An Efficient RANSAC for 3D Object Recognition in Noisy and Occluded Scenes

Abstract

In this paper, we present an efficient algorithm for 3D object recognition in presence of clutter and occlusions in noisy, sparse and unsegmented range data. The method uses a robust geometric descriptor, a hashing technique and an efficient RANSAC-like sampling strategy. We assume that each object is represented by a model consisting of a set of points with corresponding surface normals. Our method recognizes multiple model instances and estimates their position and orientation in the scene. The algorithm scales well with the number of models and its main procedure runs in linear time in the number of scene points. Moreover, the approach is conceptually simple and easy to implement. Tests on a variety of real data sets show that the proposed method performs well on noisy and cluttered scenes in which only small parts of the objects are visible.

3D Object Detection and Recognition for Robotic Grasping Based on RGB-D Images and Global Features

“…In 2010, Hinterstoisser et al proposed a template based recognition method, which used a large RGB-D template database containing objects registered from different angles [21] 220 W. Czajewski, K. Kołomyjec and Papazov and Burschka developed a method for object recognition in cluttered scenes using point clouds [35]. Less than a year later, the Point Cloud Library -an open source library of algorithms for 3D point cloud processing -was introduced [38].…”

Predicting human intention in visual observations of hand/object interactions

“…In our work, the observations are acquired using a RGBD sensor [18]. Translation is achieved by means of a novel hand-object 3D tracking system that combines state-of-the-art methods for object recognition [25], [26] and hand-object 3D tracking [23], [22].…”

Scene Perception and Recognition in Industrial Environments for Human-Robot Interaction

“…One is an extension of the ORR method [9], [10] method is more suitable for handling arbitrary object geometries and objects having few primitive shapes while the PSG method is more suitable for large models which decompose into a large number of stable primitive shapes. The pipeline for this module is shown in Fig.…”

“…The advantage of these methods, however, lies in their computational speed. Some other methods such as [7], [9], [10] provide robustness to occlusions, partial views and noisy data. However, these methods are rather slow and not suitable for real-time applications in large scenes.…”

“…However, these methods are rather slow and not suitable for real-time applications in large scenes. In this paper, an extension to the Object Recognition RANSAC (ORR) [9], [10] method has been proposed, where the effort has been directed towards a solution which enhances its robustness to noisy sensor data and also increases its speed. Another object recognition and pose estimation algorithm has been proposed, which is complementary to the PSORR method with respect to the target object geometries.…”