Generating synthetic imagery of complex scenes from ideal synthetic source imagery via MSERs on entropy imagery

摘要

Synthetic imagery generation is not a new topic; however, it has reemerged as a major focus in recent years. Thisis in part due to the success achieved by modern machine learning methodologies, in particular, deep learning.One reason these technologies have succeeded is due to the wealth of available training data. A majority ofthe available data are of generic objects or scenes. However, there are numerous applications in which dataare neither readily available nor easily obtained in large quantities. In such scenarios, synthetic imagery is anappealing choice to address this shortcoming. While still faster than the performance of data collections, physics-based models tend to have computational complexity and require extensive computational time. This work seeksto investigate the use of reduced-order modeling (ROM) of relevant objects identi ed by a maximally stableextremal region (MSER) detector from the entropy image of simple ideal high- delity, physics-based syntheticimages. Speci cally, this work will utilize MSERs to identify pertinent objects to be placed within the simplescene via ROM to produce a more complex scene. This approach has the bene t of rapidly increasing both thecomplexity of simple, ideal, high- delity, physics-based scenes and the amount of synthetic imagery generatedvia random or statistically-based placement of the objects throughout the scene.