Research On MAP Super Resolution Method For Aerial Remote Sensing

Aerial remote sensing has attracted wide attention because of its flexibility and wide range of applications.In the field of military struggle,airborne photoelectric platform can realize remote sensing of long-range reconnaissance of reconnaissance targets.It can not only detect and strike the enemy alone,but also cooperate with satellite and radar for further reconnaissance.In the field of civil production,whether it is water pollution,forest and grassland fires or pests,UAV aerial photography,aerial remote sensing has played an increasingly important role.However,due to the limitation of load quality,the imaging system often can not use the equipment with good imaging effect.We hope to improve the spatial resolution of the image through image super-resolution technology.For this reason,we propose a set registration and super-resolution super-resolution reconstruction system for aerial remote sensing.The main contents are as follows:1.The degradation factors of aerial remote sensing are proposed and analyzed.The degradation models of imaging system and aerial remote sensing are introduced.The main causes of image degradation and degradation are analyzed,and the mathematical expression of degradation model is analyzed.It mainly includes optical imaging point spread function,motion blurring and inclination of aerial remote sensing.Analysis of degradation model of aerial imaging system can analyze image degradation,which is conducive to image registration and subsequent super-resolution reconstruction.2.Aiming at the sub-pixel registration for super-resolution reconstruction,the optimization method of Taylor function is used to perform sub-pixel registration.Three sub-pixel registration methods,interpolation amplification method,frequency domain registration method and Taylor function-based optimization method are analyzed.Optimal optimization method realizes sub-pixel registration by Taylor function secondorder expansion.The computational complexity is reasonable and the principle of the algorithm is simple and clear.It is very suitable for image registration.3.A maximum posteriori probability super-resolution reconstruction model based on Markov random field theory is proposed.Markov random chain theory can well characterize the relationship between one-dimensional time series signals.We expand the use of Markov random chains and transform one-dimensional Markov random chains into two-dimensional Markov random fields,which can well characterize the local features of two-dimensional images.The relationship between low-resolution image and high-resolution image can be well established by using Markov random field.4.In order to accelerate the iteration speed of super-resolution reconstruction based on maximum a posteriori probability.We introduce the time domain correlation term in the iteration process.The information source of the image can be obtained from the front and back frames,and the structure of the image itself can be obtained.From the physical scene of super-resolution iteration of image MAP,we can find that the iteration of a picture is a continuous approach process.We can use computational mathematics knowledge to add an iteration term related to the number of iterations,which can accelerate convergence and save computational costs.