Revolutionizing Road Surface Classification: Integrating Vibration Data and MCNN for Enhanced Pothole Detection

The existing techniques for categorizing road surfaces mainly rely on visual inspection or image analysis whose reliability and performance are limited in ugly climate ones. The unique contribution of the study is a proposed learning based novel solution that combines accelerometer-based vibration data fed into multiple channels of features to accurately identify potholes and rigid surface irregularities via multi-channel neural network (MCNN). Using motion responses detected by the accelerometer in conjunction with live imagery taken from a Raspberry Pi camera, the system provides accommodating detection of variations under different scenarios. Testing reveals the proposed configuration attains a precision of $\mathbf{9 4. 5 \%}$ and an exceptionally low false positive rate of only $\mathbf{5. 4 \%}$, handily outperforming past solutions. The creative integration of diverse inputs not only strengthens detection capabilities but also enables quick processing, permitting immediate feedback to optimize upkeep, fundamentally changing infrastructure management and ensuring enhanced safety for all road users