I used BackgroundSubtractorMOG2 for foreground detection and obtain a mask to apply frames before pedestrian detection by HOG. However, I could not see any argument in the detect() method of HOG. How can I use this mask to reduce the computation time of the detection process?
You should test (for pedestrian) only foreground part of your image: get the bounding box of each connected component and apply detect on each of them. This should considerably speed up your detection, as long as the connected component extraction is not slower than the detection.
Also keep in mind that if you are not using the multi-scale approach, you have to keep connected components of (almost) the same size as the size of pedestrian you are expected.
You could use find contours, or a blob detector, or a hand made approach, it depends on the quality of your background subtraction, and the morphological operations you may have applied.
Asked: 2015-06-09 09:03:12 -0600
Seen: 310 times
Last updated: Jun 09 '15
