Road Type Classification from Tire Vibration Spectrograms using Time Frequency Representations

Abstract: This study offers a novel method for identifying the kind of road by utilising tyre vibration patterns and related sensor data that have been subjected to machine learning analysis. In order to create a classification system that can correctly recognise six different types of roads—asphalt, concrete, gravel, sand, mud, and cobblestone—we record and process signals from vehicle tires as they interact with various road surfaces. To achieve high accuracy in road surface identification, our system combines random forest classification with variables including dominant frequency, pressure variation, vibration magnitude, and statistical characteristics of the collected signals. According to experimental data, the classification accuracy is over 90% in a range of vehicle speeds and road conditions. Despite changes in tyre pressure, vehicle speed, and environmental factors, the system exhibits exceptional resilience. In order to create this trustworthy road surface identification system, this study describes the data gathering strategies, feature extraction tactics, classification algorithms, and validation processes. The suggested method uses sensors that are frequently found in contemporary cars to provide real-time information regarding road surface conditions, which has important ramifications for enhancing autonomous driving capabilities, intelligent transportation infrastructure, and vehicle safety systems. Keywords: Road Surface Detection, Tire Vibration Spectrograms, Sensor Data, Vehicle Speed, Tyre Pressure, Asphalt, Concrete, Gravel, Sand, Mud, Cobblestone.