A Precision Measurement Method for Rooftop Photovoltaic Capacity Using Drone and Publicly Available Imagery

Yue Hu, Yuce Liu, Yu Zhang, Hong‐Wei Dong, Chongzheng Li, Hongzhi Mao, Fusong Wang, Meng WangBuildings 2025 · DOI

TL;DR

This paper introduces a novel dual-optimization framework that combines monocular 3D reconstruction with a lightweight linear model to accurately estimate rooftop photovoltaic (PV) capacity from publicly available aerial imagery. By addressing geometric distortions inherent in non-orthorectified images, the method significantly improves PV area identification accuracy and provides a cost-effective solution for urban energy planning.

What Problem Does This Paper Address?

The global push for a low-carbon energy transition highlights the importance of urban rooftop photovoltaic (PV) systems. However, accurately assessing the installed capacity of these systems is a significant challenge. Publicly available aerial imagery, while cost-effective, often suffers from non-orthorectified projection effects. This means that buildings and PV arrays appear distorted due to perspective, leading to geometric inaccuracies in measurements and, consequently, errors in PV capacity prediction. Traditional methods struggle with these distortions, especially when metadata like camera parameters are unavailable. This paper aims to overcome these limitations by developing a precise and low-cost method for estimating rooftop PV capacity using readily available imagery.

Methods

The proposed solution employs a dual-optimization framework with several key components:

  • Monocular Vision-Based 3D Reconstruction: This technique leverages the inherent orthogonal structures of buildings (e.g., perpendicular walls and roof edges) to perform camera self-calibration. By identifying vanishing points, the method establishes geometric constraints to reconstruct the 3D geometry of buildings and their rooftops.
  • Vanishing Point Calibration: The paper utilizes the fspy tool to identify vanishing points from orthogonal lines in the image. These points are then used to calculate the camera's intrinsic matrix (focal length, principal point) and extrinsic matrices (rotation and translation) without requiring prior camera calibration.
  • Error Control Strategy: A crucial aspect is the multi-dimensional geometric constraint error control. After initial 3D reconstruction, the model's dimensions (roof length, width, and building height) are projected back onto the 2D image. If the residuals between the reconstructed and measured/known parameters exceed a 5% threshold, the calibration process is iteratively refined. This ensures that scale distortions are minimized.
  • Lightweight Linear Capacity Model: Once the precise PV surface area is extracted from the 3D model, a simple linear regression model is used to estimate the PV capacity. This model is trained on a dataset of manufacturer specifications, relating PV surface area directly to power generation capacity, thus avoiding complex physical modeling.

Datasets

The study utilized the following datasets:

  • Photovoltaic Array Image Dataset: This dataset comprises images of 20 PV-equipped buildings in Wuhan City. It includes publicly available Google Earth imagery and high-resolution drone images captured by a DJI Mavic 3 Classic drone.
  • Capacity-Area Manufacturer Dataset: This dataset consists of technical specifications from 19 mainstream PV manufacturers, covering over 200 PV module models. It was used to build the linear capacity-area model.

Key Insights and Ideas

  • Facade Distortion is a Major Source of Error: The study quantifies for the first time the significant impact of facade projection distortion on 2D PV area estimation. It shows a strong positive correlation between the facade area ratio and the PV area prediction error, highlighting the limitations of 2D methods.
  • Vanishing Points as Geometric Priors: Leveraging the orthogonal nature of buildings and vanishing points is an effective way to perform camera self-calibration and 3D reconstruction from a single monocular image, even without explicit camera parameters.
  • Iterative Error Control is Crucial: The proposed multi-dimensional geometric constraint error control strategy is key to mitigating scale distortions and ensuring the accuracy of the 3D reconstruction, even with imperfect initial calibration.
  • Simplicity in Capacity Modeling: A linear relationship between PV surface area and capacity, derived from manufacturer data, proves to be a robust and computationally efficient approach for capacity estimation, bypassing the need for complex physical models.

Key Results

  • Reduced Area Prediction Error: The 3D reconstruction method significantly reduced the Mean Absolute Percentage Error (MAPE) for PV area identification from 10.58% (achieved by a conventional 2D method) to 3.47%.
  • High Capacity Model Accuracy: The linear capacity-area model achieved a high coefficient of determination (R²) of 0.9548, indicating excellent predictive capability.
  • Cost-Effectiveness: The method relies on publicly available imagery and minimal building geometric priors, making it a low-cost solution for high-precision 3D data acquisition.
  • Robustness: The capacity model demonstrated robust generalization capability through rigorous five-fold cross-validation, with an average R² of 0.9503 ± 0.0108.

Limitations and Future Work

  • Limitations:
    • The linear capacity-area model is derived from current mainstream PV module technologies. Its accuracy may decrease with the emergence of significantly different future module designs (e.g., perovskite-silicon tandems, solar tiles).
  • The current study used a manual step for vanishing point selection. While the error control threshold was set at 5%, a more automated approach would be beneficial.
  • Future Work:
    • Developing more robust automatic vanishing point detection algorithms to reduce manual intervention.
  • Automating and intellectualizing the multi-dimensional geometric constraint error control process, potentially using optimization algorithms for automatic parameter iteration.
  • Periodically scraping the latest technical specifications from manufacturer websites to update the capacity-area model as PV technology evolves.

Summary updated

Abstract

Against the global backdrop of energy transition, the precise assessment of urban rooftop photovoltaic (PV) system capacity is recognized as crucial for optimizing the energy structure and enhancing the sustainable utilization efficiency of spatial resources. Publicly available aerial imagery is characterized by non-orthorectified issues; direct utilization is known to lead to geometric distortions in rooftop PV and errors in capacity prediction. To address this, a dual-optimization framework is proposed in this study, integrating monocular vision-based 3D reconstruction with a lightweight linear model. Leveraging the orthogonal characteristics of building structures, camera self-calibration and 3D reconstruction are achieved through geometric constraints imposed by vanishing points. Scale distortion is suppressed via the incorporation of a multi-dimensional geometric constraint error control strategy. Concurrently, a linear capacity-area model is constructed, thereby simplifying the complexity inherent in traditional multi-parameter fitting. Utilizing drone oblique photography and Google Earth public imagery, 3D reconstruction was performed for 20 PV-equipped buildings in Wuhan City. Two buildings possessing high-precision field survey data were selected as typical experimental subjects for validation. The results demonstrate that the 3D reconstruction method reduced the mean absolute percentage error (MAPE)—used here as an estimator of measurement uncertainty—of PV area identification from 10.58% (achieved by the 2D method) to 3.47%, while the coefficient of determination (R2) for the capacity model reached 0.9548. These results suggest that this methodology can provide effective technical support for low-cost, high-precision urban rooftop PV resource surveys. It has the potential to significantly enhance the reliability of energy planning data, thereby contributing to the efficient development of urban spatial resources and the achievement of sustainable energy transition goals.
BibTeX
@inproceedings{hu2025precision,
  author = {Yue Hu and Yuce Liu and Yu Zhang and Hong‐Wei Dong and Chongzheng Li and Hongzhi Mao and Fusong Wang and Meng Wang},
  title = {{A Precision Measurement Method for Rooftop Photovoltaic Capacity Using Drone and Publicly Available Imagery}},
  booktitle = {Buildings},
  year = {2025},
  doi = {10.3390/buildings15183377},
}
Figure