Combining Spectral and Band Angle Mapper into a One-Class Support Vector Machine for Extracting Photovoltaic Power Plants using Landsat 8 OLI Imagery

The automatic extraction of man-made objects from remote sensing imagery is a significant research area in pattern recognition. However, the spectral uncertainty of terrestrial objects, caused by environmental complexity and remote sensing image processing, poses challenges for traditional object extraction algorithms that rely on the spectral statistical properties of training samples. This uncertainty also affects the performance of the Spectral Angle Mapper (SAM) algorithm in object recognition. To address the limitations of the traditional SAM algorithm in capturing subtle spectral variability, this paper proposes a method based on the One-Class Support Vector Machine (OCSVM), which integrates SAM and Band Angle Mapper (BAM) as joint features to extract photovoltaic (PV) power plants using Landsat 8 OLI imagery. The Jiangsu coastal zone was selected as the study area due to its dynamic environment, which is more likely to induce spectral uncertainty. Accuracy assessment was conducted by comparing results with visual interpretation. Compared to traditional methods, including OCSVM-PCA, OCSVM-SMN, and OCSVM-SAM, the proposed method achieved higher overall accuracies of 85.57%, 91.87%, and 91.27% for Sheyang, Dongtai, and Haian, respectively. Additionally, the number of misclassified background pixels was significantly reduced, and the generalization capability was notably improved, confirming the validity and superiority of the proposed method.