A general discussion on synthetic scene generation methodologies for EO simulation from an AI perspective

Synthetic scene simulation plays a critical role in the design and testing processes of Electro-Optical (EO) systems by significantly reducing the need for extensive field trials. These simulations enable the controlled variation of scene components, including background, atmosphere, objects, and system-level elements of the imaging chain. Various methodologies exist in the literature for simulating radiometric scenes, each incorporating different aspects of the EO image formation process. One prominent approach involves physically-based or radiometric scene generation, which models the entire imaging chain from object space to detector output. Alternatively, when real data is available, hybrid approaches—such as composite scene generation—can be employed to merge synthetic and real elements effectively. This paper provides a general discussion of these synthetic scene generation techniques and illustrates their application in an AI-based object detection framework.