Summary. These are my assignments for the course mentioned in the title, held by professor Marcelo Gattass. The course took place during the 2006.2 semestre.
In this work, I assemble a simple video-capturing application that will serve as the basis of much work done during the course. I use the IM library to capture webcam images in real time, and GLUT for windowing.
In this work, I apply functions developed for the TecGraf project Juiz Virtual to recognize a (planar) 3-D pattern. The method used is the so-called model-based recognition, where the current set of recognized features (lines) in the captured raster image is being compared against a standard 3-D planar model, all in real-time.
 |
 |
| Figure: a "hit" (the planar 3-D model has been recognized) - white lines being used | Figure: a "miss" (the planar 3-D model has not been recognized - color lines being used for lines detected) |
In this work, I apply the Zhang camera calibration method to find intrinsic (focal length, etc.) and extrinsic (rotation, translation) parameters for three real and one virtual camera (whose calibration results are shown in the image below).
Figure: reconstructed scene
In this work, I develop a simple, show-case programs which illustrates the application of computer vision concepts to architecture. Specifically, the idea is to have a 3-D model of the future skyscraper superposed on real scene, thus giving rise to the so-called augmented reality - i.e., having a real scene combined with a 3-D model in order to see how the place (building site) would look like if the corresponding skyscraper would be constructed in that place.
Figure: real scene (left) and augmented scene (right)
Rio de Janeiro, dezembro 2006.