Rotational Projection Statistics for 3D Local Surface Description and Object Recognition

By Yulan Guo, Ferdous Sohel, Mohammed Bennamoun, Min Lu, Jianwei Wan. International Journal of Computer Vision, 2013

668 citations recorded by the source corpus.

Recognizing 3D objects in the presence of noise, varying mesh resolution, occlusion and clutter is a very challenging task. This paper presents a novel method named Rotational Projection Statistics (RoPS). It has three major modules: local reference frame (LRF) definition, RoPS feature description and 3D object recognition. We propose a novel technique to define the LRF by calculating the scatter matrix of all points lying on the local surface. RoPS feature descriptors are obtained by rotationally projecting the neighboring points of a feature point onto 2D planes and calculating a set of statistics (including low-order central moments and entropy) of the distribution of these projected points. Using the proposed LRF and RoPS descriptor, we present a hierarchical 3D object recognition algorithm. The performance of the proposed LRF, RoPS descriptor and object recognition algorithm was rigorously tested on a number of popular and publicly available datasets. Our proposed techniques exhibited superior performance compared to existing techniques. We also showed that our method is robust with respect to noise and varying mesh resolution. Our RoPS based algorithm achieved recognition rates of 100, 98.9, 95.4 and 96.0 % respectively when tested on the Bologna, UWA, Queen’s and Ca’ Foscari Venezia Datasets.

Subfields: Aerospace Engineering, Computational Mechanics, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/c8895ce929388d0499d9d85871dbd455515138eb

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 Computer Vision and Pattern Recognition

Citations in the corpus

3DmFV: Three-Dimensional Point Cloud Classification in Real-Time Using Convolutional Neural Networks (2018, background): “They can be divided into two main groups: local descriptors [8]–[11] and global descriptors [12]–[14].”
Autonomous robotic stone stacking with online next best object target pose planning (2017, methodology): “Besides the ISS descriptor we tested two additional descriptors, namely the Fast Point Feature Histogram (FPFH) descriptor [18] and the Rotational Projection Statistics (RoPS) descriptors [19].”
Learning to Detect Good 3D Keypoints (2017, background): “However, taking repeatable pairwise comparisons within 3D neighborhoods subject to arbitrary 3D rotations would require the definition of an LRF, as done bymost 3Ddescriptors, e.g., SHOT (Salti et al. 2014), MeshHoG (Zaharescu et al. 2009), ROPS (Guo et al. 2013a).”
The Perfect Match: 3D Point Cloud Matching With Smoothed Densities (2018, methodology): “Many approaches including SHOT [39], RoPS [12], USC [38] and TOLDI [43] try to ﬁrst estimate a unique local reference frame (LRF), which is typically based on the eigenvalue decomposition of the sample covariance matrix of the points in neighborhood of the interest point.”
Recognizing Objects in 3D Point Clouds with Multi-Scale Local Features (2014, methodology): “A number of feature descriptors have been proposed in the literature, including spin image [20], point’s fingerprint [28], 3D shape context (3DSC) [29], snapshot [30], variable-dimensional local shape descriptors (VD-LSD) [19], Mesh Histogram of Oriented Gradients (MeshHOG) [31], exponential map (EM”
A Comprehensive Performance Evaluation of 3D Local Feature Descriptors (2015, methodology): “…Local Surface Patch (LSP) (Chen and Bhanu 2007a), THRIFT (Flint et al. 2008), Point Feature Histograms (PFH) (Rusu et al. 2009), Fast Point Feature Histograms (FPFH) (Rusu et al. 2009), Tri-Spin-Images (TriSI) (Guo et al. 2013c), and Rotational Projection Statistics (RoPS) (Guo et al. 2013b).”
SHOT: Unique signatures of histograms for surface and texture description (2014, background): “Another hybrid descriptor is the recent Rotational Projection Statistics (RoPS) [28], which encodes robust statistics (central moments and entropy) of the distributions of vertices of the mesh, rotated along the RF axes, and projected and quantized on the 2D xy, yz, and zx planes.”
PPFNet: Global Context Aware Local Features for Robust 3D Point Matching (2018, background): “Many of those approaches try to discover a local reference frame (LRF), which is by its simplest definition, non-unique [12].”
Robust reconstruction of indoor scenes (2015, background): “In the past decade researchers have studied local shape descriptors that can be used for pruning and correspondence [20, 52, 55, 24], and proposed different types of constellations [2, 16, 43].”
3D Object Recognition in Cluttered Scenes with Local Surface Features: A Survey (2014, methodology): “and calculating the feature recall and precision under each threshold [3], [43], [44], [45], [46].”
Image Matching from Handcrafted to Deep Features: A Survey (2020, background): “2010a), rotational projection statistics (Guo et al. 2013) and tri-spin-image (Guo et al.”
RGB-D to CAD Retrieval with ObjectNN Dataset (2017, methodology): “We uniformly take random samples of 5% ≤ p Sampling ≤ 20% in each object and use RoPS [GSB ∗ 13] as the feature to describe the characteristics of the mesh clipped in a sphere with supporting radius r = 0 .”
Deep Learning Advances in Computer Vision with 3D Data (2017, methodology): “The recently proposed RoPS descriptor [Guo et al. 2013] achieved the most stable performance across different datasets.”
Deep Learning for 3D Point Clouds: A Survey (2019, background): “3D data acquired by these sensors can provide rich geometric, shape and scale information [2],”