The equivalently construction of the simulation scenes based on the measured data from typical environmental scenarios yields greater accuracy and realism than traditional radiosity-based simulation model construction. Concurrently, the feature library generated in the process of equivalent construction process can further support the verification of simulation models. Nonetheless, the persistent challenge of decoupling targets from backgrounds during scene construction has been a pivotal obstacle impeding fidelity enhancement. To this end, a novel background decoupling technique that synergizes the enhanced YOLOv8 with the Criminisi algorithm is introduced in this paper. The process begins with the precise extraction of the target mask utilizing YOLOv8. Subsequently, the refined Criminisi algorithm is employed to reconstruct the background while preserving structural integrity. Ultimately, the Poisson blending algorithm is applied to amalgamate the target and background, thereby augmenting the simulation scene's realism. Experimental results demonstrate that the simulation environment crafted via this methodology exhibits a higher resemblance to actual captured images, effectively resolving issues related to the singularity of scene targets and the diminished realism in the real-time rendering of simulation scenes.