Impacts of different transformation models on remote sensing image registration accuracy based on implicit similarity
| Autor: | Popo Gui, Qin Ye, Cuifang Ai, Yahui Yao |
|---|---|
| Rok vydání: | 2015 |
| Předmět: |
Synthetic aperture radar
Similarity (geometry) Polynomial transformation business.industry Image registration Terrain Computer Science::Graphics Transformation (function) Geography Computer Science::Computer Vision and Pattern Recognition Computer vision Artificial intelligence Affine transformation business Image resolution Remote sensing |
| Zdroj: | SPIE Proceedings. |
| ISSN: | 0277-786X |
| DOI: | 10.1117/12.2204788 |
| Popis: | How the different transformation models take effects on the registration accuracy based-on implicit similarity between the remote sensing images is the key point of this paper. For registration between SAR and optical imagery, analyze the imaging characteristic of push-broom optical satellite image and SAR image according to their imaging models; study the impacts taken by terrain fluctuation and different transformation models. The DEM and image pairs are simulated in the experiment, the results show: in region of bigger relief, the larger the registration image size, the greater impacts are taken by different transformation models on registration accuracy. Considering the polynomial transformation model leads to the low searching efficiency, affine transformation model regards as the best model for registration, but it has low accuracy and just applies to small images(such as 200x200). For large image (such as 800x800), 8-parameters transformation model is the best choice (balance accuracy and efficiency), but adding the parameters of transformation model (such as 12-parameters) again cannot significantly improve the registration accuracy. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|