Digital Curvatures Applied to 3D Object Analysis and Recognition: A Case Study

TL;DR

This paper introduces a digital curvature-based approach for 3D object analysis and recognition, leveraging global and local curvature properties to improve shape classification and face recognition tasks.

Contribution

It presents a novel multi-scale and vector-based method using digital Gaussian and mean curvatures for 3D shape analysis and recognition.

Findings

Gaussian curvatures capture global shape features

Mean curvatures identify local features and extrema

Effective for face recognition and shape classification

Abstract

In this paper, we propose using curvatures in digital space for 3D object analysis and recognition. Since direct adjacency has only six types of digital surface points in local configurations, it is easy to determine and classify the discrete curvatures for every point on the boundary of a 3D object. Unlike the boundary simplicial decomposition (triangulation), the curvature can take any real value. It sometimes makes difficulties to find a right value for threshold. This paper focuses on the global properties of categorizing curvatures for small regions. We use both digital Gaussian curvatures and digital mean curvatures to 3D shapes. This paper proposes a multi-scale method for 3D object analysis and a vector method for 3D similarity classification. We use these methods for face recognition and shape classification. We have found that the Gaussian curvatures mainly describe the global features and average characteristics such as the five regions of a human face. However, mean curvatures can be used to find local features and extreme points such as nose in 3D facial data.