In this paper, an intelligent tire algorithm based on back propagation neural network was proposed to identify the type of road peak adhesion coefficient. In the drum test, eight different vehicle conditions were taken into account for each of three simulated road conditions. Tire pressure of 900 or 1000 kPa, speed of 30 or 40 km/h, and load of 500 or 1000 kg were the parameters for the vehicle conditions. The corresponding tire x-acceleration signal was obtained using the tire acceleration sensor. Then, the original signal data was denoised using wavelet transform. The signal data was grouped by tire rotation period and the anomalous data was eliminated using the isolated forest algorithm. Following that, wavelet transform was used once more to extract the time domain and frequency domain characteristics from each batch of data, producing 45-dimensional feature parameters. Principal components analysis was applied to reduce the dimension of feature parameters to obtain 8-dimensional principal components. Finally, to identify the type of road peak adhesion coefficient, the normalized principal components and vehicle condition parameters (tire pressure, speed, load) were used as input to build a four-layer back propagation neural network, while the adam algorithm was used to optimize the learning rate. 10-fold cross validation was used to examine the model effect. The prediction results show that the model has a high identification accuracy for three types of road, with an average accuracy of 99.53%. The prediction accuracy of the models developed in 10-fold cross-validation fluctuates between ±0.2%. Under different vehicle conditions, the proposed road identification perception algorithm establishes the direct correlation between tire x-axis acceleration and road adhesion coefficient. With its high dependability and real-time capabilities, this algorithm has a wide range of possible applications in the areas of automated driving and intelligent tires.