Road Vector Data Extraction And Matching Positioning Application Based On Remote Sensing Image

As an important feature target,roads play an important role in many fields such as transportation,urban planning,and military affairs.This makes the automatic extraction of road vector data practically important.However,due to the complexity of remote sensing image scenes,it is easy to be blocked and shadowed Interference,research on efficient and accurate road vector data extraction algorithms is still very challenging.For the visual navigation system,the road vector data has significant information and a small amount of data,which can provide more flexible and larger area reference guarantees for the system.However,there is currently very little research on the application of remote sensing image matching based on road vector data.The main research work is as follows:In order to realize the extraction of road vector data of remote sensing images in complex environments,a road vector data extraction algorithm of remote sensing images based on semantic segmentation is proposed.First,the road extraction module is used to extract the road area,and then the grid is vectorized based on the extraction result,which the key points are the road extraction module and the broken connection.The road extraction module adopts the improved semantic segmentation network based on U-net network,which has strong feature expression ability and context information perception ability.The vectorization process uses the shortest path algorithm for disconnection,which improves the connectivity and integrity of the results.Experimental results show that,compared with other methods,the road extraction module of this method has strong anti-interference ability,high vectorization accuracy,and is suitable for complex environments.In order to realize the matching and positioning application of road vector data guarantee,a real-time matching and positioning scheme for remote sensing images based on road vector data is proposed.The system inputs the real-time image and the reference image,calls the road extraction module to extract the road features of the real-time image,and uses the correlation coefficient as the similarity measure for matching and positioning in the common road feature space,which the key point is the real-time road extraction module.The road extraction module adopts the method of semantic segmentation to extract road category feature,introduces deep separable convolution and improved encoder for real-time lightweighting,and uses the generative adversarial network to assist the model for weakly supervised learning.The experimental results show that,compared with other methods,the road extraction module of this method has real-time light weight and high accuracy,and the overall scheme matching accuracy is high.It can still achieve high-precision matching under large interference.