Segmented Correction Method Of Vector Data For Rural Road Census

High-precision rural road vector data is the foundation of rural informatization construction,and it is also of great significance for the country to formulate rural road development plans and implement rural revitalization strategies.However,due to the changes of ground objects,coordinate system changes,differences in the accuracy of collection equipment,and experience of collection personnel,the existing rural vector road data has serious accuracy problems.It is urgent to carry out a rural road census to provide basic data protection for the national rural revitalization strategy.However,the existing vector road data has irregular deformation,and it is difficult to clarify the complex relationship between the vector data and the road extraction method in this case.In order to ensure the correction accuracy of the existing vector data,at this stage,the internal work of the rural road census is still mainly manual,and the degree of automation is low.Therefore,in order to improve the automation degree of vector correction,this paper simulates the method of manual segmented vector data correction,and proposes a vector data segmented correction method for rural road census.The main work is as follows:(1)Preprocessing.Firstly,in order to highlight the geometrical structure features of the objects in the image,the line segment sequence algorithm is used to obtain the line segment sequence data of the object edge.The edge data can enhance the machine’s ability to express the edge of curved ground objects,and expand the range of the machine’s perception of the geometric structure of ground objects.Secondly,in order to increase the texture difference between the road and the background in the image,the multispectral image in the RGB space is converted into the HSV space to enhance the spectral heterogeneity between different objects.Finally,in order to effectively use the edge pixel data that cannot form the road edge line segment,this paper uses the phase data to express the characteristics of the road edge pixel point direction,and generates a phase image.(2)A multi-feature constrained segmented registration model is proposed.In this paper,considering that the road is an artificial feature with a large aspect ratio,the line segment sequence is used to express the edge of the road.Based on the strong ability of line segment data to express curved edges,a segmented registration correction model guided by vector data is established.Based on the relationship between road image features and vector data,as well as road texture features and geometric features in the image,the model designs line segment geometric constraints,vector data position constraints,and regional texture constraints.Then,Based on these three constraints,automatic the corrected vector data.(3)Segmented vector data correction method.Considering that the road data in the image often encounters problems such as the similarity of the road to the background,the interference of other ground objects,and the occlusion.Therefore,in order to ensure the accuracy of the corrected data and improve the automation degree of the correction method.This paper firstly designs a dynamic direction weighted matching model based on the direction information of the road to solve the problem that the road and the background cannot be distinguished due to the small heterogeneity of the road and the background texture.For long areas,a line segment sequence tracking mechanism is established to improve the efficiency of the correction process;then,for the areas where the correction direction cannot be obtained,the phase matching model is used to correct the vector;finally,the topological connection is used to correct the vector data in the completely occluded area.Through the orderly combination of preprocessing,registration model,and segmented correction methods,the algorithm can correct vector data in different types,regions,and scenarios.Experiments show that under the premise of ensuring the quality of vector correction,the efficiency and automation of the algorithm in this paper are significantly better than the existing methods.This paper has 33 figures,4 tables and 60 references.