No . 2013 - 614 3 D Object Recognition using Local Shape Descriptors ∗

3D object recognition is a challenging problem with important applications such as robotic perception. The most promising approach to solving 3D object recognition is through solving the correspondence problem. The goal of the correspondence problem in the context of 3D object recognition is to find correspondences between the objects to be recognized and the scene. If correspondences exist between an object and the scene it can be hypothesized that this object exists in the scene. Once these hypotheses are verified objects are recognized. The most common approach to solving the correspondence problem for 3D object recognition has been through techniques that try to find correspondences between local regions of the models and the scene. In this report, we focus on different local techniques and highlight their weaknesses and strengths as well as provide directions for future research.