TY - JOUR T1 - Determining regularization parameter in high resolution images georefrencing with Rational Functions TT - تعیین پارامتر پایدارسازی در زمین‌مرجع کردن تصاویر قدرت تفکیک بالا بوسیله توابع گویا JF - kntu-jgit JO - kntu-jgit VL - 4 IS - 2 UR - http://jgit.kntu.ac.ir/article-1-140-en.html Y1 - 2016 SP - 47 EP - 64 KW - High-Resolution Satellite Images KW - Rational Function KW - Ill-ondition Problem KW - regularization KW - L-curve. N2 - High-resolution satellite images are extensively used in different fields. Geo-referencing process, as innate part of extraction of topographic terrains through these images, has been studied in many researches. In geo-referencing of satellite images, different models can be used, but rational functions are the most suitable options. Determining co-efficiency of rational functions is ill-condition problem, so to solve this problem Tikhonov regularization method has been used. In such regularization method, regularization selection parameter is very important. In present study, this parameter was calculated through two methods including: minimizing root mean square of errors (RMSE) and the L-curve for determining co-efficiency of rational functions. Then these two methods have been used in least standard squares of parametric model. Also combined model has been used to determine co-efficiency of rational functions in geo-referencing process. These calculations have been done for two different control-points groups with various numbers and accuracies. Using these two models (parametric and combined), regularization parameter has been calculated through L-curve and root mean square of error methods by 55 points. The results show that the root mean square errors and L-curve methods in parametric model led to accuracy of 4.45 and 5.40 pixels, respectively. Also in the combined model, root mean square errors and L-curve methods showed accuracy of 3.42 and 5.10 pixels, respectively. Above calculations were repeated with 120 points. This time, results show approximately same accuracies for both root mean square errors and L-curve methods. M3 10.29252/jgit.4.2.47 ER -