Object Tracking

Approach Used:

• Mean Shift : Meanshift can be given a target to track, it calculates the color histogram of the target area, and then keep sliding the tracking window to the closest match (cluster center) .
• CAMShift : Just using mean shift won't change the window size if the target moves away or towards the camera, use CAMshift to update the size of window.
• Via Homography: Find the keypoints from the target and main windom, after keypoints matching utilise homography to find the target window in the main window.

Requirements:

• Python 2.7
• opencv
• Object Tracking with homography requires SIFT which is in opencv-contrib.

Usage

• Run via python filename.py
• Select a rectangular window (with help of mouse) which you want to track and hit Enter.

Result

Tracking notebook.

MeanShift

CamShift

Via Homography