This Summer I had the opportunity to spend three weeks working at the NASA Ames Research Center in Mountain View California. My stay there was part of the CRCD program

Figure 1. View of NASA Ames Research Center from across a field of mustard and wild flowers.

Abstract

In this summer project, an algorithm to segment colors from a colored glove was implemented, namely Expectation-Maximization. Test cases and examples of its implementation will be given. The issues that arose during the implementation will be discussed and recommendations are given.

Introduction

The task of detecting and representing the hand in a virtual environment is not easy. It becomes evident that the job has to be subdivided into smaller tasks. One of those parts is hand segmentation. The outcome from this part will be used in the hand reconstruction. This step is very sensitive to noise so the method has to be as accurate as possible. There are several techniques in the literature for performing color segmentation. The most robust techniques use a probabilistic model for the color. The technique used for segmenting the colored fingers in the glove is called Expectation Maximization (EM) and it belongs to the class of probabilistic methods.