Haartraining and traincascade are both systems to train object detectors, based on HAAR features for the haartraining or based on LBP/HOG/HAAR featurs for the traincascade. It is based on the cascade classifier techniques used by Viola & Jones face detection framework, which is in principle a boosted cascade of weak classifiers. Go read some documentation on the internet about this technique, before you start digging into the code. It is important that you grasp the principle.
To use the files, create empty C++ projects, add all cpp and hpp files to your project. Build and use the resulting executables to train and detect with an objectdetector. Basically traincascade is the evolved system based on haartraining, so I would just stick to that one if I was you.
More information can be found in documents
- Cascade Classification
- Cascade Classification Training
- Cascade Classification Example on Face Detection