3D Shape Analysis

11Object Recognition in 3D Scenes

11.1 Introduction

Object recognition has a large number of applications in various areas such as robotics, autonomous vehicles, augmented reality, surveillance systems, and automatic assembly systems. The task of 3D object recognition is to determine, in the presence of clutter and occlusion, the identity and pose (i.e. position and orientation) of an object of interest in a 3D scene 129. Compared to 2D images, 3D data usually provide more accurate geometrical and pose information about 3D objects. Thus, 3D object recognition in cluttered scenes is always an important research topic in computer vision. Recently, the rapid development of low‐cost 3D sensors (such as Microsoft Kinect, Asus Xtion, and Intel Realsense) has further boosted the number of applications of 3D object recognition systems.

Existing 3D object recognition methods can be classified into surface matching‐based and machine learning‐based methods. In this chapter, we first cover the surface registration‐based 3D object recognition methods (Section 11.2). We then introduce some recent advances in machine learning‐based 3D object recognition methods (Section 11.3). We finally summarize this chapter with some discussions in Section 11.4.

11.2 Surface Registration‐Based Object Recognition Methods

Most traditional 3D object recognition methods follow a hypothesize‐and‐test surface registration scheme, which usually contains three modules: (i) a feature matching module, (ii) a hypothesis ...