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

Abstract

Abstract-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.

Scale-Invariant Vote-Based 3D Recognition and Registration from Point Clouds

“…In 3D, similar success stories have been reported with methods using local features [8,15,17,20,24,41]. These approaches can integrate information from a large number of object parts.…”

“…For each interest region, i.e., a sphere centered at c with radius r, a 3D canonical orientation based on the geometry of the points inside the region is computed, for example by finding the principal directions of the points using PCA [24], or by fitting a local surface and then finding most repeatable directions on the surface from the center of the sphere [33]. A local reference frame is created as a result, hereinafter feature frame, originating at c, with one unit length equal to r, and with 3D orientation coinciding with the 3D canonical orientation.…”

“…Tombari and Di Stefano [41] first compute Hough votes over translation, assuming known scale in their 3D object recognition and registration application, then determine rotation by averaging the rotations at each mode. Geometric hashing [11,24] is a similar technique to Hough voting which reparameterizes pose in a lower dimensional space before clustering. However, all these dimensionality reduction techniques lead to an increased chance of false positive detections.…”

Rigid 3D geometry matching for grasping of known objects in cluttered scenes

“…The scenes used in the following test cases were digitized with a Minolta VIVID 910 scanner [Konica Minolta, 2011] and provided by [Mian et al, 2006]. Examples and more information about the scenes will be given in the following subsections.…”

A graph-based technique for semi-supervised segmentation of 3D surfaces

“…The complexity of such primitives can range from basic items, such as planes, cylinders or spheres, to complete parametrized models, depending on the overall goal of the pipeline. Simple primitives are fitted to the segmented parts mainly for object simplification (Lafarge et al, 2010;Baillard and Zisserman, 1999), while completed models can be used for direct 3D object recognition with resilience to clutter (Mian et al, 2006) or invariance to scale (Bariya and Nishino, 2010). Finally, another important application that needs surface decomposition is the angle and distance-preserving piecewise parametrization needed to apply textures to objects (Wang et al, 2005).…”