Research On Distributed Aperture Infrared Super-Resolution Image

With the development of non-contact infrared detection application,image plays a more and more important role in application.The application scenes of infrared image are gradually increasing,which has attracted more and more attention.Compared with the traditional visible light image,the infrared thermal imaging obtained by passive infrared technology will not be affected by various harsh environments and extreme phenomena.Imaging has strong antiinterference and is suitable for application in some special scenes.Although infrared thermal imaging has many advantages in some aspects,the wavelength of infrared light is too long.When collecting the light wave frequency,the infrared sensor array will not reach twice the image frequency,and the image information will be aliased due to under sampling,resulting in blurred infrared thermal imaging or low spatial resolution.Image blur is fatal to some high-precision and inaccessible problems,which may lead to serious consequences because some details cannot be observed.Aiming at the demand of infrared imaging system for high resolution in some occasions,this paper carries out systematic research work around the equipment assembly mode of distributed aperture,deeply studies the infrared image imaging principle,the algorithm principle and classification of image registration,the principle of image reconstruction algorithm and the selection and analysis of super-resolution reconstruction model,and formulates a complete process of image super-resolution reconstruction.Firstly,it affirms the necessity of various algorithms needed in the experiment,namely image registration algorithm,image reconstruction algorithm and super-resolution reconstruction network model,and briefly summarizes and analyzes various algorithms contained in these algorithms,so as to pave the way for the later experiments.Then the experiment is carried out.Firstly,the imaging system composed of four cameras is built,the camera is controlled and image data is collected,and the collected raw file is converted into image data format for subsequent processing.According to the principle of Fourier Merlin change algorithm,a program is written to register the frames collected by four cameras,and the registration results are analyzed.POCS algorithm is selected to integrate the information of image sequence to reconstruct and restore a high-resolution image,and the reconstruction effect is viewed and analyzed.Finally,after comparison,realsr is selected as the super-resolution reconstruction model.Firstly,the network mode is used in the training set for training,and the training results are detected through the image block of the test set,and then the high-resolution image block reconstructed from the image is used as the image input,so as to obtain the output super-resolution image block,and take the graphic block in the original image and superresolution image to realize the visual comparison display.Finally,the evaluation index method is studied,and the method in the non reference image super-resolution evaluation index is selected as the analysis method of image data visualization.The super-resolution image obtained by this study and the super-resolution image calculated by the traditional super-resolution image algorithm are comprehensively calculated through the data analysis index,and compared with the image analysis index obtained by the traditional method.The data show that this research technology is higher than the conventional super-resolution reconstruction method,which provides a new idea for the super-resolution improvement of infrared image.