A multi-scale vehicle tracking algorithm based on Structured Output SVM

A practical vehicle tracking system is required to update the appearance change of moving vehicles in real-time. The system with fixed-size tracking window can´t trace an object effectively when the scale of the object has distinct changes. Therefore, it is important to select the scale of tracking window automatically. This paper proposes an automatic updating method of tracking window and integrate it into the Struck: Structured Output Tracking with Kernels tracking algorithm based on Haar-like features. The experimental results demonstrate that the improved algorithm can automatically select the proper size of the tracking window with the increasing or decreasing of the object vehicle´s scale.