Visual Tracking Using HOG and SVM

Deepthi, Mr Mohammed, Anvar
2016 International Journal of Advanced Engineering Research and Technology (IJAERT)   unpublished
There exist a large number of visual tracking methods with revealing success. But the challenging problem in visual tracking is to handle the appearance changes of target object because of its adaptive ability. Difficulties in tracking objects can arise due to non-rigid motion, rapid movement, large variation of pose and scale, occlusion and drifts etc. One of the main reason for such failures is that, the unsuccessful image representation schemes of many algorithms. In this paper, we present a
more » ... visual tracking method using HOG and SVM along with multiple kernel algorithm. The proposed work focuses on the implementation of the HOG features (Histogram of Oriented Gradients) to distinguish the target and the background with HOG visualization and facilitate classification using Support Vector Machine (SVM). Inorder to make our method more effective we are using a boosting technique to select good SVMs. We also include an update scheme to account for object appearance variance. Our tracker is able to handle occlusion and performs against existing visual tracking algorithms in handling various conditions. Qualitative and quantitative evaluations over various challenging sequences shows the competitive performance of our tracking algorithm.
fatcat:zo7oj7ugzrep7nxzxei55bpe5a