Image Template Tracker

Template Tracking using Opencv and Python (C++)

Worked on Laptop with

• Windows 10 Home
  • python3.6
  • opencv 3.4.0

Requirement

• numpy, pandas, matplotlib
• opencv
  • Opencv package above 3.4.0 and below 4.x (Due to the feature extraction method.)
  • both opencv and opencv-contrib is needed.

How to use

python version

python TempTracking.py -d <keypoint descriptor> -f <input video file> -t <template image>

• <keypoint descriptor> support ORB, SIFT, SURF, AKAZE, KAZE feature.
• <input video files> need to be in the ascii coded folder
• <template image> is the template image you want to track
• Default values are ORB, video 0 and object.png

With the program in ScalingAndTimeMeasurement branch, it will show the calculation time and estimated scales.

c++ (opencv with cuda8.0) version

Do following to compile and just run the program.

g++ -ggdb TempTrackGPU.cpp `pkg-config --cflags --libs opencv`

It only support SURF with cuda8.0, and you need opencv3 compiled with cuda.

DEMO

• Template image(in white box).

• Tracked result looks like:

• Calculation time ~ 100ms with SIFT in core i5-3000 laptop