Three-Dimensional Model-Based Object Recognition and Segmentation in Cluttered Scenes

By Ajmal Saeed Mian, Mohammed Bennamoun, Robyn A. Owens. 2006

543 citations recorded by the source corpus.

Viewpoint independent recognition of free-form objects and their segmentation in the presence of clutter and occlusions is a challenging task. We present a novel 3D model-based algorithm which performs this task automatically and efficiently. A 3D model of an object is automatically constructed offline from its multiple unordered range images (views). These views are converted into multidimensional table representations (which we refer to as tensors). Correspondences are automatically established between these views by simultaneously matching the tensors of a view with those of the remaining views using a hash table-based voting scheme. This results in a graph of relative transformations used to register the views before they are integrated into a seamless 3D model. These models and their tensor representations constitute the model library. During online recognition, a tensor from the scene is simultaneously matched with those in the library by casting votes. Similarity measures are calculated for the model tensors which receive the most votes. The model with the highest similarity is transformed to the scene and, if it aligns accurately with an object in the scene, that object is declared as recognized and is segmented. This process is repeated until the scene is completely segmented. Experiments were performed on real and synthetic data comprised of 55 models and 610 scenes and an overall recognition rate of 95 percent was achieved. Comparison with the spin images revealed that our algorithm is superior in terms of recognition rate and efficiency

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

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 Computer Vision and Pattern Recognition

Citations in the corpus

Model Based Training, Detection and Pose Estimation of Texture-Less 3D Objects in Heavily Cluttered Scenes (2012, methodology): “Some of these methods (which assume that a full 3D model is available) include spin-images [17], point pairs [18, 19], and point-pair histograms [20, 21].”
Voting-based pose estimation for robotic assembly using a 3D sensor (2012, methodology): “In this paper, we consider geometric primitives in a pair, which we refer to as pair feature [22], [23], [24].”
Fast Global Registration (2016, methodology): “Next, we evaluate our method on the UWA dataset [27].”
On the Repeatability and Quality of Keypoints for Local Feature-based 3D Object Retrieval from Cluttered Scenes (2009, methodology): “3) or mesh representation (e.g. Mian et al. 2006b).”
Recognizing Objects in 3D Point Clouds with Multi-Scale Local Features (2014, methodology): “The UWA dataset is currently regarded as the most popular benchmark for 3D object recognition [4,10,11,13,19,27,32].”
Multimodal templates for real-time detection of texture-less objects in heavily cluttered scenes (2011, background): “We then present quantitative evaluations for challenging scenes that show that our multimodal templates outperform state-of-the-art approaches.”
Scale-hierarchical 3D object recognition in cluttered scenes (2010, methodology): “in [10] and compute their scale-dependent representation using a set of five discrete scales.”
Fast 3D recognition and pose using the Viewpoint Feature Histogram (2010, methodology): “A system for fully automatic 3D model-based object recognition and segmentation is presented in [17] with good recognition rates of over 95% for a database of 55 objects.”
Hough Transform and 3D SURF for Robust Three Dimensional Classification (2010, background): “So far, relatively few 3D categorisation methods based on local features, like tensors [5], heat kernel signatures [6], integral shape descriptors [7, 8], and scale dependent features [9] have been proposed.”
Learning to Rank 3D Features (2014, methodology): “[22] which is the standard benchmark for evaluating 3D recognition and pose estimation algorithms.”
Deep Reconstruction Models for Image Set Classification (2015, background): “Compared with single image based classification, it offers more promises and has therefore attracted significant research attention in recent years.”
DeepIM: Deep Iterative Matching for 6D Pose Estimation (2018, background): “These include point pairs (Mian et al. 2006; Hinterstoisser et al. 2016), spin-images (Johnson and Hebert 1999), and point-pair histograms (Rusu et al. 2009; Tombari et al. 2010).”
SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration (2020, methodology): “Early methods to extract local geometries include PS [9], 3DSC [19], ISS [67] and SHOT [40], which simply compute the low-level features such as faces [41, 64], corners [52], and handcrafted statistical histograms [43].”
T-LESS: An RGB-D Dataset for 6D Pose Estimation of Texture-Less Objects (2017, background): “The depth-only dataset of Mian et al. [34] includes 3D mesh models of 5 objects and 50 test depth images acquired with an industrial range scanner.”