Alexandra Popov (12216685) Leroy Chika Anozie (12400952) Marcel Bühler (99211688)
28th of October 2015 designed with Matlab R2015a and VLFeat 0.9.20
----- How to start -----
run GUI.m in Matlab
----- GitHub URL -----
https://github.com/mbbuehler/kangoroo-tracking
----- Details -----
This object tracking algorithm uses different methods for finding key points and matching them. Using the GUI the are different parameters to select:
-
Video: Choose the video you want to process, the initial rectangle is already set. You can choose from example videos from PrinctonTrackingBenchmark Website
-
Matching algorithm: Choose between two matching algorithms a) euclidean distance b) SVM
-
Display mode: Choose between rectangles or rectangles and actual key points
-
Stop after very frame: Check if you want to see every frame before continuing manually
----- Modality of the code -----
-
reading the images and preprocessing
-
finding and describing the key points with VLFeat
-
matching the key points from the neighbourhood: mode 1: matching key points with euclidean distance mode 2: matching key points using svm [trained with 99 tracking examples]
-
Rectangles and (optionally) key points are showing the track of the selected object in the video