Translator Disclaimer
24 November 2009 Motion clustering and object detection via modulated integral imaging
Author Affiliations +
A moving object detection method based on the dense velocity field detection followed by an unsupervised spatiotemporal clustering by means of the mean shift algorithm is proposed. The method firstly makes use of a novel imaging system and an algebraic solution to detect the extremely dense velocity vector distribution with a pixel-wise spatial resolution and a frame-wise temporal resolution. It is based on the complex sinusoidal-modulated imaging with a three-phase correlation image sensor (3PCIS) and an exact algebraic inversion method based on an equivalent weighted integral form of the optical flow partial differential equation. Since the inversion method is free from time derivatives, any limitations on the object velocity and inaccuracies due to approximated time derivatives are thoroughly avoided. Secondly, in order to segregate dominant velocity regions from noisy background and identify the exact moving object, the mean shift clustering method is employed to the spatially and temporarily dense velocity vector distribution. To avoid too many cluster centers or over clustering, the traditional mean shift method is improved by setting an emerging condition. An experimental system was constructed with a 320×256 pixel 3PCIS device and a standard PC for inversion operations and display. Several experimental results are shown including an application to dense motion capture of face and gesture and traffic scene, which including several independent moving objects.
© (2009) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Peizhen Wang, Shigeru Ando, and Toru Kurihara "Motion clustering and object detection via modulated integral imaging", Proc. SPIE 7513, 2009 International Conference on Optical Instruments and Technology: Optoelectronic Imaging and Process Technology, 75130S (24 November 2009);

Back to Top