Building Roof Extraction And Solar Energy Potential Assessment Based On High Resolution Remote Sensing Image

With the rapid development of the global economy,fossil energy will eventually be exhausted.All countries in the world are vigorously promoting the development of abundant and pollution-free solar energy resources.The development of solar rooftop photovoltaic has been increasingly proved to have great advantages and application prospects due to its nearby utilization,no regional restrictions,land cost saving and other characteristics.China has a vast territory and abundant solar photovoltaic resources,but it lags behind in the development and research of solar rooftop photovoltaic.Up to now,there is still a lack of a fast and accurate method to evaluate the potential of solar photovoltaic on building roofs.Based on the above problems,this paper proposes a solar roof photovoltaic potential assessment method based on high-resolution remote sensing images and full convolutional neural network.In combination with Google Earth high-resolution remote sensing image,the algorithm of deep learning is used to automatically extract the accurate contour of the building roof from the remote sensing image.Then the potential of rooftop solar energy is evaluated automatically according to the solar radiation model.By improving relevant methods or technologies,a set of available high-resolution remote sensing image area roof photovoltaic potential assessment method is finally developed to realize accurate and rapid assessment of roof photovoltaic potential and provide auxiliary decision-making for integrated development of building photovoltaic.The research content of this paper mainly includes the following aspects:(1)High-resolution remote sensing image building extraction based on full convolutional neural network.A large number of high-resolution remote sensing images in Google Earth were used to establish the building data set that met the requirements.In order to enhance the learning and generalization ability of the model,a variety of data enhancement methods were adopted for data set enlargement.Satellite image data is large in volume and low in resolution.In order to realize automatic and accurate extraction of building roof,this paper constructs full-convolution semantic segmentation model U-Net and Seg Net respectively to realize end-to-end recognition,and transforms their network layer according to the actual effect.In order to achieve the optimal model performance,the methods of different models and different network structures are respectively tried to compare the recognition accuracy of buildings on the local data set,study the functions and characteristics of each network layer,adjust the optimal network structure and super parameters,and improve the segmentation precision of the existing network.(2)Model fusion and result image post-processing.Based on the established U-Net and Seg Net building extraction models,this paper adopts the idea of integrated learning to conduct weighted voting fusion of multiple output prediction graphs of the two kinds of models(the same model uses different parameters and network layer for training and prediction),so as to further improve the robustness of building extraction model.Then,the morphological operation was used to optimize the building area contour map,eliminate the burr,hole and other noise information,and extract a relatively neat building roof contour for later solar energy potential calculation.(3)Assessment of the potential of building rooftop solar photovoltaic.The roof area,orientation and roof type data are calculated according to the extracted roof results,and the roof shape changes are simulated through grid division.The solar radiation model was designed to calculate the solar radiation amount per unit area of the roof and the radiation proportional coefficient of different roof cell grids were calculated,and the plane area was converted based on the proportional coefficient.Then,different photovoltaic cell array paving methods are set for flat roof and non-flat roof,and the area of solar panels that can be installed on the roof is calculated.Finally,the potential of building roof photovoltaic was evaluated by integrating roof photovoltaic area and photovoltaic system parameters.