I am currently testing Flann feature matching with OpenCV in python, and do not fully understand what some of the parameters actually do. Here is a snippet of code copied from the OpenCV docs. The full code can be found here.
# FLANN parameters
FLANN_INDEX_KDTREE = 1
index_params = dict(algorithm = FLANN_INDEX_KDTREE, trees = 5)
search_params = dict(checks=50) # or pass empty dictionary
flann = cv2.FlannBasedMatcher(index_params,search_params)
matches = flann.knnMatch(des1,des2,k=2)
I am curious what the inputs algorithm, trees, and checks mean. The default values found in the OpenCV docs are commonly adopted by others code I have read on here, so I am therefore wondering if it is worthwhile to manipulate and test different values or if it recommended to stick with these?
What are the key differences between the various index parameter algorithms? Why is
FLANN_INDEX_KDTREEthe most commonly used?I assume that
trees = 5means that 5 KDTrees are made. What does changing this value mean in practice?The OpenCV docs mention that increasing the value of checks will give better precision but slow down the code. It says: “If you want to change the value, pass
search_params = dict(checks=100).” Is it worth exploring values other than 50 or 100 for checks?
And finally, aside from Brute Force Matcher and the Flann algorithms, are there other common feature matching algorithms?
