Apparatus and method for extracting correspondences between aerial images

• US 9,390,330 B2
• Filed: 12/14/2011
• Issued: 07/12/2016
• Est. Priority Date: 12/16/2010
• Status: Active Grant

• Current Assignee: Electronics and Telecommunications Research Institute
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: Park, Il-Kyu, Chu, Chang-Woo, Cha, Young-Mi, Koo, Bon-Ki
• Primary Examiner(s): Vu, Kim
• Assistant Examiner(s): HU, FRED H.
• Application Number:
  US13/325,957
• Publication Number:
  US 20120155745A1
• Time in Patent Office: 1,672 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06T 17/05 Geographic models
• CPC Class Codes: G06T 2207/10032 Satellite or aerial image; Remote sensing
• CPC Class Codes: G06T 2207/30184 Infrastructure
• CPC Class Codes: G06T 7/33 using feature-based methods
• CPC Class Codes: G06V 20/176 Urban or other man-made structures

Abstract

Disclosed herein is an apparatus and method for extracting correspondences between aerial images. The apparatus includes a line extraction unit, a line direction determination unit, a building top area extraction unit, and a correspondence extraction unit. The line extraction unit extracts lines corresponding buildings from aerial images. The line direction determination unit defines the directions of the lines as x, y and z axis directions based on a two-dimensional (2D) coordinate system. The building top area extraction unit rotates lines in the x and y axis directions so that the lines are arranged in parallel with the horizontal and vertical directions of the 2D image, and then extracts building top areas from rectangles. The correspondence extraction unit extracts correspondences between the aerial images by comparing the locations of the building top areas extracted from the aerial images.

Claims

1. An apparatus for extracting correspondences between aerial images, comprising:

one or more processors configured to;

extract lines corresponding buildings from two or more aerial images;

define directions of the lines as x, y and z axis directions based on a coordinate system of a two-dimensional (2D) image;

rotate lines in the x axis direction and lines in the y axis direction so that the lines in the x axis direction are arranged in parallel with a horizontal direction of the 2D image and the lines in the y axis direction are arranged in parallel with a vertical direction of the 2D image, and then extract building top areas from rectangles that are formed by the rotation;

extract correspondences between the aerial images by comparing locations of the building top areas extracted from the aerial images;

create a direction distribution histogram of the lines, and define directions having three highest peak values in the direction distribution histogram as x, y, and z axis directions, respectively;

define one of the highest peak value directions closest to a vertical direction as the z axis direction and two remaining highest peak value directions as the x and y axis directions, respectively; and

define an arbitrary one of the two remaining highest peak value directions as the x axis direction and a remaining highest peak value direction as the y axis direction.

2. The apparatus as set forth in claim 1, wherein the lone or more processors are further configured to extract lines corresponding to edges and contours of the buildings from among lines that are selected from among the lines of the aerial images and fall within a preset reference range.

3. The apparatus as set forth in claim 1, wherein the one or more processors are further configured to, when each of the building top areas is not a rectangle, create a rectangle by extending the lines in the x and y axis directions.

4. The apparatus as set forth in claim 1, wherein the one or more processors are further configured to define central points of the building top areas as feature points and extracts correspondences between the aerial images based on locations of the feature points and the edges of the building top areas.

5. The apparatus as set forth in claim 4, one or more processors are further configured to calculate a transformation matrix according to rotation and translation transformation of the buildings between the aerial images, from sums of squares of distances between coordinates of the feature points of the aerial images.

6. The apparatus as set forth in claim 5, wherein the one or more processors are further configured to extract corresponding points of another aerial image for the feature points using the transformation matrix, and compare locations of the feature points with locations of the corresponding points of the latter aerial image.

7. A non-transitory computer readable medium encoded with a computer program for extracting correspondences between aerial images, the program when executed by a computer causes the computer to perform a method comprising:

extracting lines corresponding buildings from two or more aerial images;

defining directions of the lines as x, y and z axis directions based on a coordinate system of a 2D image;

rotating lines in the x axis direction and lines in the y axis direction so that the lines in the x axis direction are arranged in parallel with a horizontal direction of the 2D image and the lines in the y axis direction are arranged in parallel with a vertical direction of the 2D image, and then extracting building top areas from rectangles that are formed by the rotation;

extracting correspondences between the aerial images by comparing locations of the building top areas extracted from the aerial images;

creating a direction distribution histogram of the lines, and defines directions having three highest peak values in the direction distribution histogram as x, y, and z axis directions, respectively;

defining one of the highest peak value directions closest to a vertical direction as the z axis direction and two remaining highest peak value directions as the x and y axis directions, respectively; and

defining an arbitrary one of the two remaining highest peak value directions as the x axis direction and a remaining highest peak value direction as the y axis direction.

8. The non-transitory computer readable medium as set forth in claim 7, wherein the extracting lines extracts lines corresponding to edges and contours of the buildings from among lines that are selected from among the lines of the aerial images and fall within a preset reference range.

9. The non-transitory computer readable medium as set forth in claim 7, wherein the extracting building top areas comprises, when each of the building top areas is not a rectangle, creating a rectangle by extending the lines in the x and y axis directions.

10. The non-transitory computer readable medium as set forth in claim 7, wherein the extracting correspondences comprises defining central points of the building top areas as feature points, and extracts correspondences between the aerial images based on locations of the feature points and the edges of the building top areas.

11. The non-transitory computer readable medium as set forth in claim 10, further comprising calculating a transformation matrix according to rotation and translation transformation of the buildings between the aerial images, from sums of squares of distances between coordinates of the feature points of the aerial images.

12. The non-transitory computer readable medium as set forth in claim 11, wherein the extracting correspondences further comprises:

extracting corresponding points of another aerial image for the feature points using the transformation matrix; and

comparing locations of the feature points with locations of the corresponding points of the latter aerial image.