ShapeFit and ShapeKick for Robust, Scalable Structure from Motion

摘要

We introduce a new method for location recovery from pair-wise directionsthat leverages an efficient convex program that comes with exact recoveryguarantees, even in the presence of adversarial outliers. When pairwisedirections represent scaled relative positions between pairs of views(estimated for instance with epipolar geometry) our method can be used forlocation recovery, that is the determination of relative pose up to a singleunknown scale. For this task, our method yields performance comparable to thestate-of-the-art with an order of magnitude speed-up. Our proposed numericalframework is flexible in that it accommodates other approaches to locationrecovery and can be used to speed up other methods. These properties aredemonstrated by extensively testing against state-of-the-art methods forlocation recovery on 13 large, irregular collections of images of real scenesin addition to simulated data with ground truth.