Thermal imaging for enhanced foreground background segmentation

Foreground  background segmentation is the primary step of most automated video monitoring system aiming at object tracking, event detection or scene interpretation. In uncontrolled environments, with dynamic background and lighting changes, this basic task is very challenging. This work is based on the hypothesis that the combination of LWIR (8-12 m) and colour cameras can significantly improve the robustness of foreground  background segmentation. An acquisition unit with co-aligned thermal and visible fields of view is used. Starting from a state-of-the-art algorithm for moving objects extraction in colour video, we adapted the method for processing of “RGBT” video format. Pros and cons of using thermal imagers in outdoor video monitoring applications are discussed. A preliminary objective performance evaluation of detection accuracy is also presented.

[1]  KimKyungnam,et al.  Real-time foreground-background segmentation using codebook model , 2005 .

[2]  Gérard G. Medioni,et al.  Detecting and tracking moving objects for video surveillance , 1999, Proceedings. 1999 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Cat. No PR00149).

[3]  Clemente Ibarra-Castanedo,et al.  Advanced surveillance systems: combining video and thermal imagery for pedestrian detection , 2004, SPIE Defense + Commercial Sensing.

[4]  D. Rensch,et al.  Comparative Studies of Extinction and Backscattering by Aerosols, Fog, and Rain at 10.6 micro and 0.63 micro. , 1970, Applied optics.

[5]  Xin Li,et al.  Layered Representation for Pedestrian Detection and Tracking in Infrared Imagery , 2005, 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Workshops.

[6]  Keiichi Yamada,et al.  Comparison between infrared-image-based and visible-image-based approaches for pedestrian detection , 2003, IEEE IV2003 Intelligent Vehicles Symposium. Proceedings (Cat. No.03TH8683).

[7]  Xia Liu,et al.  Pedestrian detection and tracking with night vision , 2005, IEEE Transactions on Intelligent Transportation Systems.

[8]  Massimo Bertozzi,et al.  Stereo Vision-based approaches for Pedestrian Detection , 2005, 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Workshops.

[9]  I. Haritaoglu,et al.  Background and foreground modeling using nonparametric kernel density estimation for visual surveillance , 2002 .

[10]  Larry S. Davis,et al.  Real-time foreground-background segmentation using codebook model , 2005, Real Time Imaging.

[11]  Hervé Sizun,et al.  Fog attenuation prediction for optical and infrared waves , 2004 .

[12]  James W. Davis,et al.  Fusion-Based Background-Subtraction using Contour Saliency , 2005, 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Workshops.

[13]  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).