A Mobile Vision System for Urban Object Detection with Informative Local Descriptors

Publication from Digital

Fritz G., Seifert C., Paletta L., Seifer C.

, 2006


We present a computer vision system for the detection and identification
 of urban objects from mobile phone imagery, e.g., for the application
 of tourist information services. Recognition is based on MAP decision
 making over weak object hypotheses from local descriptor responses
 in the mobile imagery. We present an improvement over the standard
 SIFT key detector [7] by selecting only informative (i-SIFT) keys
 for descriptor matching. Selection is applied first to reduce the
 complexity of the object model and second to accelerate detection
 by selective filtering. We present results on the MPG-20 mobile phone
 imagery with severe illumination, scale and viewpoint changes in
 the images, performing with ? 98% accuracy in identification, efficient
 (100%) background rejection, efficient (0%) false alarm rate, and
 reliable quality of service under extreme illumination conditions,
 significantly improving standard SIFT based recognition in every
 sense, providing - important for mobile vision - runtimes which are
 ? 8 (?24) times faster for the MPG-20 (ZuBuD) database.