遥感影像道路断线修复方法研究OACSTPCD

In remote sensing image road extraction,the road break caused by trees or building shadows greatly affects the integrity and connectiv-ity of road,which makes the results unable to provide reliable data support for the analysis and processing methods based on the road network.Based on the road extraction results from remote sensing images of deep semantic segmentation network,we proposed a road break repair meth-od,which combining the location,direction,texture and connectivity of the road break points,and combining the matching road break points with cubic curve fitting to optimize the connection.Experimental results show that the method improves the F1-Score of DeepGlobe Road datas-et by 1.27%compared with the original result.The accuracy and reliability of original results are greatly improved.Moreover,the results after break repair have good integrity and connectivity,which reflects the reliability and practicability of this method for remote sensing image road break repair.