Research On Deblurring Of Multi-source Remote Sensing Image Based On Double Deep Convolutional Neural Network

With the continuous development of remote sensing technology and the research and development of various high-resolution remote sensing platforms,remote sensing technology has gradually played an important role in the fields of science and technology,social economy,and national defense construction.However,due to the influence of unavoidable factors such as platform shake and atmospheric turbulence,remote sensing images will be blurred,the image quality will be reduced,and some important information will be lost.The images cannot accurately reflect the actual situation of the target,which will lead to subsequent applications such as environmental monitoring and disasters.Detection etc.are affected.Therefore,the application value of remote sensing image deblurring is also increasing.At present,in the field of deblurring,most scholars' research directions are camera motion blur in daily life,a single image deblurring method,and there are few related researches on remote sensing image deblurring.Because the single image deblurring method lacks the necessary prior knowledge,the deblurring effect is not satisfactory,the real-time performance and stability are poor,and it takes up a lot of time and computing resources.In order to further enhance the applicability and stability of the deblurring algorithm,this article first deeply studies the principles and framework of the Maximum A Posteriori-based image deblurring method and the convolutional neural network-based deblurring method,and compares the experiments to summarize the advantages and disadvantages of the algorithms..In order to avoid the shortcomings of the above algorithms,this thesis designs a new multi-source remote sensing image deblurring algorithms based on double-deep convolutional neural network.According to the characteristics of multi-source remote sensing images,a dual convolutional neural network is designed to better obtain external priors,thereby improving image restoration capabilities;this paper proposes a new activation function,which has stronger image extraction Ability to solve the problem that the Re LU activation function treats the details of the image as a fuzzy part "by mistake",and has a better ability to remove noise in the experiment.By comparing with other deblurring methods,the results prove that the method proposed in this paper has better image deblurring ability.