This study aims to identify and separate solar panels in South Asia using semantic segmentation and scale optimization with the MRS algorithm. The main goal is to process satellite images with a mosaic approach to detect photovoltaic panels. The study focuses on selecting the best scales in the MRS algorithm to improve separation accuracy. Scales of 2, 4, 6, 8, 10, 12, and 14 meters are tested. Key metrics such as Precision Segmentation Error (PSE), Noise-to-Signal Ratio (NSR), and Edge Discrepancy (ED) are used for evaluation. Results show that medium scales, like 6 meters, perform best for identifying individual panels, with a minimum ED of about 0.3. Larger scales, such as 200 meters or more, are better for analyzing groups of panels and estimating energy production. Variations in metrics across scales highlight the method’s sensitivity to scale selection. A weight of 0.7 is prioritized for spectral data and 0.3 for panel shape. Given the rapid growth of solar energy in South Asia, this approach enables fast mapping, energy production estimation, and efficient resource management. The findings emphasize that scale choice depends on the analysis goal (individual or group panels) and regional features. This method can support sustainable energy policies.

Solar energy, photovoltaic panels, semantic segmentation, MRS algorithm, PSE, NSR, ED.