Depth estimation and implementation on the DM6437 for panning surveillance cameras

M. Asif, J.J. Soraghan

Research output: Contribution to conferencePaper

4 Citations (Scopus)

Abstract

Real-time efficient video analytics (VA) for surveillance system requires 'in-camera' or 'at camera edge' decision-making. Video Analytics is required to work on good resolution footage in order to register tiny but important information that may otherwise be lost during subsequent compression process, especially in temporally exploited streams. This paper presents a novel approach to estimate 3D location coordinates of selected control points from panning camera footage. The two frames at different panning angles are considered to be two images captured from two different coplanar viewpoints with some translational distance between their optical centers. To simulate panned images as coplanar the information required is the panning angle. With the help of the panning angle transformational matrix two image planes are calculated. The approach developed here will help extract motion descriptors, including and not limited to, motion activity and direction of motion activity. The performance of the algorithm is evaluated in the paper using several test sequences. This paper provides a comprehensive implementation issues onto TI DaVinci platform.

Conference

Conference16th International Conference on Digital Signal Processing
CountryGreece
CitySantorini
Period5/07/097/07/09

Fingerprint

Cameras
Decision making

Keywords

  • video analytics
  • EVM
  • evaluation module
  • DM6437
  • feature detection
  • background model
  • disparity
  • panning

Cite this

Asif, M., & Soraghan, J. J. (2009). Depth estimation and implementation on the DM6437 for panning surveillance cameras. 739-745. Paper presented at 16th International Conference on Digital Signal Processing, Santorini, Greece. https://doi.org/10.1109/ICDSP.2009.5201224
Asif, M. ; Soraghan, J.J. / Depth estimation and implementation on the DM6437 for panning surveillance cameras. Paper presented at 16th International Conference on Digital Signal Processing, Santorini, Greece.6 p.
@conference{07207cb4abc44af2b45e2949dd90dcab,
title = "Depth estimation and implementation on the DM6437 for panning surveillance cameras",
abstract = "Real-time efficient video analytics (VA) for surveillance system requires 'in-camera' or 'at camera edge' decision-making. Video Analytics is required to work on good resolution footage in order to register tiny but important information that may otherwise be lost during subsequent compression process, especially in temporally exploited streams. This paper presents a novel approach to estimate 3D location coordinates of selected control points from panning camera footage. The two frames at different panning angles are considered to be two images captured from two different coplanar viewpoints with some translational distance between their optical centers. To simulate panned images as coplanar the information required is the panning angle. With the help of the panning angle transformational matrix two image planes are calculated. The approach developed here will help extract motion descriptors, including and not limited to, motion activity and direction of motion activity. The performance of the algorithm is evaluated in the paper using several test sequences. This paper provides a comprehensive implementation issues onto TI DaVinci platform.",
keywords = "video analytics, EVM, evaluation module , DM6437 , feature detection, background model, disparity, panning",
author = "M. Asif and J.J. Soraghan",
year = "2009",
month = "7",
doi = "10.1109/ICDSP.2009.5201224",
language = "English",
pages = "739--745",
note = "16th International Conference on Digital Signal Processing ; Conference date: 05-07-2009 Through 07-07-2009",

}

Asif, M & Soraghan, JJ 2009, 'Depth estimation and implementation on the DM6437 for panning surveillance cameras' Paper presented at 16th International Conference on Digital Signal Processing, Santorini, Greece, 5/07/09 - 7/07/09, pp. 739-745. https://doi.org/10.1109/ICDSP.2009.5201224

Depth estimation and implementation on the DM6437 for panning surveillance cameras. / Asif, M.; Soraghan, J.J.

2009. 739-745 Paper presented at 16th International Conference on Digital Signal Processing, Santorini, Greece.

Research output: Contribution to conferencePaper

TY - CONF

T1 - Depth estimation and implementation on the DM6437 for panning surveillance cameras

AU - Asif, M.

AU - Soraghan, J.J.

PY - 2009/7

Y1 - 2009/7

N2 - Real-time efficient video analytics (VA) for surveillance system requires 'in-camera' or 'at camera edge' decision-making. Video Analytics is required to work on good resolution footage in order to register tiny but important information that may otherwise be lost during subsequent compression process, especially in temporally exploited streams. This paper presents a novel approach to estimate 3D location coordinates of selected control points from panning camera footage. The two frames at different panning angles are considered to be two images captured from two different coplanar viewpoints with some translational distance between their optical centers. To simulate panned images as coplanar the information required is the panning angle. With the help of the panning angle transformational matrix two image planes are calculated. The approach developed here will help extract motion descriptors, including and not limited to, motion activity and direction of motion activity. The performance of the algorithm is evaluated in the paper using several test sequences. This paper provides a comprehensive implementation issues onto TI DaVinci platform.

AB - Real-time efficient video analytics (VA) for surveillance system requires 'in-camera' or 'at camera edge' decision-making. Video Analytics is required to work on good resolution footage in order to register tiny but important information that may otherwise be lost during subsequent compression process, especially in temporally exploited streams. This paper presents a novel approach to estimate 3D location coordinates of selected control points from panning camera footage. The two frames at different panning angles are considered to be two images captured from two different coplanar viewpoints with some translational distance between their optical centers. To simulate panned images as coplanar the information required is the panning angle. With the help of the panning angle transformational matrix two image planes are calculated. The approach developed here will help extract motion descriptors, including and not limited to, motion activity and direction of motion activity. The performance of the algorithm is evaluated in the paper using several test sequences. This paper provides a comprehensive implementation issues onto TI DaVinci platform.

KW - video analytics

KW - EVM

KW - evaluation module

KW - DM6437

KW - feature detection

KW - background model

KW - disparity

KW - panning

U2 - 10.1109/ICDSP.2009.5201224

DO - 10.1109/ICDSP.2009.5201224

M3 - Paper

SP - 739

EP - 745

ER -

Asif M, Soraghan JJ. Depth estimation and implementation on the DM6437 for panning surveillance cameras. 2009. Paper presented at 16th International Conference on Digital Signal Processing, Santorini, Greece. https://doi.org/10.1109/ICDSP.2009.5201224