Feature-Matching

using numpy and opencv

USAGE

• add the trained image to be matched with in model folder.
• replace the query_image with the image you want to match.
• open jupyter notebook in the folder.
• change the name of the image in the codes for static matching and real-time matching i.e img = cv2.imread('model\abc.jpg',1) 1 in argument represents the scale image i.e BGR that can be altered as opted
• press esc key to clear output window

feature matching:

• the features founded of both the trained image and the real-time frame were the object is to be found are matched with thevcomputed descriptors.

A descriptor provides a representation of the information given by a feature and its surroundings. Which is abstracted to a feature vector a vector that contains the descriptors of the keypoints found in the image with the reference object.

Feature matching can be implemented using Brute-Force based matching, FLANN based matching and many a more effective algorithm. However, I find FLANN based matching somewhat accurate giving a ratio of 500 best matches with approximately false matches in range of 1 to 5 which is quite amazing!!!

So, I have implemented BF based matching in static frames and FLANN in real-time-frame matching. There you go with the code ...