proj_match_points_ransacT_proj_match_points_ransacProjMatchPointsRansacProjMatchPointsRansacproj_match_points_ransac (算子)

名称

proj_match_points_ransacT_proj_match_points_ransacProjMatchPointsRansacProjMatchPointsRansacproj_match_points_ransac — 通过寻找点之间的对应关系来计算两幅图像之间的投影变换矩阵。

签名

描述

Given a set of coordinates of characteristic points (Cols1Cols1Cols1Cols1cols1cols_1,Rows1Rows1Rows1Rows1rows1rows_1) and (Cols2Cols2Cols2Cols2cols2cols_2,Rows2Rows2Rows2Rows2rows2rows_2) in both input images Image1Image1Image1Image1image1image_1 and Image2Image2Image2Image2image2image_2, proj_match_points_ransacproj_match_points_ransacProjMatchPointsRansacProjMatchPointsRansacProjMatchPointsRansacproj_match_points_ransac automatically determines corresponding points and the homogeneous projective transformation matrix HomMat2DHomMat2DHomMat2DHomMat2DhomMat2Dhom_mat_2d that best transforms the corresponding points from the different images into each other. The characteristic points can, for example, be extracted with points_foerstnerpoints_foerstnerPointsFoerstnerPointsFoerstnerPointsFoerstnerpoints_foerstner or points_harrispoints_harrisPointsHarrisPointsHarrisPointsHarrispoints_harris。

The transformation is determined in two steps: First, gray value correlations of mask windows around the input points in the first and the second image are determined and an initial matching between them is generated using the similarity of the windows in both images.

The size of the mask windows is MaskSizeMaskSizeMaskSizeMaskSizemaskSizemask_size x MaskSizeMaskSizeMaskSizeMaskSizemaskSizemask_size. Three metrics for the correlation can be selected. If GrayMatchMethodGrayMatchMethodGrayMatchMethodGrayMatchMethodgrayMatchMethodgray_match_method has the value 'ssd'"ssd""ssd""ssd""ssd""ssd", the sum of the squared gray value differences is used, 'sad'"sad""sad""sad""sad""sad" means the sum of absolute differences, and 'ncc'"ncc""ncc""ncc""ncc""ncc" is the normalized cross correlation. For details please refer to binocular_disparitybinocular_disparityBinocularDisparityBinocularDisparityBinocularDisparitybinocular_disparity。The metric is minimized ('ssd'"ssd""ssd""ssd""ssd""ssd", 'sad'"sad""sad""sad""sad""sad") or maximized ('ncc'"ncc""ncc""ncc""ncc""ncc") over all possible point pairs. A thus found matching is only accepted if the value of the metric is below the value of MatchThresholdMatchThresholdMatchThresholdMatchThresholdmatchThresholdmatch_threshold ('ssd'"ssd""ssd""ssd""ssd""ssd", 'sad'"sad""sad""sad""sad""sad") or above that value ('ncc'"ncc""ncc""ncc""ncc""ncc").

To increase the algorithm's performance, the search area for the matching operations can be limited. Only points within a window of points are considered. The offset of the center of the search window in the second image with respect to the position of the current point in the first image is given by RowMoveRowMoveRowMoveRowMoverowMoverow_move and ColMoveColMoveColMoveColMovecolMovecol_move。

If the transformation contains a rotation, i.e., if the first image is rotated with respect to the second image, the parameter RotationRotationRotationRotationrotationrotation may contain an estimate for the rotation angle or an angle interval in radians. A good guess will increase the quality of the gray value matching. If the actual rotation differs too much from the specified estimate the matching will typically fail. The larger the given interval, the slower the operator is since the entire algorithm is run for all relevant angles within the interval.

