We propose an automatic camera calibration algorithm using generated model of a scene by user. In our approach, for each frame, the algorithm detects feature points in image and the model using SIFT or any other feature detectors. Matching can be performed using ANN search method, SIFT matching algorithm or any other matching algorithm. Non-geometrically matched features points are suppressed by using "histogram based suppression", "RANSAC/LMedS iteration and fundamental matrix calculation" and "iterative refinement". The detected feature points are used for camera calibration. Experimental results show that the proposed method can robustly detect geometrically matched feature points in the case of occlusion and bad lightening condition; therefore it can be used for precise moving camera calibration.
展开▼