Driving video recorders (DVRs) are recently in general use, and can be expected to become vehicle necessities. Due to the widespread deployment and high mobility, DVRs are practical to monitor the conditions of road surfaces. Hence, we propose a DVR-based road bump detection technique, and employ clustering to reduce false alarms caused by image noises. Involved with GPS data, detected road bumps can be labeled on a map, which can be sent to the municipality for the reference of road repair or inform drivers to drive more carefully. Under a normal lighting environment, the precision of bump detection can be higher than 80%, and the computation speed can be up to 300 fps for the image resolution of 960´480.
Fig (a) Analyzing the driving video for vehicle shaking detection. (b) Road bumps labeled on a map with different colors according to the vehicle shaking amplitudes.