Design And Implementation Of An Embedded Three Dimensional Palmprint Recognition System

The conventional two-dimensional (2D) palmprint recognition is a fast and effective personal authentication method using 2D images. At present, the Equal Error Rate (EER) of 2D palmprint recognition on a medium size palmprint database can reach 0.1% or less. It is hard to improve the accuracy by just using 2D images. Although 2D palmprint recognition can achieve high accuracy, the 2D palmprint images can be easily counterfeited and much three-dimensional (3D) depth information is lost in the imaging process. With the development of 3D techniques, it is possible to capture 3D palmprint online. 3D palmprint recognition has opened up a new field of palmprint recognition. It not only makes up for some lack of 2D palmprint recognition, but also gives its contributions to other 3D biometric authentication at the method level.We design and finish the system, this dissertation includes the following parts:1. Design an embedded 3D palmprint system. The system contains embedded engine and I/O devices two parts. The embedded engine uses DM642, and the I/O devices include capture part, projecting part, store part and internet part.2 Develop and build a 3D palmprint collecting system. In order to make a balance between the accuracy, speed and cost, we adopted structured-light scanning technique to build the embedded 3D palmprint capturing device.3 Devise the 3D palmprint local feature extraction and matching algorithm. Local features extracted from 3D palmprint contain line and texture features. It is difficult to extract line and texture features from 3D data directly, because even for the most prominent principal lines, the depth variation is very small. For this reason, we defined Mean Curvature Image (MCI) and Gaussian Curvature Image (GCI) based on surface curvature calculation. Beside MCI and GCI, we extracted the line and texture features of 3D palmprint. Then use them to recognize people.