Farneback Optical Flow Application for Automotive Onboard Vibration Monitoring

This research explores a vision-based vibration monitoring algorithm that may be used to detect and measure vibrations on a vehicle with an onboard camera. To achieve this, we explored Optical Flow algorithms techniques such as Lucas-Kanade, Horn-Schunck and Farneback to analyze pixel motion between frames and relate it to vibrations. For testing and validation, CARLA simulator was used to acquire the front camera image and IMU data in scenarios where the vehicle drives over road bumps and potholes. The visual vibration estimation was compared with the z-axis acceleration data of an IMU and the results shows that it is possible to use optical flow methods to detect vibration on the vehicle.