Pothole Detection in Asphalt: an Automated Approach to Threshold Computation based on the Haar Wavelet Transform

With the increasing deployment of vehicles embedded technology and with the imminent availability of the autonomous vehicles on the market, solutions for detecting potholes in roads have gained attention from industry and academia. The current work proposes an automated system for pothole detection by the one dimensional Haar Wavelet Transform (HWT) applied to accelerometer signals. In this context, the proposed methodology explores the advantage of low cost processing in both stages, in the signal acquisition and during the analysis. The analysis of the wavelet coefficients is done through a two-step threshold procedure that enables the identification of strong variations within data, here related to the potholes. Since the threshold values are estimated adaptively, the detected variations can also identify the normal signal pattern, associated to the accepted road conditions. Thus, no manual threshold calibration is required. Overall, we found that our proposed methodology is efficient not only for a controlled environment scenario but also for a real scenario.