SIFT-Based Error Compensation for Ear Feature Matching and Recognition System

摘要

The current ear feature matching and recognition system, based on Scale Invariant Feature Transform (SIFT) image matching algorithm, can realize the human ear feature matching and detect the displacement of the human ear so as to reproduce the human ear position and posture. However, due to the influence of image acquisition equipment performance and lighting conditions, too dark or too bright background could bring the locally underexposed or overexposed image. This could result in the loss of some image details so as to make it impossible to identity the image and the recognition rate would be reduced. In this talk, the application of image gray level normalization processing can reduce the sensitivity of imaging to light intensity. Accordingly, it will greatly improve the recognition rate of human ears. Furthermore, it has been found that even if the object is stationary, the image matching results still have certain fluctuation changes, which could be caused by the system error. In order to reduce the error, the Background-based Compensation Model (BCM) has been established based on the investigation of the system error brought by the working environment changes. The results show that, BCM can be used to compensate the system errors of ear recognition matching and further improve the matching accuracy of human ear.