Panorama Stitching in Planetary Environments

While panorama stitching is considered largely solved, the background of planetary environments adds some challenging aspects. One facet is the need of scientifically exploitable panoramas. They provide visual and geometric context for measurement data of other sensors, which makes them the main sense on planetary missions. Some images are taken in different spectral ranges to determine the composition of rocks. These images may also need to be embedded into a wider context for a better understanding. The process of panorama stitching has to be traceable and documentable like every process in science. Any blending or unconstrained warping could corrupt the result and lead to avoidable uncertainties in interpretation. This scientific integrity of panoramas is paramount for reliable further analysis by geologists. Another aspect of planetary environments on the other hand are the large data sets of currently up to 1000 images per panorama. In order to reduce the amount of unneeded data which is transmitted, the cameras typically have a very narrow field of view. This allows a high granularity of the mapped area however it increases the required capabilities of the stitching framework. In this thesis we will present
 a flexible and scalable pipeline for panorama stitching in planetary environments. The modular design allows individual stages to be modified  or completely exchanged without consequent changes on other modules. As the main objective is the flexibility and long-term usability, also for other purposes in the industrial domain, some modules only receive basic functionality. However for the stages directly involved in the stitching process we evaluate state of the art solutions for implementation and extended them.