1

I'm working on a real time system and matching ORB features using BFMatcher(cv::NORM_HAMMING)

I'm matching a single image features to 9 others in a loop using 

matcher->knnMatch(m_descriptorMat, matches, 2)

Usually the matching takes 1-2 ms per loop iteration (1 image per iteration)

matcher is trained before starting the matching.

On occasions, a single iteration would slow down and take ~200-300 ms (and the following iteration would again be 1-2 ms)

The size of each descriptor is ~300-400 features (8-bit)

Any ideas why repeated calls to the same matcher would suddently slow down?

La bla bla
  • 8,558
  • 13
  • 60
  • 109
  • 1
    [Possibly related (although I have no idea)](https://stackoverflow.com/questions/11227809/why-is-it-faster-to-process-a-sorted-array-than-an-unsorted-array) – GPPK Nov 14 '17 at 12:48

0 Answers0