| |
| Acadêmico(a): Lucas Eduardo Schlogl |
| Título: BIBLIOTECA PARA DETECÇÃO DE RELEVOS BASEADA EM LUZ ESTRUTURADA E PADRÃO DE PROJEÇÃO BINÁRIO |
| |
| Abstract: |
| This work presents the development of a relief reconstruction library based on structured light and binary patterns that emerges as an alternative to the work started by Storz (2017). By analyzing the distortions of the binary patterns projected onto the scene and performing the decoding of the captured images it is possible to successfully reconstruct the 3D scene. Compared with Storz (2017) the developed library changes the form of calibration for variables defined as Z value and white and black thresholds, eliminating the time-consuming process of calibrating with chessboard. In order to view the reconstructed scene, it’s possible to export the point clouds to XYZ and PLY files in addition to being able to get the disparity map, and with these point clouds it’s possible to import them in other programs such as Unity and Meshlab. Currently the library is developed in C ++ and uses OpenCV to perform operations on the images and all code runs only on CPU. The library portability was also performed with the goal of reaching as many operating systems as possible, since the Storz (2017) library works only on Windows. Analyzing the results, it was possible to notice that the implemented algorithm is approximately 310 times faster when compared to the Storz (2017) library, but the process of image capture is longer due to having to capture multiple images. The developed library can’t be used in real-time scenarios since the capture time of the images takes approximately 5 seconds. We also analyzed the impact of lighting on the reconstruction and it can be verified that the impact was minimal. |
| |
|
