Lithology classification in semi-arid area combining multi-source remote sensing images using support vector machine optimized by improved particle swarm algorithm

• DOI: 10.1016/j.jag.2023.103318
• Publication Date: 2023-04-28T17:41:40Z
• Authors: Jiaxin Lu; Ling Han; Lei Liu; Junfeng Wang; Zhaode Xia; Dingjian Jin; Xinlin Zha
• Keywords: Environmental sciences; Physical geography; Lithology classification; Particle swarm optimization; Multi-source remote sensing; Feature selection; Semi-arid area; GE1-350; 01 natural sciences; GB3-5030; 0105 earth and related environmental sciences

The development of multi-source remote sensing technologies is helpful for geologists to obtain more comprehensive and complete lithological maps. In recent years, establishing automatic classification models based on Machine Learning (ML) algorithms has become an important approach identify various lithologies supported by data. Aiming at the specific geological geographical conditions in a semi-arid area, Duolun County, Inner Mongolia Autonomous Region, China, this paper integrated GaoFen-2 (GF-2), Sentinel-2A, Advanced Spaceborne Thermal Emission Reflection Radiometer (ASTER) GaoFen-3 (GF-3) data, used Support Vector (SVM) classifier basis Particle Swarm Optimization (PSO) carry out lithology classification. Firstly, removing interference vegetation information from acquired 63-dimensional candidate feature sequence was constructed extracting spectral, backscattering, polarization texture features. Secondly, improved PSO algorithm with Inertia Factor changing S-curve Decreasing function (SDIF-PSO) proposed, basis, selection using SVM two-layer SDIF-PSO designed. Finally, iterative optimization process multiple model parameters accuracy before after were compared. experimental results showed that proposed had best capability, highest cross-validation 90.90%, which 3.85% than Grid-Search (GSO) algorithm, 0.15% Linear (LDIF-PSO) Concave (CDIF-PSO). dimension combination reduced 35 through selection, convergence reaches 92.14%, 1.24% all features same algorithm.