Tracks selection for robust, efficient and scalable large-scale structure from motion

A Generic Bundle Adjustment Methodology for Indirect RPC Model Refinement of Satellite Imagery

“…Our pipeline employs the feature tracks selection method from [12], summarized in Algorithm 4, to select an optimal subset of feature tracks (or equivalently tie points) for the bundle adjustment. V and E denote the nodes/cameras and edges of EG list of tracks, {T j } j=1,...,N tracks number of spanning trees, K EG [default value = 60] output : subset of tracks, S T ranked = set of ranked tracks in T in decreasing priority (length-scale-cost criterion) Initialize k = 0 k is the counter of spanning trees Initialize S = {} S is the output subset of tracks while k < K EG do Compute camera weights Equation ( 8) C root = camera with largest weight, Create inverted_list using T ranked a list that sorts the tracks seen in each camera using T ranked Set l = 1 l counts the layers in the current tree…”

“…where mean(C i ) and std(C i ) correspond to the mean and the standard deviation of the average reprojection errors of visible tracks in C i . The scalar µ is a balancing factor set to 3 [12].…”

“…Ultimately, K EG indirectly controls the number of selected tracks, since higher values will result in the union of subsets found by a wider variety of spanning trees. Further details and other toy examples similar to Figure 5 describing Algorithm 4 can be found in [12].…”

“…A DLT resectioning algorithm is used to estimate a projective matrix using the 3D to 2D correspondences [25]. The center C of the projective model is used as center of rotation in (12). This is a coarse approximation, similar to [7], but from our experience the exact value of C has little impact on the bundle adjustment solution.…”

Evaluating Tie Points Distribution, Multiplicity and Number on the Accuracy of Uav Photogrammetry Blocks

“…Based on the analysis of error upper bound, Liu et al (2014) and similarly Cui et al (2015) selected a subset of matches that has a good quality vs. quantity trade-off to enhance the accuracy of two-view SfM. In Cui et al (2017) a fast tracks selection method to improve both efficiency and robustness of the bundle adjustment is proposed. In their method, three selection criteria of Compactness, Accurateness and Connectedness are introduced: the first two are used to calculate a selection priority for each track and the third is to guarantee the completeness of scene structure.…”