PedCross: Pedestrian Crossing Prediction for Auto-Driving Bus

The topic of study, Intelligent Transportation System (ITS), focuses on using vehicle-to-environment communication to address severe traffic problems, such as safety and congestion difficulties. Pedestrians remain the most Vulnerable Road Users (VRUs). Moreover, pedestrian movements are more challenging to predict because humans can quickly change their direction and status (e.g., walking or stopping). In this research, we build a system called PedCross, which uses human image semantic information to predict the behavior of pedestrians (i.e., crossing or not crossing). In PedCross, images of pedestrians are first used to detect skeletons. The features in the detected skeletons are then extracted for model training. Two types of models, Random Forest and LSTM, are considered for pedestrian crossing prediction. To further improve the efficiency and accuracy of PedCross, Skip Frame, Head Orientation, and Warning/Dangerous Zones are integrated. PedCross is tested with the collected ITRI dataset and deployed on the auto-driving bus for a road test. The road test indicates that PedCross achieves all the requirements set forth by ITRI and outperforms Free Space, a baseline system developed by ITRI.