The content of this article comes from the "Surveying and Mapping Bulletin" No. 3 of 2024, drawing review number: GS Jing (2024) No. 0499
Impervious surface extraction and expansion analysis based on SVM hybrid nucleus
Ji Jianren1, Wang Jingxue1,2, Wang Liqin3
1. College of Geomatics, Geomatics and Geosciences, Liaoning Technical University, Fuxin 123000, Liaoning, China;2. Collaborative Innovation Institute of Geospatial Information Service, Liaoning Technical University, Fuxin 123000, Liaoning, China;3. Liaoyang Institute of Land and Resources Exploration and Planning, Liaoyang 111000, China
Funds: National Natural Science Foundation of China (41871379); Liaoning Province Xingliao Talent Program (XLYC2007026); Liaoning Provincial Applied Basic Research Program (2022JH2/101300273)
Key words: impervious surface, texture features, support vector machine, hybrid kernel function, spatiotemporal analysis
Citation format: Ji Jianren, Wang Jingxue, Wang Liqin. Impervious Surface Extraction and Expansion Analysis Based on SVM Hybrid Kernel[J]. Bulletin of Surveying and Mapping, 2024(3): 43-48.doi: 10.13474/j.cnki.11-2246.2024.0308
Abstract:There are problems of high time complexity and low extraction accuracy when extracting impervious surfaces by using a single kernel function in support vector machines. In order to solve this problem, this paper introduces the polynomial kernel function on the basis of the radial basis kernel function, and proposes a method for extracting impervious surface with mixed kernel functions. Firstly, because the features with different attributes have similar spectral information, the combination of spectral information and image entropy texture information in the feature extraction process can more clearly distinguish the categories of various objects. Then, the polynomial kernel is introduced on the basis of the radial basis kernel function, which can obtain the feature information of the image from the local and global perspectives respectively, and improve the extraction accuracy of impervious surface. Finally, the spatiotemporal evolution analysis was carried out on the basis of the results of impervious surface extraction. In this paper, Landsat images from 2009 to 2021 in the main urban area of Fuxin City were used for experiments. The results show that the combination of spectral and entropy texture can improve the feature extraction effect and improve the accuracy of impervious surface extraction. Compared with the single kernel function extraction method, the accuracy of impervious surface extraction is improved by 2.5%, which proves the effectiveness of the proposed method.
About author:JI Jianren (1999—), female, master student, main research direction is remote sensing image processing. E-mail:[email protected] Corresponding author: Wang Jingxue, E-mail:[email protected]
First instance: Yang Ruifang review: Song Qifan
Final Judge: Jin Jun
information