Centre of mass model - A novel approach to background modelling for segmentation of moving objects

This paper describes a novel method, centre of mass model, to detect moving objects in a dynamic scene based on background subtraction. Any displacement of the position of centre of mass (CoMs) in two consecutive frames is the indicator of a moving object in a scene. Dividing a scene into subregions and modelling them as individual masses allow segmentation of the moving object(s). In the proposed scheme, an image is divided into blocks that are called super-pixels and each super-pixel is represented with the x and y components of CoM of a block. The segmentation is achieved by taking the absolute difference between CoM of current super-pixel and the mean of CoMs of previous corresponding super-pixels, and thresholding the difference with a dynamically updated value. A comparative work has been carried out to evaluate the performance of the proposed model and the previously reported seven different methods. The model produced consistent outputs for the images taken in different environmental conditions. The moving objects were successfully segmented with no post-processing operations. Centre of mass model demonstrated better overall performance than the methods previously reported. Its output was superior for auto-focused video images.

[1]  Stephen J. McKenna,et al.  Tracking interacting people , 2000, Proceedings Fourth IEEE International Conference on Automatic Face and Gesture Recognition (Cat. No. PR00580).

[2]  Alex Pentland,et al.  Pfinder: Real-Time Tracking of the Human Body , 1997, IEEE Trans. Pattern Anal. Mach. Intell..

[3]  Ramesh Jain,et al.  Introduction to Machine Vision , 1995 .

[4]  Rangachar Kasturi,et al.  Machine vision , 1995 .

[5]  Yee-Hong Yang,et al.  The background primal sketch: An approach for tracking moving objects , 1992, Machine Vision and Applications.

[6]  Larry S. Davis,et al.  W/sup 4/: Who? When? Where? What? A real time system for detecting and tracking people , 1998, Proceedings Third IEEE International Conference on Automatic Face and Gesture Recognition.

[7]  Azriel Rosenfeld,et al.  Tracking Groups of People , 2000, Comput. Vis. Image Underst..

[8]  Xiaobo Li,et al.  Detection of vehicles from traffic scenes using fuzzy integrals , 2002, Pattern Recognit..

[9]  Tieniu Tan,et al.  Recent developments in human motion analysis , 2003, Pattern Recognit..

[10]  Terrance E. Boult,et al.  Frame-rate omnidirectional surveillance and tracking of camouflaged and occluded targets , 1999, Proceedings Second IEEE Workshop on Visual Surveillance (VS'99) (Cat. No.98-89223).

[11]  Kentaro Toyama,et al.  Wallflower: principles and practice of background maintenance , 1999, Proceedings of the Seventh IEEE International Conference on Computer Vision.

[12]  O. Silven,et al.  A real-time system for monitoring of cyclists and pedestrians , 1999, Proceedings Second IEEE Workshop on Visual Surveillance (VS'99) (Cat. No.98-89223).

[13]  Larry S. Davis,et al.  View-based detection and analysis of periodic motion , 1998, Proceedings. Fourteenth International Conference on Pattern Recognition (Cat. No.98EX170).

[14]  Hironobu Fujiyoshi,et al.  Moving target classification and tracking from real-time video , 1998, Proceedings Fourth IEEE Workshop on Applications of Computer Vision. WACV'98 (Cat. No.98EX201).

[15]  Carlos Orrite-Uruñuela,et al.  Detected motion classification with a double-background and a Neighborhood-based difference , 2003, Pattern Recognit. Lett..

[16]  W. Eric L. Grimson,et al.  Adaptive background mixture models for real-time tracking , 1999, Proceedings. 1999 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Cat. No PR00149).