Research On Key Technologies Of Infrared Guidance For Suppressing Cloud And Window Aero-Optical Effect

The infrared imaging terminal guidance(IITG)with typical characters including excellent concealment,fast imaging,relatively high imaging resolution and anti-interference efficiency,will demonstrate a valuable application in high-speed moving target detection and recognition fighting in atmosphere.However,the IITG will be disturbed by window aero-optical effect and clouds,so as to greatly hinder its efficient processes in target detection.As demonstrated,the traditional methods mainly consider the thin cloud situation,and thus the researches on the thickness and the type of the cloud layer,which will greatly influence the efficiency of the IITG,is insufficient.The key technologies for suppressing the cloud and window aero-optical effect,which will remarkably disturb the target detection and recognition in the process of conducting IITG under the high-speed moving conditions,are researched in this thesis.The main contents are as follows:First of all,the research of cloud classification algorithm is carried out by using statistical analysis and machine learning combined with various cloud map data of meteorological satellites,so as to provide a basis and then guidance for improving the efficiency of the IITG.The average accuracy of the cloud classification algorithm for the IITG proposed in this thesis is already more than 80%.Secondly,different algorithms are used to correct the window aero-optical effect according to the target background.In the past,the correction algorithms in which the targets are only in the ground background.are considered mainly.If the targets are in the sky background,there still existing non-sufficient correction of the aero-optical effect.The image data used to test the correction algorithm were obtained using an aero-optical mathematical model,and the infrared spectral image inversion algorithm was also used to augment the data set.Then,the window aero-optical effect is corrected by using the conditional adversarial networks against the ground background according to comprehensive considering both the aero-transmission effect and the aero-heating effect and also ignoring the system noise.The mapping relationship between the picture s involving the window aero-optical effect and the clear pictures is learned and thus extracted through training a large amount of data employing the conditional adversarial networks for effectively correcting the patterns involving aero-optical effect.The typical value of the structural similarity between the original pictures involving the window aero-optical effect and the clear pictures is about 50%,and also that between the pictures involving the weak window aero-optical effect corrected by the adversarial networks and the clear pictures is about 70%.Finally,the window aero-heating effect is corrected by the high-heat non-uniform background suppression algorithm proposed against the sky background,and also the system noise is removed,and further the MAP-HU moment algorithm is used to correct the aero-transmission effect.Thereby,the aim for effectively improving the signal-to-noise ratio of the image and then retaining the weak target characters and thus improving the detection and recognition efficiency,is achieved.