| With the emergence of climate change,environmental pollution and primary energy exhaustion,the focus of energy exploitation and use in China has gradually shifted from fossil energy to renewable new energy.As a representative of new energy,photovoltaic power station is developing rapidly in China,with an increasing number of installations every year.At the same time,the development of high-resolution remote sensing satellite technology in recent years has played a huge role in promoting the improvement of China’s earth observation level,enabling China to obtain more accurate remote sensing data sources from many aspects.Therefore,in this paper,a variety of remote sensing data sources were integrated,and the area of photovoltaic power station in Quyang County,Baoding City,Hebei Province and its vicinity were taken as the research area.Photovoltaic power station targets in World View-Ⅱ remote sensing images were detected by the improved feature pyramid method.Gaofen-1 and Landsat 8 remote sensing images were used to study land cover and surface temperature changes around photovoltaic power stations in mountainous areas.This paper aims to detect photovoltaic power stations through remote sensing images and analyze certain environmental images,providing some ideas and references for the planning and construction,management decisions and environmental protection of photovoltaic power stations in the future.The work done in this paper is as follows:1.In view of the shortage of target data sets of photovoltaic power stations in remote sensing images,this paper annotated the target data of photovoltaic power stations in pre-processed World View-Ⅱ remote sensing images according to the construction method of public data sets,and constructed target detection data sets for subsequent training,testing and verification of photovoltaic power station detection.2.Aiming at the problem that the FPN(Feature Pyramid Network)model has low target detection accuracy for photovoltaic power station due to the complex surrounding environment and irregular distribution of power station cluster,this paper improved the FPN model and proposed the structure of multistage feature extractor and error screening mechanism.Conv2 and Conv3,Conv4 and Conv5 were fused into shallow feature map and deep feature map respectively to improve the feature extraction effect of the proposed method for photovoltaic power station targets.Through experimental verification and result analysis,it can be seen that the detection accuracy AP of the proposed method for photovoltaic power station targets reaches 83.6%,which is improved by 8.3% compared with the original FPN model,and can achieve effective detection of photovoltaic power station targets.3.There are few quantitative studies on land cover types and surface temperature changes in the surrounding environment caused by the massive construction of photovoltaic power stations in mountainous areas.Thus,this paper used pre-processed medium-high resolution Gaofen-1 remote sensing images to conduct land cover research in the study area through supervised classification method,and made land cover change maps and statistical data to calculate land transfer matrix.Using infrared information from medium resolution Landsat 8 images,land surface temperature inversion was carried out in the study area by atmospheric correction method.By superposition analysis of temperature retrieval results and land cover changes,the corresponding surface temperature of each land type and the area proportion of each land type under different temperature classification can be obtained,so as to focus on the influence of the distribution and proportion of photovoltaic power stations on the spatial pattern of thermal environment in the region.Through experimental verification and result analysis,it can be seen that under the current geographical location and time conditions in the study area,the proportion of photovoltaic power stations shows a strong correlation with average surface temperature,which is one of the main factors causing temperature rise in mountainous areas. |
