Robust Extraction of Vertices in Range Images by Constraining the Hough Transform

Abstract

Abstract. We describe a technique for extracting vertices from range images of cluttered box-like objects. Edge detection is performed and an edge map is acquired. Extraction of vertices is carried out using the edge map and comprises two steps: Linear boundary detection in 3D and boundary grouping. In order to recover the four parameters of a 3D linear segment, we decompose the problem in two 2D subproblems, each recovering two line parameters. These subproblems are solved by means of the Hough Transform, constrained in this way so that accurate and efficient propagation of the edge points localization error is achieved. Pairs of orthogonal boundaries are grouped to form a vertex. The orthogonality of a boundary pair is determined by a simple statistical test. Our strategy comprises many advantages, the most important of which robustness, computational efficiency and accuracy, the combination of which is not to be found in existing approaches.

Model globally, match locally: Efficient and robust 3D object recognition

“…The two-point feature used in our method is based on the idea of surflet pairs. Several approaches detect objects using a variant of the Generalized Hough Transform [8,14,25] but are limited to primitive objects as the recovery of a full 3D pose with 6 degrees of freedom is computationally too expensive. Schnabel et al [18] detect primitives in point clouds by using an efficient variant of RANSAC.…”

“…Global approaches [8,13,14,18,23,25] are typically neither very precise nor fast, and are limited mainly to the classification and recognition of objects of certain type. By contrast, local approaches that are based on local invariant features [1,4,5,6,7,10,15,17,19,20,21,24] became extremely popular and proved to be quite efficient.…”

Reliable recovery of piled box-like objects via parabolically deformable superquadrics

“…All possible pairs of detected lines are considered and orthogonal pairs are inserted to the set of the detected vertices. Vertex detection is described in detail in [13].´ µ: Surfaces of superquadric seeds are aligned to the detected vertices.´ Úµ: Seed evolution takes place, in the context of which the model parameter recovery problem is decomposed into two subproblems:…”

Localization of Piled Boxes by Means of the Hough Transform

“…Pairs of lines found to be orthogonal, along with their intersection point are grouped to a vertex. Line detection in 3d is performed via a series of Hough Transforms (see [6] for details). An interesting feature of our vertex detector is that constrains the transform in this way, so that it allows for efficient and accurate propagation of the edge points localization error in the parameter space.…”

“…The reader may have already observed that not all the linear boundaries and as a consequence not all of the vertices of the graspable surfaces have been recovered. The adopted line detection guarantees detection of all boundaries in the image up to a user defined probability of success (see [6], [9]). The execution time of the algorithm depends exponentially on this number.…”