This project is based on ARToolkit 5.3.2, replace its force-match image retrieval with libvot(vocabulary treee). The main framework is aimed to provide multiple threads for tracking and a simple detect thread for image retrieval.
1.extract key points and descriptors
To start this program, you first need to learn how to use ARToolkit, and check out some of its workflow, and also you must be able to complie the original examples without errors. And you need to learn how to use libvot(https://github.com/psastras/vocabtree), the sift files and tree are generated by it. For this program, I use some another libaries, such as Opencv, Opengl, GLUT, etc.You need to install opencv 2.4.13 (I found some errors occured with different opencv version.) and make sure the makefile is no problem (replace the lib and include directories). The OS I used is mac osx 10.12.6, but i think it could be complied on linux if the required libraries and include files are okay. Remember to modify "Makefile" and "nftSimple.cpp", cause I just use my local absolutely directory.
Once you unzip the ARToolkit package, cd examples/nftSimple, rm -rf * , then git clone this program.
Then try make...
You can get ../../bin/nftSimple if no errors occured.
And then you could start it under termial.
I recommended to use "640*480" resolution.