Base On The Fundamental Matrix Study Of Matching Algorithm

Pattern Matching in computer vision in the resumption of the corresponding relationship between two images pixel, has been the focus of the study. Collected from different viewpoints of the two same object space image point correspondence between certain constraints, This is the epipolar geometric constraint. basic mathematical expression of matrix is this kind of geometric constraint. Fundamental matrix contains a different viewpoint are the important geometrical relation between image and described the corresponding points should meet the epipolar constraint conditions; contains two cameras or between a single motion parameters of the camera. An accurate estimate of the fundamental matrix is an important task in computer vision. This Article respectively on computer image processing of image matching involved in technology, Feature Point detection technology and solution of fundamental matrix of three conduct more in-depth studies, compared to all kinds of algorithms, said their advantages and disadvantages, and put forward an improved algorithm. On feature points extraction, as SIFT algorithm for calculating long time, and Harris algorithm and graph rotation noise does not fit, put forward a combination of advantages of the two Harris-SIFT algorithm, that is, the first establish a scale space, use Harris algorithm of Feature extraction and then to each feature point feature vector descriptor. Algorithm of time has been reduced on the one hand, and algorithm robustness has been improving. On the solution of the fundamental matrix, put forward a 5-point algorithm for solving the essential matrix, and then use the basic nature of matrices and matrix calibration relationship derived the fundamental matrix and on misunderstanding points in a 5-point algorithm for detection of Polar distance minimum principle, improve the accuracy rate of the 5-point algorithm.