Advice tracking pedestrians
Hello, I have to write a program to track pedestrians, I thought about :
Using HOG to detect pedestrians in a given frame Associate detected pedestrians from two consecutive frames by euclidean distance. Kalman filter to smooth the trajectory for when hog doesn't work.
Is there an advice you could give me or is this a right approach ?
Thanks
That is the most frequently used approach you can find. However it will fail in many cases because HOG is far from an effective pedestrian detector. In OpenCV however it is the best one implemented as far as I know. But there are many other algorithms out there like part based models / latentSVM of Felzenswalb and Integral Channel Features of Piotr Dollar!