Flagge AT Sprache auf Deutsch ändern
 JS Switch to Ajax-Version

Portfolio

Vehicle counting on traffic videos
used technique
Matlab

year
year 2009

description:
Coding of a vehicle counting algorithm in the lecture 'From Fundamental Issues to Recent and Future Developments of Automated Video Surveillance' at the TU Vienna.

arrow Implementation and adaption of different approaches
arrow Use of the HOG Descriptor [1]
Description of a vehicle texture features by gradient magnitude and gradient orientation
arrow Classification via Support Vector Machine (SVM)
Training of two classes: vehicle - non vehicle
arrow Evaluierung of the performance on the frame level
Region 1
(left)
Region 2
(right upper)
Region 3
(right lower)
Correct2396 %938 %2191 %
Incorrect14 %1562 %29 %
Total24100 %24100 %23100 %

arrow Evaluation of the performance on the vehicle counting level
Region 1
(left)
Region 2
(right upper)
Region 3
(right lower)
Correct2396 %2083 %23100 %
Incorrect14 %417 %00 %
Total24100 %24100 %23100 %


Detailed information can be found in the below attached PDF document.

[1] Dalal N.; Triggs B., “Histograms of Oriented Gradients for Human Detection”, INRIA Rhone-Alps

video evaluation video
further information

keywords: vehicle counting, vehicle detection, HoG, SVM

screenshot(s)
training data set: class car   training data set: class no car   Screenshot of the classified video  

previous project   project 4 of 17   next project
© 2018 Johannes Perl | All rights reserved | Adminlogin | Valid XHTML 1.0 Transitional Valid W3C WAI-A WCAG 1.0 Tableless Layout