1 / 18

Tracking Humans using Multiple pairs of PTZF Cameras and Wide-Angle Cameras

Tracking Humans using Multiple pairs of PTZF Cameras and Wide-Angle Cameras. Author: Abhilash Jindal , Y7009 Brajesh Kushwaha , Y7119 Supervisor: Dr. K. S. Venkatesh Dr. Krithika Venkataramani. Aim. Identifying and tracking a VIP using 3 pairs of PTZF and wide-angle cameras.

jase
Download Presentation

Tracking Humans using Multiple pairs of PTZF Cameras and Wide-Angle Cameras

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Tracking Humans using Multiplepairs of PTZF Cameras andWide-Angle Cameras Author: AbhilashJindal, Y7009 BrajeshKushwaha, Y7119 Supervisor: Dr. K. S. Venkatesh Dr. KrithikaVenkataramani

  2. Aim Identifying and tracking a VIP using 3 pairs of PTZF and wide-angle cameras. The final system's performance can be can be described as:- • Detecting all the humans in the field of view of the wide-angle cameras. • Targeting people one by one by the wide-angle cameras. • Passing the track to the PTZF camera from the corresponding wide-angle camera. • Simultaneous zooming of all the PTZF cameras onto each person's face. • Cross-checking the combined outputs of the PTZF cameras against a human face-database to recognize our VIP. • Tracking the identified VIP by the PTZF cameras simultaneously.

  3. Overview of the work The work has been divided into 5 parts: • Control of the PTZF cameras. • Human-Tracking using single camera. • Transformation of the pixels in wide-angle camera to PTZF camera. • Fusion of data from 3 wide-angle camera for improved tracking. • Recognizing individual from the output of 3 PTZF cameras. The last part is being done as a part of a different B.Tech Project under the supervision of Dr. KrithikaVenkataramani.

  4. Background subtraction and Contour evaluation Original Frame Fore-ground Tracked Object with contour drawn

  5. Camshift Tracking • It is based on the photometric cues of the image frame. Taking color sample Histogram of the selected part

  6. Improved Tracking Original Frame Masked Frame Fore-ground

  7. Apply Camshift on each part Confidence Evaluation Divided image frame Tracked aligned parts

  8. Confidence Evaluation Histogram (Frame1) Real no. [0,1] Cross- Correlation Histogram (Current-Frame )

  9. Aligning Trackers • If confidence(tracker Legs) < threshold, flag(Legs)=0; • If(flag(Legs)==0), if(flag(torso)!=0) align(Legs, torso); else align(Legs, Head); Similarly for the other two trackers.

  10. Kalman Filtering where, zk: Measurementxk: stateuk: control input wk: process noisevk: measurement noise F: transfer matrix

  11. where, R: measurement error matrix / covariance of vkQ: covariance of wkP: error covariance • The measurement error(R) has been made inversely proportional to the confidence. An increased error ensures less importance is given to the current measurement whose confidence is low. • The weights in the Transfer matrix (F) have been set heuristically.

  12. xk = State of the model (after kth update) zk = kth measurement of parameters

  13. Current Progress on full occlusion

  14. Work to be done • Designing a controller for the PTZF camera for a better time-response during tracking. • Transforming wide-angle camera co-ordinates to the corresponding PTZF camera. • Extending the single camera tracking to multi-camera tracking.

  15. References • A. Ariel, G. Mikhail, et al. Robust Real-Time background subtraction based on Local Neighborhood patterns. EURASIP Journal on Advances in Signal Processing, 2010, 2010. • M.D. Dixit, Combining edge and color features to track partially occluded humans, M.Tech thesis, Department of Electrical Engineering,May 2009

More Related