Object Recognition Using a Human-like Vision Analysis System

Abstract

This paper addresses the problem of recognizing three-dimensional objects, in which it considers not only the shape of objects but also the surface conditions. These two information sources are combined by using a human-like vision analysis, which can detect the shape features in 2D and the surface condition in 3D. Here we create two kinds of database from CAD models and real world objects for Hidden Markov Model. One is the 3D to 2D omni-directional projected shape data and the other is the surface condition data extracted from each projected direction. The shape feature and the surface condition are encoded experimentally by an easy way and saved as HMMs' observation data. By performing the same process, the hidden state dataset can be obtained from the real world objects. The result shows the surface condition is a very valid argument for an object recognition system

Keywords

• 3D point cloud
• feature extraction
• HMM
• object recognition