The Study of Fast and Precise Positioning Techniques for Laser Speckle Patterns
| Autor: | Ping-Yen Chen, 陳秉彥 |
|---|---|
| Rok vydání: | 2015 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 103 Laser speckle patterns provide high resolution and invariant properties and this can offer precise positioning information. However, as the positioning resolution become higher, the computation complexity is also increased. Furthermore, the will slow down the positioning speed. In this thesis, GPU and SURF methods are used to accomplish fast and precise positioning of laser speckle images. Speckle patterns are invariant to rotation and translation generated from a laser device. Therefore, it is suitable for precise positioning. For image matching, SURF is used to calculate image feature points and the image offset. Furthermore, CUDA architecture is used to achieve the fast matching. The matching procedures are uploading images to GPU, offset calculation, and then downloading images to CPU. Experimental results have shown that GPU can achieve better performance than CPU in calculation speed. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|