Large-scale solar panel mapping from aerial images using deep convolutional networks

Yuan, Jiazheng; Yang, Hsiuhan Lexie; Omitaomu, Olufemi A.; Bhaduri, Budhendra

doi:10.1109/bigdata.2016.7840915

Cited by 52 publications

(29 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, fully supervised segmentation has relatively poor computation efficiency. 6,7 To enable efficient solar panel identification and segmentation, DeepSolar first utilizes transfer learning 12 to train a CNN classifier on 366,467 images sampled from over 50 cities/towns across the US with merely image-level labels indicating the presence or absence of panels. Segmentation capability is then enabled by adding an additional CNN branch directly connected to the intermediate layers of the classifier, which is trained on the same dataset to greedily extract visual features to generate clear boundaries of solar panels without any supervision of actual panel outlines.…”

Section: Scalable Deep Learning Model For Solar Panel Identificationmentioning

confidence: 99%

“…Such a result is significantly higher than previous reports. [6][7][8]13 Furthermore, our performance evaluation guarantees far more robustness since their test sets were only obtained from one or two cities but ours are sampled from nationwide imagery. Mean relative error (MRE), the area-weighted relative error, is used to measure size estimation performance.…”

Section: Scalable Deep Learning Model For Solar Panel Identificationmentioning

confidence: 99%

“…5 The availability of satellite imagery with spatial resolution less than 30 cm for the majority of the US, which is annually updated, offers a rich data source for solar installation detection based on machine learning. Existing pixel-wise machine learning methods 6,7 suffer from poor computational efficiency, and relatively low precision and recall (cannot reach 85% simultaneously), while existing image-wise approaches 8 cannot provide system size or shape information. Google's Project Sunroof utilizes a proprietary machine learning approach to report locations without any size information.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

DeepSolar: A Machine Learning Framework to Efficiently Construct a Solar Deployment Database in the United States

Wang

Majumdar

et al. 2018

Joule

236

130

View full text Add to dashboard Cite

We developed an accurate deep learning framework to automatically localize solar photovoltaic panels from satellite imagery and estimate their sizes. We used it to construct a comprehensive and publicly available solar installation database of the contiguous US. We demonstrated its value by identifying key environmental and socioeconomic factors correlating with solar deployment, such as income and education. We also found that the solar deployment density can be accurately estimated at the microscopic level with these factors using a novel predictive model.

show abstract

Section: Scalable Deep Learning Model For Solar Panel Identificationmentioning

confidence: 99%

Section: Scalable Deep Learning Model For Solar Panel Identificationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

DeepSolar: A Machine Learning Framework to Efficiently Construct a Solar Deployment Database in the United States

Wang

Majumdar

et al. 2018

Joule

236

130

View full text Add to dashboard Cite

show abstract

“…Some previous deep learning-based semantic segmentation methods have been applied to the identification of distributed photovoltaic power stations. Yuan [55] was the first to introduce an FCN model for distributed photovoltaic power station identification. However, the adopted FCN model requires up-sampling by a large multiple, which may cause the loss of feature information.…”

Section: Introductionmentioning

confidence: 99%

Combined Multi-Layer Feature Fusion and Edge Detection Method for Distributed Photovoltaic Power Station Identification

Jie

Yue

et al. 2020

Energies

View full text Add to dashboard Cite

Distributed photovoltaic power stations are an effective way to develop and utilize solar energy resources. Using high-resolution remote sensing images to obtain the locations, distribution, and areas of distributed photovoltaic power stations over a large region is important to energy companies, government departments, and investors. In this paper, a deep convolutional neural network was used to extract distributed photovoltaic power stations from high-resolution remote sensing images automatically, accurately, and efficiently. Based on a semantic segmentation model with an encoder-decoder structure, a gated fusion module was introduced to address the problem that small photovoltaic panels are difficult to identify. Further, to solve the problems of blurred edges in the segmentation results and that adjacent photovoltaic panels can easily be adhered, this work combines an edge detection network and a semantic segmentation network for multi-task learning to extract the boundaries of photovoltaic panels in a refined manner. Comparative experiments conducted on the Duke California Solar Array data set and a self-constructed Shanghai Distributed Photovoltaic Power Station data set show that, compared with SegNet, LinkNet, UNet, and FPN, the proposed method obtained the highest identification accuracy on both data sets, and its F1-scores reached 84.79% and 94.03%, respectively. These results indicate that effectively combining multi-layer features with a gated fusion module and introducing an edge detection network to refine the segmentation improves the accuracy of distributed photovoltaic power station identification.

show abstract

“…To overcome this problem, [5] integrates feature maps from different stages of CNN to produce a prediction map. This approach achieves PV array mapping by pixel-wise prediction based on hierarchically integrated feature maps.…”

Section: Introductionmentioning

confidence: 99%

Automatic Detection and Mapping of Solar Photovoltaic Arrays with Deep Convolutional Neural Networks in High Resolution Satellite Images

Zhang

2020

2020 IEEE 4th Conference on Energy Internet and Energy System Integration (EI2)

View full text Add to dashboard Cite

The locations and capacities of household rooftop solar photovoltaic (PV) arrays are important for power grid planning. However, it is hard to collect such information manually as a significant number of PV arrays are distributed dispersedly in residential areas. With the development of deep learning model on image recognition, it brings an opportunity to build an intelligent detector that is able to automatically identify and delineate PV arrays in satellite images. Convolutional neural networks (CNN) are ideally suited for this task as CNN has capability of capturing spatial information of digital images by convolution operation. In this work, we trained a deep CNN with manually annotated satellite images taken from a region of Manchester (UK), and validated our detector with another set of satellite images taken from the same city. Our results indicate that the detector is capable to identify PV arrays with a high accuracy and delineate them in pixel-wise with high precision, showing the feasibility of our approach.

show abstract

Large-scale solar panel mapping from aerial images using deep convolutional networks

Cited by 52 publications

References 6 publications

DeepSolar: A Machine Learning Framework to Efficiently Construct a Solar Deployment Database in the United States

DeepSolar: A Machine Learning Framework to Efficiently Construct a Solar Deployment Database in the United States

Combined Multi-Layer Feature Fusion and Edge Detection Method for Distributed Photovoltaic Power Station Identification

Automatic Detection and Mapping of Solar Photovoltaic Arrays with Deep Convolutional Neural Networks in High Resolution Satellite Images

Contact Info

Product

Resources

About