Study On Modeling And Simulation Of Optical Remote Sensing System And Image Processing Technology

With the development of optical remote sensing imaging technology, spaceremote sensing systems become more complex, and are changing towards higherspatial resolution, higher spectral resolution, and higher radiometric resolution.However, the miniaturization and light-weight of remote sensors are conflicting withhigh-resolution imaging, which increases the difficulty of designing them. Due tothe limitations of observing conditions, it is impossible to acquire remote sensingdata beyond various conditions. Therefore, we can lower the cost to acquire variousdemanding images by means of laboratory simulation from sensors, satelliteplatforms, deep space environment, atmospheric transport, and surface scenes. Thus,optical remote sensing simulation technology plays an important role in remotesensing mission predictions, imaging system design, image quality assessment,image processing algorithm verifications, and image interpretation trainings. It hasalready become a hot research topic at home and abroad.After the modeling and simulation of optical remote sensing systems, variousimages are obtained. We need to need to use a variety of appropriate optical remotesensing image processing techniques to process them to extract useful information.This realizes the diversity of application of remote sensing images. In this paper,following the completion of the image processing techniques research:1) Based on empirical summarized MTF mathematical models, we designedremote sensing imaging chain simulation V1.0. The software can implementdynamic simulation of atmospheric, optical systems, platforms, CCD detector,defocus, image motion, and electronics system in a fully digital environment, andshow all affects of each system on the final remote sensing images quality. Thishelps to evaluate imaging performance and achieve optimal design of remotesensing sensors. 2) In the MTF algorithm based blade edge method, a novel iterativealgorithm for nonlinear least squares to Fermi function is proposed. It can eliminatethe shortcomings of simulated annealing algorithm to calculate the coefficientfunctions. Thus, this algorithm has greater improvement in terms of both accuracyand speed.3) Based on the analysis and comparison of various conventional colorinterpolation algorithms, a novel algorithm based on weighted edge-directed,weighted color-diference interpolation, and well-designed terminal condition foroptimal performance is proposed. Considering the color of the laws of the algorithm,the weighting algorithm, the adaptive algorithm and iterative optimization algorithm,this algorithm increases the correlations of R, G, B color components, and selectsthe appropriate interpolation direction to effectively suppress moiréfringe color.4) After pretreatment of astronomy star charts, star point coarse positioning,star point fine positioning, we proposed a new feature matching method using thecenter of gravity of congruent triangles feature points. Then, based on the focus ofthese triangles point set, we use RANSAC algorithm to calculate the registrationtransformation model to complete the registration process. The whole process isusing the GPU to design parallel programming for the star map registrationalgorithm, and using MEX file compiled by Matlab nvmex complier to effectivelyaccelerate the registration speed of astronomical star chart. Total time speedup hasreached29.043times.5) Based on home-made multi-band polarization CCD ground experimentaldevice, we acquire polarization spectral images under incandescent lamp. Thisexperiment is designed to tell the real flowers from artificial flowers. Compared withgrayscale images, it is easier to distinguish them using DOLP as evaluation index.