Pavement Defect Detection Based On Vehicle-mounted Mobile Devices Vibration Signal

In recent years,highways as one of the infrastructure projects in China have played a very important role in improving people’s lives and promoting the development of various fields in the country.With the passage of time,the early completion of the highways due to the limitations of construction technology,the impact of natural conditions and high load use of the highways,resulting in different degrees of damage to the highway surface structure,and eventually in the form of potholes.In order to ensure the safety,stability and comfort of highway traffic,highway surface maintenance work needs to be carried out in a timely and effectively manner.However,due to the huge number of highway mileage in China,Manual inspection of highway surface potholes is inefficient,The automatic detection method based on video images requires the installation of specific equipment,and the huge amount of video data and the high cost of transmission,storage and analysis cannot meet the basic requirements of fast and efficient road maintenance work in China.Faced with all these challenges,this paper proposes a highway surface defect detection method based on the vibration signal of vehicle-mounted mobile devices,aiming to speed up the defect detection efficiency and improve the safety of road driving.The main research contents are as follows:(1)Independent research and development of data collection APP and construction of highway surface defect vibration signal dataset.Due to the relative novelty of the subject,no datasets related to highway surface defect vibration signal was found.Therefore,it was decided to develop vibration signal collection APP independently using mobile device acceleration sensor combined with GPS positioning system,camera,Gaode offline map and other related technologies,and complete the data collection work with ordinary cars.The collected raw vibration signals are processed through labeling,extraction,data cleaning,data filling to build a high-quality vibration signal dataset for highway surface defects.In this paper,we use the mean and variance in statistics and draw the data change curve to analyze the statistical characteristics and actual regular characteristics of each type of data in the dataset respectively.(2)The vibration signal defect detection model based on the fusion of temporal and morphological features is proposed.According to the characteristics of each type of data in the dataset,an algorithmic model suitable for dealing with the problems related to this topic is searched.In order to solve the problem that the existing models extract single features of temporal data and the existing networks do not perform well on the dataset constructed by this paper,we propose LCN(LSTM and CNN’s Networks,LCN)and TCN(Transformer and CNN’s Networks,TCN),which are multi-feature fusion networks.Both LCN and TCN networks use a two-branch structure,where the LSTM/Transformer branch mainly extracts the correlation features between data,and the CNN-Inception branch mainly extracts the overall morphological features of data,and adds the channel attention mechanism to the CNN-Inception branch to improve the attention of important channels.Finally,the features extracted by the two branches are fused to improve the completeness and comprehensiveness of the features.The experimental results prove that the model built in this paper works better on the self-built dataset compared with the reference model,and the overall average improvement in classification accuracy is more than4%.(3)In order to make the model better implemented,the ”Highway surface Defect Intelligent Detection System” is designed and implemented.The system integrates the core functions of highway surface vibration signal upload and retrieval,vibration signal visualization,vibration signal pre-analysis,highway surface defect detection and visualization,and highway surface defect statistics and visualization.It realizes one-stop and quick detection of highway surface defects,and lays a solid foundation for later maintenance work.