On ear-based human identification in the mid-wave infrared spectrum

In this paper the problem of human ear recognition in the Mid-Wave Infrared (MWIR) spectrum is studied in order to illustrate the advantages and limitations of the ear-based biometrics that can operate in day and night time environments. The main contributions of this work are two-fold: First, a dual-band database is assembled that consists of visible (baseline) and midwave IR left and right profile face images. Profile face images were collected using a high definition mid-wave IR camera that is capable of acquiring thermal imprints of human skin. Second, a fully automated, thermal imaging based, ear recognition system is proposed that is designed and developed to perform real-time human identification. The proposed system tests several feature extraction methods, namely: (i) intensity-based such as Independent Component Analysis (ICA), Principal Component Analysis (PCA), and Linear Discriminant Analysis (LDA); (ii) shape-based such as Scale Invariant Feature Transform (SIFT); as well as (iii) texture-based such as Local Binary Patterns (LBP), and Local Ternary Patterns (LTP). Experimental results suggest that LTP (followed by LBP) yields the best performance (Rank1 = 80.68%) on manually segmented ears and (Rank1 = 68.18%) on ear images that are automatically detected and segmented. By fusing the matching scores obtained by LBP and LTP, the identification performance increases by about 5%. Although these results are promising, the outcomes of our study suggest that the design and development of automated ear-based recognition systems that can operate efficiently in the lower part of the passive IR spectrum is a very challenging task.