Abstract: In order to provide security assurance for aircraft ground movement, wingtip tracking is proposed which is based on video image processing and analysis techniques. Due to the poor performance of the existing tracking features for a wingtip, this paper presents a new tracking algorithm within a particle filtering tracking framework which adopts oriented gradient local binary pattern(OG_LBP) descriptors to integrate texture and contour information. The OG_LBP descriptor not only reduces the histogram dimensionality but also expresses both local and holistic features of the wingtip image to enhance the discrimination. In addition, via combining the observation information of the current frame, the new tracking algorithm establishes a model of the target state. Furthermore, the radius of particle propagation is updated adaptively by the similarity between the target model and each hypothesis of the particle, which can overcome the degeneracy problem in particle filtering and result in low computational cost. Experimental results show that the proposed algorithm can achieve an accurate and robust tracking performance for different wingtips with complex background in real-time.

Key words: visual tracking, particle filter, Bayesian network, local binary pattern, oriented gradient