Presenting an Effective Algorithm for Tracking of Moving ObjectBased on Support Vector Machine

  Hamed Mohammadi Azni - Computer and IT engineering Faculty, Islamic Azad University, Qazvin Branch, Iran
  Fariborz Mahmudi - Computer and IT engineering Faculty, Islamic Azad University, Qazvin Branch, Iran
Shahrbano Akbarpoor - Islamic Azad University, Jouybar Branch, Jouybar, Iran

In this paper, an effective algorithm has been presented for object tracking in videoimages using color and texture features and with the help of accelerated support vector machine. In the proposed method, object region is firstly determined by user in the first frame; then, a region as area as the object and around it is considered as background. After that, color and texture features are extracted from object and background regions, trained to the support vector machine having been accelerated by subtracting training vectors, and tested. The first output will be a binary image in which the object has been exactly separated from its surrounding background. Then, color and texture features of the exact background region being obtained in the previous stage are developed in order to become resistant against background changes in next frames. In the following, features of object and developed background are used again for training of support vector machine and they have been used for recognition of object pixels in the next frame. In the proposed method, center of gravity of object and mean shift process have been used for object locating.

computer vision, object tracking, Support Vector Machine,texture, mean shift