You can try several heuristics once you get the features. Some ideas:
- Discard items that are too long/short. Maybe calculate the mean and standard deviation and discard features that are more than mean + 2standard deviation or less than mean - 2standard deviation.
- If you have a sequence, tracking vector from N to N+1 should be similar to N-1 to N.
- Try different feature tracking by OpenCV - FAST, SIFT, SURF etc. Each one has advantages/disadvantages based on the features you are trying to track.
