Image Fusion Technology In Ultraviolet Imager

Discharge will occur when electrical equipment is affected by environmental,installation and other factors during operation.Long-term discharge will increase the occurrence of faults,and threaten the safe and stable operation of electricity and railways.As an excellent instrument for detecting weak corona discharge,the ultraviolet imager has been gradually applied to power inspection in recent years.The ultraviolet imager can respectively imaging the ultraviolet light signal generated by the discharge and the visible light signal of the surrounding background,and fuse the two images to achieve the purpose of locating the discharge position.The related technologies of image fusion and the real-time video fusion of ultraviolet and visible images of UV imager are mainly conceded in this paper.In this paper,the registration of UV and visible images is studied from the feature-based and mutual information based.This paper analyzes the feature-based image registration process,and extracts and matches the UV and visible images by SIFT,SIFT+RANSCA and SURF.In addition,this paper proposes an ultraviolet-visible image registration algorithm based on mutual information,and introduces the principle of mutual information-based algorithm from the registration process.And studies the similarity measure based on mutual information,uses the 1+1 evolutionary algorithm to find the optimal similarity measure,finally uses the rigid body transformation parameters to registrating the UV image.The simulation results of several registration algorithms based on features and mutual information-based registration algorithms are carried out on UV and visible images.The results show that the registration algorithm based on mutual information has a good result in UV image registration.After strict registration,the basic conditions for image fusion have been established.In this paper,an ultraviolet and visible image fusion algorithm based on Non-subsampled Shearlet Transform(NSST)and adaptive sparse representation is proposed.By performing NSST transformation on two images,different high frequency coefficients and low frequency coefficients use different fusion rules respectively.In this paper,the standard deviation of the low-frequency sub-band images of ultraviolet and visible light is studied,and the fusion rule of large deviation is adopted.The training of adaptive sparse dictionary and the fusion rule of high frequency are analyzed for high-frequency sub-bands.Compared with the fusion algorithms mentioned in other literatures,the proposed algorithm has double superiority in subjective visual and objective evaluation.However,the above algorithm is inferior in real-time.In order to meet the real-time video image fusion in the UV imager,according to the characteristics of the UV and visible images,an image fusion method based on pixel comparison and weighting is also studied,which provides the direction for next real-time fusion.Finally,according to the needs of the project,the hardware processing system was selected as the HI3531 D of the Hi Silicon platform,and the application of image registration in the ultraviolet imager was completed,realizing the real-time video image fusion based on the Hi Silicon platform.The results show that the image registration and image fusion technology of the UV imager can effectively obtain the registration parameters,and the fusion effect is better.The test results under the hardware platform met the design requirements.