CYCLOPS: A cyclists' orientation data acquisition system using RGB camera and inertial measurement units (IMU)

Area of Science:

• Robotics and Autonomous Systems
• Computer Vision
• Machine Learning

Background:

• Predicting cyclist intentions is crucial for urban road safety.
• Existing methods lack comprehensive data capture for cyclist behavior analysis.
• Machine learning models require robust datasets for accurate motion prediction.

Purpose of the Study:

• To introduce CYCLOPS, an original system for acquiring synchronized image and inertial measurement unit (IMU) data from moving cyclists and vehicles.
• To provide a dataset for training machine learning models to predict cyclist motion intentions.
• To detail the hardware and software design for a complete, open-source data acquisition system.

Main Methods:

• Developed a two-device system: one for the bicycle (IMU, microcontroller, transceiver) and one for the vehicle (IMU, microcontroller, RGB camera).
• Integrated components for real-time data acquisition and storage on a laptop.
• Detailed hardware architecture, device installation, IMU configuration, and software setup.

Main Results:

• Successfully developed and documented the CYCLOPS acquisition system.
• The system captures synchronized cyclist motion (acceleration, orientation) and vehicle data, including images.
• All design and software files are publicly available under a CC BY 4.0 license.

Conclusions:

• CYCLOPS provides a foundational, open-source tool for generating valuable datasets for cyclist intention prediction research.
• The system's design facilitates data collection for advancing AI in intelligent transportation systems.
• Availability of the system promotes further research and development in road safety and autonomous systems.