When it comes to more autonomous unmanned aerial vehi- cles, enhanced sensory and perceptive capabilities need to be integrated and qualified for mission scenarios of larger scale. In this context, recent developments in embedded technologies now allow the use of onboard image processing on such airborne platforms. However, the acquisition of mission relevant imagery and video test data necessary to develop and verify such processing algorithms can be complicated and costly. There- fore we are interested in the usability of commercial-of-the-shelf virtual simulation environments for generation of test and training data. To yield general acceptance, the relevance and comparability to real world imagery needs to be investigated. We pursue a multi-level approach to analyze differences between real and coherently simulated imagery and measure respective influence on image processing algorithm performance, taking into account typical visual database and rendering benchmarks such as level of detail, texture composition and rendering details. More specifically, in this paper we analyze corresponding real and synthetic footage using image descriptors from the content based image retrieval domain introduced in the MPEG7 standard. This allows us to compare the appearance of images in regard to specific image properties without disregarding their overall content. In future work it is planned to apply and evaluate the test subject, a computer vision algorithm on real and synthetic imagery. These evaluations are compared to allow detection of specific image properties influencing the performance of the test sub- ject and therefore will help in identifying differences in the synthetically generated image. The results will provide insight on how to specifically trim image generation methods to reach equal processing performance with both image sets, mandatory to justify usage of synthetic footage for algorithm development and qualification    «

When it comes to more autonomous unmanned aerial vehi- cles, enhanced sensory and perceptive capabilities need to be integrated and qualified for mission scenarios of larger scale. In this context, recent developments in embedded technologies now allow the use of onboard image processing on such airborne platforms. However, the acquisition of mission relevant imagery and video test data necessary to develop and verify such processing algorithms can be complicated and costly. There- fore we are i...    »