Effect of combining algorithms in smartphone based pothole detection

Pavement surface condition monitoring plays a crucial role in ensuring comfort and safety to drivers and pedestrians. Furthermore, detailed information on existing conditions of pavement is vital to pavement managers in guaranteeing adequate maintenance. Presently, the concerned authorities expend a tremendous amount of time, finance and labor by employing the traditional methods for pavement distress detection. Potholes are one of the most common distresses in flexible pavement as they can lead to trip and fall accidents and are hence a liability for both public and private properties. The current work describes a novel method for detecting potholes by deploying smartphone accelerometers. Various researchers have formulated algorithms for detecting potholes using smartphone sensors. However, due to factors such as speed, extent of the distress, etc., individually these algorithms fail to provide results with desired accuracy. The present paper focuses specifically on the sensing component, reorientation of the smartphone-accelerometer with respect to the axes of the vehicle in which the smartphone is fixed, pothole detection algorithms and its threshold values and most importantly, identifying the best combination and threshold values of different algorithms to upgrade the results. The accuracy of the smartphone-accelerometers is validated with an external tri-axis accelerometer. The result obtained from smartphone-based method was compared with simultaneously collected videos of the road stretch. The efficiency of the method is affirmed again by calculating the true-positive and false-positive values. The ideal combination of algorithms and its threshold values suggested in this paper exhibit a true-positive of 93.18% and false-positive of 20%.