Authors

J. S. S. Prasanna

PG Scholar, Department of Computer Science and Engineering, University College of Engineering Kakinada (A), JNTUK, India

S. Chandra Sekhar

Assistant Professor, Department of Computer Science and Engineering, University College of Engineering Kakinada (A), JNTUK, India

Abstract

The increasing installation of rooftop solar photovoltaic (PV) systems has created a need for automated techniques that can identify photovoltaic panels from aerial and satellite imagery and convert image-based detections into useful solar-energy information. Manual rooftop inspection is time-consuming and can be inconsistent when panels are small, closely arranged, partially obscured, or surrounded by visually similar structures. This paper presents a Hybrid Transformer-Based Solar PV Panel Detection and Energy Estimation System that combines convolutional feature extraction with Transformer-based contextual learning for pixel-level photovoltaic segmentation. A pretrained ResNet-18 backbone extracts hierarchical spatial features from rooftop imagery, while a Transformer Encoder models relationships among spatially separated image regions. A transposed-convolution decoder reconstructs the representation into a binary segmentation mask. The predicted mask is subsequently processed using contour extraction to obtain panel-level bounding boxes. The segmented area is converted into physical area using image spatial resolution and is then used to estimate installed photovoltaic capacity and expected daily electricity generation. The complete workflow is integrated into a Streamlit application for interactive image upload, detection visualization, panel localization, performance evaluation, and solar analysis. The study demonstrates that combining local CNN representations with global contextual modelling provides an effective basis for integrated rooftop PV detection, localization, and solar-energy assessment.

Keywords

Solar Photovoltaic Detection Rooftop Imagery Hybrid CNN-Transformer ResNet-18 Transformer Encoder Semantic Segmentation Solar Energy Estimation

Citation of this Article

J. S. S. Prasanna, & S. Chandra Sekhar. (2026). Hybrid Transformer-Based Solar PV Panel Detection and Energy Estimation System. Journal of Artificial Intelligence and Emerging Technologies (JAIET). 3(9), 36-48. Article DOI: https://doi.org/10.47001/JAIET/2026.309005 

Licence Copyright (c) 2026 Journal of Artificial Intelligence and Emerging Technologies. This work is licensed under a Creative Commons Attribution Non Commercial 4.0 International Licence.

References

  1. J. García, M. Fernandez, and L. Gomez, “Generalized deep learning model for photovoltaic module segmentation in aerial imagery,” Solar Energy, vol. 275, pp. 112–125, 2024.
  2. Y. Zhang, H. Li, and X. Wang, “Deep learning-based rooftop photovoltaic extraction framework using aerial imagery and building footprint data,” ISPRS International Journal of Geo-Information, vol. 14, no. 3, pp. 101–118, 2025.
  3. M. Kruitwagen, D. M. J. Tax, and R. van de Weijer, “Space-scale exploration of the poor reliability of deep learning models: The case of rooftop photovoltaic detection,” Environmental Data Science, vol. 4, pp. 1–20, 2025.
  4. S. Khan, A. Rehman, and M. Shah, “A comprehensive review of remote sensing and artificial intelligence for image analysis,” Sensors, vol. 25, no. 19, pp. 5965–5988, 2025.
  5. R. Sharma, P. Kumar, and S. Singh, “Survey of photovoltaic panel detection methods using image processing and deep learning,” Renewable and Sustainable Energy Reviews, vol. 178, pp. 113–130, 2024.
  6. L. Chen, Q. Wang, and Y. Li, “Deep learning in remote sensing image classification: A review,” Neural Computing and Applications, vol. 36, pp. 10235–10260, 2024.
  7. H. Zhao, X. Liu, and Y. Chen, “CPVPD-2024: A large-scale photovoltaic power plant dataset and deep learning detection framework,” Scientific Data, vol. 12, pp. 145–160, 2025.
  8. A.Patel, S. Mehta, and R. Gupta, “Deep learning-based rooftop photovoltaic detection and energy assessment using aerial imagery,” Sustainability, vol. 17, no. 15, pp. 6853–6870, 2025.
  9. T. Nguyen, J. Lee, and K. Park, “Comparative study of deep learning models for photovoltaic panel detection in aerial imagery,” IEEE Access, vol. 13, pp. 45021–45035, 2025.
  10. F. Müller, S. Bosch, and T. Erbertseder, “Monitoring rooftop photovoltaic development using deep learning and aerial imagery,” Applied Energy, vol. 356, pp. 121981, 2024.
  11. P. Li, X. Zhang, and Y. Wang, “Automated detection and tracking of photovoltaic modules using 3D remote sensing data,” Applied Energy, vol. 348, pp. 120–135, 2024.
  12. M. Santos, R. Torres, and J. Silva, “Fast automatic solar module geo-labeling from UAV thermal imagery using deep learning,” Solar Energy Advances, vol. 5, pp. 100–115, 2025.
  13. K. Patel, A. Sharma, and R. Singh, “Solar photovoltaic assessment using large language models and computer vision,” arXiv preprint arXiv:2507.19144, 2025.
  14. D. Kim, H. Park, and J. Kim, “Enhanced rooftop photovoltaic detection using CNN feature aggregation techniques,” arXiv preprint arXiv:2501.02840, 2025.
  15. Y. Zhou, Z. Wang, and Q. Li, “Survey of deep learning methods for remote sensing satellite image analysis,” IEEE Geoscience and Remote Sensing Review, vol. 61, pp. 1–20, 2023.
  16. S. Gupta, A. Verma, and R. Kumar, “AI-based solar panel detection using high-resolution satellite imagery,” Journal of Renewable Energy, vol. 2025, pp. 1–12, 2025.
  17. M. Rahman, K. Ahmed, and J. Kim, “Solar photovoltaic potential assessment using deep learning and remote sensing,” Energy Reports, vol. 11, pp. 245–260, 2025.
  18. L. Wang, Y. Chen, and H. Li, “Deep learning-based object detection in remote sensing imagery: A review,” IEEE Transactions on Geoscience and Remote Sensing, vol. 61, pp. 1–18, 2023.
  19. J. Peters, M. Brown, and S. Clark, “Automated rooftop solar panel detection using convolutional neural networks,” Remote Sensing, vol. 16, no. 4, pp. 850–865, 2024.