A method of object tracking based on particle filter and optical flow to avoid degeneration problem

In this paper, we propose the technique for tracking the velocity of objects that combines the particle filter with optical flow. If the feature of the target is unknown, some conventional particle filters are hard to apply. This problem can be solved by estimation method of optical flow. Moreover, we introduce a new idea. If particles relocate out of the image, particles are salvaged. When a new object appear in the image, new particles are reassigned. Experimental results show that unstructured multiple objects can be tracked.