This research focuses on reidentifying vehicles between advance and stop-bar detectors at signalized intersections using high-resolution traffic data. An optimization framework, proposed on top of ML models that predict the travel time from advance to stop-bar locations, is used to evaluate the accuracy of correct match pairs between the two detectors. The ML models are trained using semi-ground-truth data, while the accuracy of the match pairs are tested on video-verified ground-truth data.
- preprocess_training_data.py
- process_events.py
- generate_candidate_matches.py
- match_pairs_train_dataset.py
- process_match_pairs.py
- feature_extraction.py
- ML_models.py