Research On Real Time Image Matching Metdod For A New Image Interpretation And Draughting Application

For present, there are three main methods used in the work of Interpreting and Draughtingobjects in Remote Sensing imagery. This work is an important procedure in photogrammetryproduction work flow. However, each method of the three has its own shortcomings. With thispaper we want to propose a new method, which is supposed to be more efficient than the threetraditional methods from some aspects. The key point for new method is nothing but anapplicable matching strategy. With working on image matching strategies in photogrammetryand computer version domain, some improvements in theory and arithmetic were achieved.1. A two stage matching strategy is designed. The first stage called coarse-matchingstage aims to get dense supportive correspondences by virtue of which the secondstage can be more fast as well as robust in correspond procedures.2. In coarse-matching stage, a divisional iterative matching algorithm based oninvariant feature is proposed. Experiments proved that reliability and density ofcorrespondences is promoted.3. A neighboring quasi-epipolarline constrain is proposed in the study of automaticrelative orientation. A multi-constrain strategy is used to eliminate mismatches, andthus reliability of correspondences is promoted.4. To define the most reasonable searching area, we designed an experiment-proveneffective parameters updating method which differentiate three point types.5. On the basis of intensive analysis about past Color Geodesic Distance (CGD)algorithm, we proposed a more fast one which is kernel of a new adaptive blockmatching algorithm called adaptive Segmentation NCC algorithm based onCGD(SCGD). SCGD is experimentally proved to be more robust on large scaleimagery. While on middle or small scale case, traditional NCC can NOT performbetter either.6. Subpixel procedure based on quadratic curve fitting is performed for correctcorrespondences7. With analysis on relative methods for multi-frame matching, we designed andimplemented a method for the integrated application of both two-frame andmulti-frame matching.8. Designed and implemented an applicable image matching solution for DigitalImagery Interpreting System(DIIS).