Building height information is the basic data of urban 3D modeling,but most of the existing building height estimation studies use remote sensing images such as LiDAR and SAR.With the rapid development of computers and the internet,street view data has become an emerging data source for building height estimation due to its easy collection and low cost.In this study,we propose a method for estimating the height of buildings in street view images.Firstly,the segment anything model is used to extract the pixel height of the building in the image,then,the image metadata and electronic map data are used to obtain the distance between the building and the camera,and the image focal length.According to the geometric relationship between the street view image and the building entity,the pinhole camera model is improved,and the building height estimation method is constructed.Finally,the Mapillary street view images of Beijing and Berlin are selected for experimental verification.The results show that the accuracy of height estimation is significantly improved compared with that before the improvement of the pinhole camera model,the RMSE is reduced by 11.31 meters,and the R2 is increased by 0.4,which is of practical value.
维普
万方数据