| Currently,automatic or semi-automatic extraction of objects from very high resolution(VHR)remotely sensed imagery has been one of the topical research themes in the communities of photogrammetry and remote sensing,computer vision,pattern recognition,artificial intelligence etc.Research on automatic extraction method of high-resolution remote sensing images has been carried out at home and abroad,but there has been no related breakthrough progress.On the contrast,the semi-automatic road extraction method has made some remarkable achievements due to the combination of human cognitive ability and computer analysis ability.However,the existing semi-automatic road extraction methods has lower generality and automation.For this reason,this paper proposes a semi-automatic road extraction method based on circular template matching,in order to get stronger generality and automation.Firstly,with related contrast filtering experiments,the LO filter is selected to smooth the image.Secondly,the seed point is selected in the human-computer interaction mode.The prior information of the path is acquired by the adaptive mechanism,and the circular template is initialized.Then,establishing a gray histogram of circular template,calculating the gray value of circular template and the peak ratio of histogram,combining the texture variance value of circular template is used as the basis for judging the matching of the circular template.Furthermore,according to the analysis of the extraction of road interference template tracking during the matching process,dynamic mechanism,construction of obstacle crossing road main road direction and template update mechanism.At the same time,elimination of false seed points based on topological connectivity.The experienments of extraction by several groups of different types of high resolution remote sensing image show that,compared with other methods,this method can effectively extract road information,and has certain superiority and robustness. |
PDF Full Text Request