1 | Calibration, point matches, and line segment matches |
---|
2 | of a close-range architectural image set |
---|
3 | ---------------------------------------------------- |
---|
4 | |
---|
5 | This set contains images of an Oxford University building, taken by |
---|
6 | Olympus C-820L digital camera. The camera was held by a person |
---|
7 | standing on the ground. The following was applied to the set: |
---|
8 | |
---|
9 | 1. Automated wide baseline stereo: affine & scale invariant features |
---|
10 | were detected in each image and matched by RANSAC, yielding projective |
---|
11 | camera matrices and a cloud of projective 3D points. |
---|
12 | |
---|
13 | 2. Metric upgrade from vanishing points: vanishing points computed |
---|
14 | from detected line segments and (along with square pixels assumption) |
---|
15 | used to upgrade the projective reconstruction to metric one. |
---|
16 | |
---|
17 | 3. Automated line segment matching: the line segments matched and |
---|
18 | reconstructed, using geometrical constraints on lines and photommetric |
---|
19 | information in their neighborhood. |
---|
20 | There are some mismatches among the matched lines, mainly due to |
---|
21 | ill conditioning of the geometry for horizontal lines (because camera |
---|
22 | centers all lie in about horizontal plane). |
---|
23 | |
---|
24 | |
---|
25 | References: |
---|
26 | |
---|
27 | 1. Werner, T. and Zisserman, A. |
---|
28 | New Techniques for Automated Architecture Reconstruction from Photographs |
---|
29 | Proc. 7th European Conference on Computer Vision, Copenhagen, Denmark, 2002. |
---|
30 | http://www.robots.ox.ac.uk/~vgg/publications/papers/werner02.{pdf,ps.gz}. |
---|
31 | |
---|
32 | 2. Werner, T. and Zisserman, A. |
---|
33 | Model Selection for Automated Architectural Reconstruction from Multiple Views |
---|
34 | Proc. British Machine Vision Conference, 2002. |
---|
35 | http://www.robots.ox.ac.uk/~vgg/publications/papers/werner02a.{pdf,ps.gz}. |
---|