Support vector machine with Histogram of oriented gradient trained near online, and tracker.
Building the SVM detector based on the HOG feature is a relatively simple process. When is not necessary to be robust and detector is focused only on one object. You can build this by combining several OpenCV available tutorials and source codes distributed in Opencv samples. There is maybe one thing that is not natural and cannot be taken from examples and tutorials. Train set in online training is only 20 positive images warp over positive windows and 30 random negative samples.
What do you think?
What do you think?
Tutorial on SVM with HOG, tracker soon
This is just an example. The tutorial will be available later. The code is a little bit complex.