ABSTRACT

This paper proposes a novel pavement transverse crack detection model based on time–frequency analysis and convolutional neural networks. The accelerometer and smartphone installed in the vehicle collect the vibration response between the wheel and the road, such as pavement transverse cracks, manholes, and normal pavement. Since the original vibration signal can only contain a one-dimensional domain (time–acceleration). Time–frequency analysis, including Short-Time Fourier Transform and Wavelet Transform, can transfer the one-dimensional vibration signal into a two-dimensional time–frequency-energy spectrum matrix. The energy spectrum matrix obtained from STFT and WT can effectively obtain different signal features in terms of time and frequency features. If STFT and WT are further combined with CNN models, STFT-CNN and WT-CNN, respectively, pavement transverse cracks can be detected more accurately. In this study, the reliability of the developed pavement transverse cracks detection model was evaluated based on the data collected by conducting a road driving test. Analysis results of the developed model show that the accuracies of WT-CNN and STFT-CNN are 97.2% and 91.4%, respectively. The F1 scores to analyse the practicability and the adaptability of the crack detection model of WT-CNN and STFT-CNN are 96.35% and 89.56%, respectively.

摘要

本文提出了一种基于时频分析和卷积神经网络的新型路面横向裂缝检测模型。安装在车辆上的加速度计和智能手机收集车轮与路面之间的振动响应，例如路面横向裂缝、检修孔和正常路面。由于原始振动信号只能包含一维域（时间-加速度）。时频分析，包括短时傅里叶变换和小波变换，可以将一维振动信号转化为二维时频能量谱矩阵。从STFT和WT得到的能谱矩阵可以有效地从时间和频率特征上得到不同的信号特征。如果将STFT和WT与CNN模型进一步结合，分别为STFT-CNN和WT-CNN，可以更准确地检测路面横向裂缝。在这项研究中，根据通过道路行驶试验收集的数据，评估了所开发的路面横向裂缝检测模型的可靠性。所开发模型的分析结果表明，WT-CNN 和 STFT-CNN 的准确率分别为 97.2% 和 91.4%。分析WT-CNN和STFT-CNN裂缝检测模型的实用性和适应性的F1分数分别为96.35%和89.56%。