Key Techniques Studying On High-resolution Remote Sensing Imaging Simulation

As more and more high-resolution satellites are launched successively,remote sensing observation has long entered the "high-resolution era".Using computer simulation technology to simulate the process of remote sensor observation,it can directly and accurately analyze and evaluate the internal and external factors and mechanisms that affect image quality in the imaging process,and provide basis and guidance for applications such as sensor design,remote sensing data processing,and image quality evaluation,which has a very important application value.The coverage area of a single pixel of high-resolution remote sensing image is small,and the geometric and radiation characteristics of the object are more highly recognized.Compared with the low-and medium-resolution,the factors that should be considered in the high-resolution imaging simulation are more complex.Therefore,this dissertation combines the computer simulation technology and the remote sensing imaging mechanism to study several key techniques involved in the high-resolution remote sensing imaging simulation.The specific research works are listed as fellow:The research on radiation characteristics model of high-resolution digital scene.A high-definition 3D digital scene is established based on a real scene.A method for inversion of characteristic parameters of BRDF model was proposed.The effect of atmospheric scattering on the inversion of BRDF characteristic parameters was deducted,and the intrinsic directional reflection characteristics of the object were obtained.A material texture mapping method was proposed to associate the intrinsic directional reflection characteristics of different materials with the 3D model,thereby establishing a high-precision 3D digital scene with radiation characteristics model,providing the basis for subsequent calculation at-sensor radiance under different illumination,atmosphere,and observation conditions.The calculation of at-sensor radiance of high-resolution remote sensing imaging.A fast calculation method of optical radiation parameters--such as atmospheric downward radiance,transmittance,path radiance,and optical thickness--based on neural network is proposed,takes the optical radiation parameters calculated by MODTRAN under different conditions as training samples,constructs and trains the neural network to quickly calculate optical radiation parameters,providing the necessary input parameters for parallel radiation transfer calculation.Large amounts of light are emitted by Monte Carlo method and the interaction between and the lights and three-dimensional scene and atmosphere are traced,simulating the multiple reflections between the surfaces and the spatial modulation of the atmosphere.Finally,the at-sensor radiance is calculates,and the high-resolution radiance image of ideal sensor is obtained.Based on the research results of the above-mentioned key techniques,a high-resolution remote sensing imaging simulation program was developed,and the simulation results were compared with the radiance images acquired by GF-2.The results showed good radiation and spatial consistency between the simulated radiance images and the real at-sensor radiation images of satellite.This article systematically studies several key techniques in high-resolution remote sensing imaging simulation.The research result of the dissertation is great significant for the development of remote sensing imaging simulation platform and the promotion of application process of simulation technology.