
Urban growth as a determining factor of social welfare and environmental sustainability is considered very vital. In recent years, remote sensing data has been used as a tool to determine the size of urban expansion and monitor urban growth. The purpose of the present research is to compare the efficiency of pixel-based vector machine kernel functions and object-oriented fuzzy operators in monitoring Tabriz urban growth changes using Sentinel 2 satellite. in order to classify each image based on pixel base support vector machine kernel functions, using support vector machine kernels, the image classification process was performed and the urban growth map of each function was produced for the years under study. ecognition software was used for object-oriented classification. At this stage, segmentation was done based on different scales, shape factor and compression ratio to reach the optimal segmentation mode. Then, teaching points were identified and classification was done using fuzzy operators. Based on the results of this research, the AND fuzzy operator with an overall accuracy of 96.49% and a kappa coefficient of 0.9688 for the image of 2014 and an overall accuracy of 97.31% and a kappa coefficient of 0.9725 for the image of 1403, the accuracy value provided more. Therefore, considering the greater accuracy of object-oriented fuzzy operators , it can be stated that object-oriented processing algorithms of satellite images in the classification of digital satellite images, compared to support vector machine algorithms, make it possible to achieve higher accuracy in extracting the urban area of Tabriz.
Keywords
Main Subjects
)