An Examination of Different Vision based Approaches for Road Anomaly Detection

Recent advances in vehicular technology and sensing devices has led to the proliferation of different approaches for road surface condition monitoring and road anomaly detection. These approaches can be categorized into vision-based and accelerometer-based techniques. These techniques are designed to enable vehicles detect road anomalies and notify drivers of such, thereby reducing the rate of anomalous induced accidents. Also, these approaches can be incorporated into autonomous vehicles towards navigating through road terrains with anomalies such as potholes, speed bumps and rutting. The vision-based techniques have received wide acceptance among academia and industry players alike because of their inherent ability to detect road anomalies in real-time and notify drivers easily. In this regard, this paper presents a mini-survey of various state-of-the-art vision-based techniques proposed in the literature for road surface condition monitoring and anomaly detection. The merits and drawbacks of these techniques are highlighted. Furthermore, open research issues are presented. Our effort forms part of a larger goal aimed at developing robust vision processing approaches for road anomaly detection, as well as benefiting researchers involved in similar pursuits.