We propose a framework for three-dimensional (3D) model reconstruction
using structure-from-motion (SfM). Our primary application
is scanning forest canopies for ecological research (as shown in
Figure 1) using photos taken from a radio-controlled helicopter, although
the methods apply to any series of photographs taken along
a path with the camera pointing perpendicular to the path.
Previous works in this area include Snavely et al.’s Bundler [2006]
where the datasets are unordered photos of famous landmarks
downloaded from the Internet; using structure-from-motion, the geometries
are reconstructed, but this process takes a long time. Our
approach is based on the fact that the photos are taken in a continuous
path, so we can utilize the image adjacencies. The major
contributions of our work are 1) introduce a linear time complexity
algorithm that reduces number of image pairs to match; 2) present
a high performance framework for 3D reconstruction using SfM.