Research On Positioning And Block Adjustment Technology Of Airborne Three Linear Array CCD Images And Corresponding Acceleration Strategies

Airborne three line CCD camera can acquire high spatial resolution and high overlap ratio panchromatic and multispectral images in a fast and convenient approach.Thus it has a wide application in surveying and mapping filed.Image matching and area block adjustment are the key parts and indispensable steps in the following airborne three-line array CCD image processing.Accurate,reliable and effective image matching can provide a large number of joint points for bundle block adjustment.High precise and efficient block adjustment technology can effectively improve the ground positioning accuracy of airborne three-line CCD image,thereby ensuring the ground accuracy of subsequent DOM image products.In this thesis,the airborne three-line CCD camera is used as research object.In comparison with the typical airborne three-line ADS40 image,the study focus is put on image matching and area block adjustment research for the domestic developed airborne three-line array camera.In the thesis,the main research contents include strict geometric model construction,direct geo-positioning,accelerated matching technique and fast bundle block adjustment based on sparse matrix.The main work and innovations are done there:1.Firstly,an in-depth analysis is given on the current status,developing trend,and existing problems of matching and block adjustment technologies for airborne three-line CCD images.For domestic airborne three-line stereo surveying and mapping camera,there are some problems urgently to be solved.In combination with these problems,the main research contents of the thesis are determined.2.Based on the analysis for the imaging principle and characteristics analysis of airborne three-line CCD camera,an introduction is given on the various coordinate systems involved in the geometric model constructions.Then the conversion models from the POS navigation angles to the exterior orientation elements are deduced under two different angle rotation systems.3.Upon the above research fundamentals,strict geometric imaging models are built for typical ADS40 image and the new domestic airborne three-line array camera.And linearization froms are derived for strict geometric model under two different angle rotation systems.Moreover,the multi-photo space intersection approach is studied.The correctness and precision of the models are verified by direct positioning test,which lays the foundation for the subsequent block adjustment processing.4.For the domestic airborne three-line CCD images,the data amount is huge.And the uneliminated error matching points is especially serious for subsequent block adjustment.To solve these problems,the thesis put forward an improved GPU-accelerated SIFT matching method in combination with the epipolar constraint and RANSAC algorithm.Experiments are carried out on ADS40 and domestic three-line CCD images,which verifies the efficiency and reliability properties of proposed matching method.5.An analysis is given on the principles of three commonly used block adjustment models for airborne three-line CCD images,and a detail introduction is given on the fundamentals of sparse matrix calculation technique.Then the sparse matrix technique is applied to the normal bundle block adjustment to accelerate the computation speed.Experimental results verify that in comparison with the traditional block adjustment algorithm the improved algorithm has 2 to 4 times acceleration ratio effect and keep high precision.