A Brief Survey on 2D and 3D Feature Extraction

Abstract

The detection, description, and comparison of data contents is of great importance for a variety of research domains. In this chapter, we intend to enlighten these issues in the context of 2D imagery and 3D point cloud data. Hence, we address fundamental ideas by focusing on several questions: How can we detect an object? How can we describe an object? What makes an object memorable? According to which criteria can we recognize the same object or similar objects? How could similarity be defined? In order to answer such questions which, in turn, allow us to infer a deeper understanding of the respective scene, we describe how distinctive characteristics contained in respective 2D or 3D data may be represented as features, and we thereby categorize features according to different feature types. For this purpose, we first derive a general definition for characterizing a feature. Subsequently, we focus on feature extraction from 2D imagery as well as feature extraction from 3D point cloud data. Based on these findings, we discuss the motivation for involving specific features in the scope of our work and, finally, we provide concluding remarks.