LKT Tracking Reinforced using HSV Histogram Backprojection - Fixed Object
-- Using SIFT to first recognize the object.
-- Populating the obtained bounding box on the object, via SIFT & Homography Matrix, with GFTT points.
--Tracking the GFTT points via LKT and putting a bounding box over them in successive frames.
--It was observed (rather naively) that most of the GFTT points get lost in translation & track is lost when the object goes out of frame.
--Created a HueSaturation-histogram model of the object when it is first identified.
--Backprojected on the frame (every 10) and the bounding box is placed on that.
--Manually adjust the HS scale for higher accuracy.
--Good Results even if the object goes out of frame !
-- Used OpenCV C++ libraries
Location - UTRGV-Brownsville Campus
Source Code :
Github - https://github.com/bassamarshad/Track...