Picture geographic location method and system based on large-scale street view data

Abstract

The invention discloses a picture geographic location method based on large-scale street view data and a system for realizing the method. The method comprises a real-time matching part, namely the following steps: extracting local feature points from a queried picture as query feature points, and generating a query feature set; in an index structure, searching K nearest neighbor feature points foreach query feature point, and generating a nearest neighbor feature distribution table; obtaining a nearest neighbor feature set of a candidate picture by virtue of the nearest neighbor feature distribution table; and selecting a candidate picture corresponding to an optimal feature set in the nearest neighbor feature set as an optimal matched picture, wherein a geographic location correspondingto the matched picture is regarded as the geographic location of the queried picture, and K is an integer. The method also can comprise a preprocessing part preferably, namely a time-consuming 'indexstructure creating part' which is separately processed in advance. The method and system which are disclosed by the invention has the characteristics that location accuracy is high, matching speed ishigh and globally optimal solution can be found.

Description

technical field [0001] The present invention relates to the technical field of image positioning, in particular to a method and system for image geographic positioning based on large-scale street view data. Background technique [0002] Image geolocation is to determine or estimate the geographic location of an image based on its visual information. This problem has many application scenarios, such as automatic driving of cars, mobile terminal navigation, video geographic trajectory generation, etc. In recent years, street view images have been continuously updated and improved, and in some cities, full coverage of street views can be achieved. There are many identifiable visual information in the picture, such as buildings, billboards, etc. We can use the method based on feature matching to identify these information, so as to achieve the purpose of geographical location of the picture. [0003] In the field of academic research, authors such as Zamir proposed that each q...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problems of slow matching speed of street view pictures, only local optimal solutions can be obtained and low positioning accuracy, the present invention discloses a picture geolocation method based on large-scale street view data, which can perform geolocation only based on the visual information of pictures, It has the characteristics of fast matching speed, global optimal solution and high positioning accuracy

Images