Point-based iterative remote sensing image road extraction method

Abstract

The invention discloses a point-based iterative remote sensing image road extraction method. Compared with on-site collection, automatic extraction of the road centerline map from the remote sensing image is higher in efficiency and lower in cost. In order to improve road connectivity and keep accurate alignment between a road map and a real road center line, the invention provides a point-based iterative road map exploration method using segmentation clue guidance, variable step length and trajectory exploration. Wherein the segmentation clues are embodied as follows: center line segmentationand intersection point segmentation are used as supervision information in the neural network; wherein the variable step length is embodied in that an adjustable step length training neural network is used at a road intersection point, a road end point and a connection point, and the trajectory exploration method is embodied in that a next drop point set starting from an image center point according to a time sequence is obtained by utilizing one-time remote sensing image input.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a method and application for extracting roads from remote sensing images by using a neural network. Background technique [0002] Extracting roads from remote sensing images is a research topic in the field of remote sensing. Traditional approaches construct road maps through various techniques, such as utilizing prior knowledge of nearby buildings and vehicles (hinz et al.) [ISPRS journal of photogrammetry and remote sensing, 2003, 58(1-2):83-98], shape factor (song et al.) [Photogrammetric Engineering & Remote Sensing, 2004, 70(12): 1365-1371], simulated annealing technique (stoica et al.) [International Journal of Computer Vision, 2004, 57(2): 121-136], Spectral contrast and local linear trajectories (das et al.) [IEEE Transactions on Geoscience and Remotesensing, 2011, 49(10):3906-3931]. Furthermore, minimum spanning trees (turetken et al.) [IEEE Confere...

AI Technical Summary

Problems solved by technology

[0008] The purpose of the present invention is to solve the problem that the existing methods are difficult to achieve both graph-level connectivity and pixel-level accuracy, and propose an iterative road that uses road segmentation as a guide, supplemented by dynamic step size and trajectory exploration Graph Exploration Methods

Images