Prediction of International Roughness Index Based on Convolution Long Short Time Memory Network
The rapid development of highway brings the demand for rapid detection and analysis of various pavement indexes.According to the characteristics of international pavement roughness indexes,the combination of convolution neural network and long-term and short-term memory neural network(CNN-LSTM)is proposed to predict the international pavement roughness index-es.Convolution neural network and long-term and short-term memory neural network learn the spatial dimension of lidar range data respectively according to the characteristics of roughness and time dimension,the prediction of flatness index is completed.The ex-perimental results show that,compared with LSTM network,the MAPE value of CNN-LSTM model is only 2.3488,and the accura-cy and recall rate are 90.61%and 87.89%respectively.By comparing the real value with the predicted value,it can be found that CNN-LSTM is more suitable for the prediction of international roughness index.
long short memory neural networkinternational roughness predictionconvolutional neural networkpavement roughness
国家科技期刊平台
NSTL
万方数据
