| |
| Acadêmico(a): Kevin Stortz |
| Título: Biblioteca para reconstrução de relevos |
| |
| Abstract: |
| Terrain recognition and 3D reconstruction are computer processes that have a high degree of computational complexity. Because they need to perform many correlation algorithms between two distinct images. Concepts as Epipolar Geometry aim to decrease the complexity to perform it and improve the definition of correlation from two distinct visualizations of the same scene. It will be used a camera pinhole model, that will be needed the calibration process, to define the intrinsic parameters matrix, the radial distortion coefficient and the extrinsic parameters matrix. Also need to be made the projector calibration, so it can be used as an inverse camera. It will also defined by the stereo calibration process the fundamental matrix that encapsulate the Epipolar Geometry. With the Epipolar Geometry it is possible to map the pixels two distinct views from the same scene. So, it is possible to emit a structured light pattern over the scene and with the camera perceiving the distortion of this pattern over the 3D model, so it will be able to reconstruct it. The last steps consist to correlate the pixels from the two views based in the Euclidian distance from the colors of the two pixels and generating a disparity image to analyses the correlation process. Some results can be observed in the retroprojection errors in calibration process that are the same from library as so from the software used as reference, and in the disparity image generated was observed that can be an error in the epipolar line definition process or the correlation step, because the scene terrain cannot be observed in disparity image. |
| |
|
