METHOD OF EXTRACTING THREE-DIMENSIONAL OBJECTS INFORMATION FROM A SINGLE IMAGE WITHOUT META INFORMATION

• US 20110216962A1
• Filed: 08/03/2010
• Published: 09/08/2011
• Est. Priority Date: 12/16/2009
• Status: Active Grant

• Current Assignee: INHA Industry Partnership Institute
• Original Assignee: INHA Industry Partnership Institute
• Inventors: Kim, Taejung, Lee, Taeyoon
• Granted Patent: US 8,437,554 B2
• Time in Patent Office: Days
• US Class Current: 382/154
• CPC Class Codes: G06T 2207/10032 Satellite or aerial image; Remote sensing
• CPC Class Codes: G06T 2207/30184 Infrastructure
• CPC Class Codes: G06T 7/507 from shading G06T7/586 takes precedence

Abstract

Disclosed herein is a method of extracting 3-dimension object information by a shadow analysis from a single image without meta information, and a technical problem to be solved is to extract three-dimension information of an object such as a height of the object and a footprint surface position of the object from a single image without meta information.

Claims

1. An three-dimension object information extraction method by a shadow analysis from a single image without meta information, comprising:

a first step of inputting a single image without meta information;

a second step of selecting a reference object from the image inputted through the first step, measuring a vertical line and a shadow line relative to the reference object from the input image and inputting a height of the reference object;

a third step of estimating azimuth and elevation angles or elevation angle proportion constants, that can replace the elevation angles, of a camera and a light source at image acquisition time using vertical-line and shadow-line information of the reference object measured from the input image and the height of the reference object;

a fourth step of selecting a specific object for finding a footprint position and the height of the object from the inputted image and extracting a roof surface outline of the selected object;

a fifth step of adjusting a height value of the object selected;

a sixth step of projecting a vertical line and a virtual shadow of the object corresponding to a height value adjusted through the fifth step onto the input image;

a seventh step of determining if the vertical lines or virtual shadow projected through the sixth step matches against the vertical lines or shadow of the input image; and

an eighth step of extracting a height and a footprint surface outline position of the object selected, when determined that the vertical lines of virtual shadow projected through the sixth step matches against the vertical lines or shadow of the image inputted through the seventh step.

2. The object information extraction method as set forth in claim 1, wherein the second step selects the reference object from an image inputted through the first step and measures a vertical line connecting a roof corner of the reference object and a footprint corner corresponding to the roof corner and a shadow line connecting a shadow-line end point corresponding to the roof corner and a footprint corner corresponding to the roof corner.

3. The object information extraction method as set forth in claim 1, wherein the third step estimates azimuth angles of a camera and a light source at image acquisition time using vertical line and shadow line, estimates elevation angles of the light source and the camera at that photographing time using a height of the reference object inputted through the second step, ground sampling distance of the image, the vertical line and the shadow line information of the reference object measured from the input image, or estimates elevation angle proportion constants replaceable of the elevation angles of the light source and the camera if ground sampling distance of the image is unknown.

4. The object information extraction method as set forth in claim 1, wherein the sixth step generates a vertical line and a virtual shadow of the object corresponding to a height value adjusted through the fifth step using azimuth and elevation angles or elevation angle proportion constants replaceable of the elevation angles of the camera and the light source at image acquisition time estimated through the third step, and projects a vertical line and virtual shadow of the object onto the input image.

5. The object information extraction method as set forth in claim 1, wherein the seventh step determines whether the vertical lines or virtual shadow projected through the sixth step matches against the vertical lines of shadow of the inputted image, and in a case of a mismatch, re-adjusts a height value of the object by repeating the fifth step procedure.

6. The object information extraction method as set forth in claim 1, wherein the eighth step extracts a height of the object adjusted through the fifth step and determines it as an actual height of the object when determined through the seventh step that the vertical lines or virtual shadow projected at the sixth step matches against the vertical lines or shadow of the image inputted through the first step, and extracts a footprint surface outline position of the object by parallel-moving a roof surface outline of the object according to a vertical line, and determines a position of the object in the input image through the footprint outline position extracted.