Abstract:
With the continuous increase of constructions in highway, road maintenance
has become more and more important. Thus, it is of great significance to
develop the rapid, intelligent and real-time detection technologies for road surface
conditions. This paper used the self-developed driving data acquisitionAPP to collect
the vibration acceleration data during driving, and carried out the feasibility study on
the evaluation method of pavement rutting using smartphones. Firstly, the collected
vibration acceleration data are de-noised, and the vibration characteristics under
different working conditions are analyzed. Secondly, seven time-domain vibration
acceleration indexes with high correlation with pavement rutting are extracted, and
the dimensions of seven primary indexes are reduced to two independent principal
components by principal component analysis. Finally, the rutting evaluation model
based on convolutional neural network is established and compared with the results
of back propagation neural network and multilayer perceptron neural network. The
results show that the average relative error of the rutting evaluation model based on
the convolutional neural network is 16.6%, which is lower than the other twomodels.
It indicates that the pavement rutting can be evaluated satisfactorily by smartphones.
In addition, this paper divided the evaluation results of rutting into four grades (Excellent,
Good, Medium and Poor) and displayed them in different colors on the map.
This study is of great significance to improve the level of intelligent detection of road
rutting and road maintenance management.
