Infrared unmanned aerial vehicle group target simulation method under complex background based on image derivation

Abstract

The invention belongs to the field of infrared image simulation, and discloses an infrared unmanned aerial vehicle group target simulation method under a complex background based on image derivation. The method comprises the following steps: (S1) preprocessing an infrared video to obtain an original unmanned aerial vehicle template data set and a background template data set; (S2) making a fuzzy-restored data set; (S3) performing supervised pre-training and unsupervised fine tuning on the evolution-degradation adversarial network model; (S4) carrying out unmanned aerial vehicle template extraction, background superposition and image restoration by using the adversarial network model; (S5) jointly solving a Gaussian-Poisson equation to obtain a simulation image; and (S6) multiple original unmanned aerial vehicle templates being mixed for multiple times to obtain an infrared unmanned aerial vehicle group simulation image. According to the method, background information in the infrared unmanned aerial vehicle image can be eliminated, the unmanned aerial vehicle template with high definition and signal-to-noise ratio is extracted to be mixed with the background template, dark artifacts of the synthetic image are removed, and the synthetic image with high harmonicity and visual authenticity is obtained.

Description

technical field [0001] The invention belongs to the field of infrared image simulation and image editing, relates to infrared image editing and artificial neural network training, and realizes an infrared unmanned aerial vehicle group target simulation method under complex background based on image derivation. Background technique [0002] The technology of modern small drones is becoming more and more mature. Because of its ability to mount dangerous items such as guns and explosives, it poses a serious threat to social public order and national defense security. Infrared video images containing UAVs and UAV groups in different backgrounds are the basis for the research and verification of infrared UAV target recognition and tracking algorithms, but are limited by the high complexity and high consumption of manpower and material resources of UAV collaborative technology. , so that more researchers choose to use simulation methods instead of measured data shooting. In order...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problems of difficult separation of target templates, low harmony of synthetic images, and inconsistent characteristics in the infrared UAV swarm target simulation method based on image mixing, the original template image is clarified by using the evolutionary generator of the dual generative adversarial network evolution-degradation network ; Then extract the clear target template and superimpose it with the background image, and then use the degradation generator of the evolution-degeneration network to restore the superimposed image to obtain the restored image of the composite image; finally, the above output result is used as a grayscale constraint item, The gradient feature of the original UAV template image is used as the gradient constraint item, and the Gauss-Poisson equation is jointly optimized to obtain a closed-form solution, and the target-background mixed image is obtained, and the above process is repeated to realize the target image of the UAV group in a complex background simulation of

Images