Simulation And Recognition Of Multi-View Warship Infrared Images

Infrared imaging technology is widely used in military and civilian fields,and has been paid attention by various countries for a long time.In the military field,due to the limitation of many special conditions,we cannot directly obtain a large number of infrared image data.The development of computer technology makes it possible for people to simulate infrared image by computer.Specially in the task of warship detection and recognition in this paper,there is a lack of large-scale real-world infrared images.Therefore,we obtain infrared images of warships through infrared image simulation to provide data support for subsequent detection and recognition.In this paper,we firstly use the SE-Workbench infrared image simulation software to simulate the multi-view infrared image of the warship.First of all,we establish the threedimensional model of the warship by Multigen Creator software,the temperature field model of the warship surface and the atmospheric thermal model of the scene by SEWorkbench infrared image simulation software.Then,we carry out the multi-view infrared images of the warship based on the established model,and finally develop a large-scale multi-view infrared image database of the warship.However,the traditional infrared image simulation method relies on three-dimensional model of target scene,temperature field model of target surface and OpenGL graphics rendering,thus,resulting in a high-cost,complex and time-consuming process.In order to improve the efficiency of simulation,we further study the method of infrared image simulation based on generative adversarial networks.Using the conditional generative adversarial network trained on the paired visible and infrared image dataset,the visible images of the warship can be simulated to generate the corresponding infrared images.Furthermore,since paired images will not be available,this paper proposes an infrared image simulation method based on cycle generative adversarial network,and extends it to the infrared image simulation of complex ground scenes.This method utilizes a large number of unpaired natural scene visible and infrared images,and realizes the conversion of visible image from complex ground scene to corresponding infrared medium wave image.Based on the establishment of multi-view warship infrared image database,this paper carries out the research of multi-view image cooperative detection and recognition method.In the aspect of detection,we use collaborative segmentation to model the conformal complementary information of multi-view images to detect warship in multi-view image sequence.On the basis of detection results,we exploit joint sparse representation to mine the correlation information between views,which significantly improves the recognition rate of multi-view warships.Finally,this paper develops the software that integrates the cooperative detection and recognition of multi-view warship.