Once the initial matching is complete, a randomized search algorithm (RANSAC) is used to determine the transformation matrix HomMat2DHomMat2DHomMat2DHomMat2DhomMat2Dhom_mat_2d. It tries to find the matrix that is consistent with a maximum number of correspondences. For a point to be accepted, its distance from the coordinates predicted by the transformation must not exceed the threshold DistanceThresholdDistanceThresholdDistanceThresholdDistanceThresholddistanceThresholddistance_threshold。

Once a choice has been made, the matrix is further optimized using all consistent points. For this optimization, the EstimationMethodEstimationMethodEstimationMethodEstimationMethodestimationMethodestimation_method can be chosen to either be the slow but mathematically optimal 'gold_standard'"gold_standard""gold_standard""gold_standard""gold_standard""gold_standard" method or the faster 'normalized_dlt'"normalized_dlt""normalized_dlt""normalized_dlt""normalized_dlt""normalized_dlt". Here, the algorithms of vector_to_proj_hom_mat2dvector_to_proj_hom_mat2dVectorToProjHomMat2dVectorToProjHomMat2dVectorToProjHomMat2dvector_to_proj_hom_mat2d are used.

Point pairs that still violate the consistency condition for the final transformation are dropped, the matched points are returned as control values. Points1Points1Points1Points1points1points_1 contains the indices of the matched input points from the first image, Points2Points2Points2Points2points2points_2 contains the indices of the corresponding points in the second image.

The parameter RandSeedRandSeedRandSeedRandSeedrandSeedrand_seed can be used to control the randomized nature of the RANSAC algorithm, and hence to obtain reproducible results. If RandSeedRandSeedRandSeedRandSeedrandSeedrand_seed is set to a positive number, the operator yields the same result on every call with the same parameters because the internally used random number generator is initialized with the seed value. If RandSeedRandSeedRandSeedRandSeedrandSeedrand_seed = 0, the random number generator is initialized with the current time. Hence, the results may not be reproducible in this case.

执行信息

• 多线程类型：可重入（与非独占算子并行运行）。
• 多线程作用域：全局（可从任何线程调用）。
• 未采用并行化处理。

参数

可能的前趋

points_foerstnerpoints_foerstnerPointsFoerstnerPointsFoerstnerPointsFoerstnerpoints_foerstner, points_harrispoints_harrisPointsHarrisPointsHarrisPointsHarrispoints_harris

可能的后继

projective_trans_imageprojective_trans_imageProjectiveTransImageProjectiveTransImageProjectiveTransImageprojective_trans_image, projective_trans_image_sizeprojective_trans_image_sizeProjectiveTransImageSizeProjectiveTransImageSizeProjectiveTransImageSizeprojective_trans_image_size, projective_trans_regionprojective_trans_regionProjectiveTransRegionProjectiveTransRegionProjectiveTransRegionprojective_trans_region, projective_trans_contour_xldprojective_trans_contour_xldProjectiveTransContourXldProjectiveTransContourXldProjectiveTransContourXldprojective_trans_contour_xld, projective_trans_point_2dprojective_trans_point_2dProjectiveTransPoint2dProjectiveTransPoint2dProjectiveTransPoint2dprojective_trans_point_2d, projective_trans_pixelprojective_trans_pixelProjectiveTransPixelProjectiveTransPixelProjectiveTransPixelprojective_trans_pixel

替代

hom_vector_to_proj_hom_mat2dhom_vector_to_proj_hom_mat2dHomVectorToProjHomMat2dHomVectorToProjHomMat2dHomVectorToProjHomMat2dhom_vector_to_proj_hom_mat2d, vector_to_proj_hom_mat2dvector_to_proj_hom_mat2dVectorToProjHomMat2dVectorToProjHomMat2dVectorToProjHomMat2dvector_to_proj_hom_mat2d

另见

proj_match_points_ransac_guidedproj_match_points_ransac_guidedProjMatchPointsRansacGuidedProjMatchPointsRansacGuidedProjMatchPointsRansacGuidedproj_match_points_ransac_guided

参考文献

Richard Hartley, Andrew Zisserman: “Multiple View Geometry in Computer Vision”; Cambridge University Press, Cambridge; 2000.
Olivier Faugeras, Quang-Tuan Luong: “The Geometry of Multiple Images: The Laws That Govern the Formation of Multiple Images of a Scene and Some of Their Applications”; MIT Press, Cambridge, MA; 2001.

模块

匹配

---