Vibration vs. vision: best approach for automated pavement distress detection

Abstract

Roads are the largest component of infrastructure; they directly impact people’s life by providing mobility and connectivity. To ensure consistent surface quality, roads must be monitored continuously and repaired when necessary. Presently, authorities spend substantial amount of time, finance and labor for pavement distress detection by employing traditional manual and instrumented methods which are generally tedious, and time-consuming. To overcome these drawbacks, various automated techniques like Ground Penetrating Radar, Laser-Imaging-Systems, etc. are deployed. Recently, image-processing and smartphone-based systems are being devised for pavement distress detection. Here, a vibration-based method using smartphone accelerometer and gyroscope, and a vision-based method using video processing for automated pavement distress detection are designed and compared to identify the more suitable one. Both experiments are performed on same roads and results are validated by manual surveying. Accuracy of vibration-based method for detecting potholes, patches and bumps is found as 80%. Accuracy for detecting cracks, potholes and patches using vision-based method is identified as 84%. An additional effort is taken to estimate the extent of pavement distresses using vision-based approach and validate it using manual stripping method. The study reveals that, vibration-based-analysis is sufficient for routine monitoring purposes whereas vision-based-method is more appropriate for detailed analysis.