Fast 3D recognition and pose using the Viewpoint Feature Histogram

By Radu Bogdan Rusu, Gary R. Bradski, Romain Thibaux, John M. Hsu. 2010

974 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computational Mechanics, Geology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computational Mechanics

In Geology

Citations in the corpus

Learning human activities and object affordances from RGB-D videos (2012, background): “The high-level activities are {making cereal, taking medicine, stacking objects, unstacking objects, microwaving food, picking objects, cleaning objects, taking food, arranging objects, having a meal}.”
Data-Driven Grasp Synthesis—A Survey (2013, positive): “Rusu et al. [104, 105] provide extensions of [106] for either recognizing or categorizing objects and estimating their pose relative to the viewpoint.”
A Global Hypotheses Verification Method for 3D Object Recognition (2012, methodology): “, [9, 10], compute a single descriptor which encompasses the whole object shape: this requires, in presence of clutter and/or occlusions, the scene to be pre-processed by a suitable 3D segmentation algorithm capable of extracting the individual object instances.”
Ensemble of shape functions for 3D object classification (2011, background): “• VFH: The Viewpoint Feature Histogram is a descriptor based on normal vectors and was introduced in [7].”
3DNet: Large-scale object class recognition from CAD models (2012, methodology): “Another available post-processing step for ﬁltering is available in the framework using the approach proposed in [24], where the similarity of nearby views is used to ﬁlter accidental matches.”
GDRNPP: A Geometry-Guided and Fully Learning-Based Object Pose Estimator (2021, positive): “Given that direct pose regression networks currently exhibit suboptimal performance, most methods still resort to traditional techniques to varying degrees.”
CAD-model recognition and 6DOF pose estimation using 3D cues (2011, background): “Though VFH showed promising results in [10] , it has a few shortcomings: • it is invariant to the size of the object as the compound histogram is normalized by the total number of points in the partial view; • it is invariant to rotations around the camera's view di rection, so it does not allow…”
BoVDW: Bag-of-Visual-and-Depth-Words for gesture recognition (2012, background): “been recently developed [9, 10], which are based on codifying the distribution of normal vectors among regions in the 3D space.”
Performance of global descriptors for velodyne-based urban object recognition (2014, background): “Generally, these descriptors can be roughly divided into two groups: the local descriptors for every point of the object [7], [8], [9], [10], [11], [12], [13] and the global descriptors for the whole object [3], [5], [14], [15], [16].”
Unsupervised discovery of object classes with a mobile robot (2014, methodology): “We use the Viewpoint Feature Histogram [2], which computes a single 308-element descriptor for each segment, and compute a distance between segments using the χ 2 distance.”
Deep Learning Advances in Computer Vision with 3D Data (2017, methodology): “On the other hand, global descriptors-based methods (e.g., Rusu et al. [2010], Aldoma et al. [2012b], Tang et al. [2012], and Wang et al. [2013]) employ global descriptors and require segmentation of the scene in candidate objects, since only one descriptor is computed for each input point cloud.”
Unsupervised Point Cloud Pre-training via Occlusion Completion (2020, methodology): “We compare OcCo with Jigsaw and two hand-crafted point cloud global descriptors: viewpoint feature histogram (VFH) [40] and M2DP [17] in Table 7.”
3D Object Recognition in Cluttered Scenes with Local Surface Features: A Survey (2014, background): “Examples in this category include geometric 3D moments [18], shape distribution [19], viewpoint feature histogram [20], and potential well space embedding [21].”
3D Object Detection with Multiple Kinects (2012, methodology): “Spin image [11], VFH [20], or 3D shape context [5] are prominent examples.”
Scan Context++: Structural Place Recognition Robust to Rotation and Lateral Variations in Urban Environments (2021, methodology): “For the initial 3D LiDAR place recognition methods [22, 23, 24], local keypoint-based approaches were used, similar to following the early history described above in the visual domain.”
A category-level 3-D object dataset: Putting the Kinect to work (2011, background): “D shape context [14], and the recent VFH model [22]–but this list is not exhaustive.”
Paris-Lille-3D: A large and high-quality ground-truth urban point cloud dataset for automatic segmentation and classification (2017, methodology): “The detailed processing pipeline is: extraction of the ground by region growing on an elevation map, segmentation of objects by connectivity of the remaining point cloud, computation of descriptors on each object (some simple geometric descriptors inspired by [SM14] and some 3D de- scriptor of the l”