Intelligent system for cross-spectral and cross-distance face matching

Making face recognition system more and more intelligent is an active research area that has great challenges like identifying faces captured from long distances and at night time. This task involves cross-spectral and cross-distance comparison of facial probe images with the gallery facial data that are taken indoor and from shorter distances. In this paper, a new approach for night time and long distance face recognition has been proposed. The major stages of face recognition including pre-processing, feature extraction and matching has been investigated. The proposed approach is tested using the Long Distance Heterogeneous Face (LDHF) Database comprising visual (VIS) and Near Infra-Red (NIR) images taken at distances 1 m, 60 m, 100 m and 150 m. The combination of wavelet based Histogram of Oriented Gradients (HOG) feature extractor and Local Binary Pattern (LBP) features has yielded comparatively better results for long distances than the competing techniques.