Improved CAMshift Based on Supervised Learning
CAMshift algorithm refers on back-projected distribution of target object’s colour to locate the location of the target object in the subsequent frame. However, this mechanism becomes inaccurate when one or more foreign objects that share the same colour features with the target object are very close to one another, resulting these objects are in the same search window. Therefore, this study proposed the embedment of two binary classifiers trained by SVM into the existing CAMshift. These classifiers were modeled to verify the back-projected distribution under 4 types of representations and to distinguish target and non target objects. The aim is to maintain the search window to cover only the target object during tracking. Experiments were conducted to verify the performance of the classifier under three environments namely easy, adjacent and cluttered. Results have shown that the classifier has managed to classify true detection with up to 80%.
KeywordsCAMshift Support Vector Machine object tracking
Unable to display preview. Download preview PDF.
- 2.Bradski, G.R.: Computer Vision Face Tracking for Use in Perceptual User Interface. Intel Technology Journal 2(2), 12–21 (1998)Google Scholar
- 3.Huang, S., Hong, J.: Moving Object Tracking System Based on CAMshift and Kalman Filter. In: International Conference on Consumer Electronics, Communications and Networks, pp. 1423–1426 (2011)Google Scholar
- 4.Exner, D., Bruns, E., Kurz, D., Grundhoefer, A., Bimber, O.: Fast and Robust CAMshift Tracking. In: IEEE International Workshop on Computer Vision for Computer Games, pp. 9–16 (2010)Google Scholar
- 5.Nicole, M.A.: A Comparison of Mean-shift Tracking Methods. In: 12th Central European Seminar on Computer Graphics, pp. 197–204 (2008)Google Scholar
- 6.Horn, B.K.P.: Robot vision. MIT Press, Cambridge (1986)Google Scholar
- 8.Anthony, G., Gregg, H., Tshilidzi, M.: Image Classification using SVMs: One-Against-One vs One-Against-All. In: Proceedings of the 28th Asian Conference on Remote Sensing (2007)Google Scholar