Axial motion stereo vision and structured light for three-dimensional acquisition of the human ear.

摘要

Our objective is the complete surface scan of the canal, concha and external lobe of a human ear. We present two applications, (1) canal and concha scanning for hearing aid design and, (2) biometric detection, tracking and discrimination of the ear lobe shape. Over recent years, the possibility of reconstructing three dimensional models from images has lead to an active research field. There are two common non-contact methods to extract 3D surface contours. They are stereo vision and structured light illumination. Much of the theory of stereo vision is very well understood, meanwhile, there still remain numerous problems associated with axial motion stereo vision. An ear scanning system is built for biomedical and biometrics applications. There are many challenging problems in this system such as small space, concavity, accuracy, and scale range from canal to entire lobe, as well as the merging of all the data. The traditional stereo vision methods cannot be used in this system because of these problems. The axial stereo vision method is used for canal scanning and structured light illumination method is used for concha scanning. The mathematical and geometric models of axial motion stereo vision system and structured light illumination system are presented in this study and the advantages of the stereo vision and structured light illumination systems are examined. The ear scanning prototype system and the experimental results are shown in this dissertation. The surface scan of canal and concha can be used for biomedical application to create computer aided design model for hearing aid. The surface scan of concha and lobe can be used for biometrics application for tracking and identification.;Given the 3D scan results of the ear concha and lobe, a new and efficient distortion-invariant "super image" is introduced to track and identify the biometrics. It is based on linear phase coefficient composite filter. The super image consists of a weighted sum of training images chosen to span the distortion range under analysis. Unlike correlation based composite filter design, the super image is implemented using a complex vector inner product operation. A super image vector inner product is implemented by element-wise multiplying a super image template by a window of interest in the input scene and then summing the element-wise operations. The resulting amplitude indicates target detection and the resulting phase indicates the value of scale, orientation or movement of the target object. The mathematical characteristics of super image vector inner product are presented and its application is demonstrated.;Keywords. Structured Light, Axial Stereo Vision, Composite Filters, Synthetic Discriminant Functions, Target Tracking