| With the rapid development of network technology and geographic information technology,more and more people can participate in the sharing of resources in the Internet,and user-created content has become an important content in Internet information.OpenStreetMap(OSM)road network data is a kind of road network data that is publicly participated and freely obtained.It can be released regularly,and the update speed is fast and strong.It is a good geographic data resource.However,because the uploader of OSM road network data does not have professional data collection knowledge,the uploaded road network data can not be compared with professional surveying and mapping data,and there are quality problems such as data irregularity or even data falseness.Therefore,it is necessary to evaluate the quality of the OSM road network before using it.Many scholars have proposed OSM road network data evaluation methods compared with high-quality vector reference data,but highprecision reference data sets are often difficult to obtain.Such methods cannot be widely used in the evaluation of OSM data quality.The remote sensing image has the advantages of convenient acquisition,wide coverage and strong current situation,which can be used as reference data for OSM road network data quality evaluation.Therefore,based on remote sensing imagery,this paper proposes a method for acquiring road network skeleton of remote sensing image based on Bayesian network verification and deep learning network model.The obtained road network skeleton is used as the reference data of OSM road network data,and the evaluation model of fuzzy reasoning is established to evaluate the quality of OSM road network data.The quality of OSM road network data is evaluated.The research content of this paper mainly focuses on the following aspects.Firstly,a quality evaluation method based on remote sensing image is proposed.Aiming at the difficulty of obtaining high-quality vector reference data and the difficulty of current situation,an evaluation method using remote sensing image as reference data is proposed.Two typical OSM data quality evaluation methods are analyzed.Then the quality evaluation methods of remote sensing images are proposed.The quality evaluation system of fuzzy reasoning is established.Finally,the flow of remote sensing image quality evaluation methods is summarized.Secondly,a data authentication method based on Bayesian network for OSM road network is proposed.The OSM road network data is simplified,the initial road skeleton is obtained,and then the road network and the image are registered.Aiming at the low reliability of road network data,this paper proposes a method to construct a Bayesian network model for verification by using the edge features,spectral features and vegetation characteristics of the road,and using the verified road segments as the basis road skeleton;Thirdly,a new road extraction method based on improved U-Net network model is proposed.Aiming at the new road that OSM road network can't cover,an improved U-Net network model extraction method is proposed.The existing sample library is updated by using the basic road skeleton acquisition sample to train the improved UNet network model,and the remote sensing image is predicted.The obtained road network is verified and merged with the existing road network to obtain a complete road network skeleton.Fourthly,the prototype system of OSM road network data evaluation based on remote sensing image was designed and developed.The method of this paper was tested by platform to verify the effectiveness of the proposed method.The method described in this paper uses a remote sensing image to evaluate the quality of OSM road network data from a new perspective.The experimental results show that the method is effective. |
PDF Full Text Request