Research On Slope Deformation And Crack Monitoring Based On Computer Vision

With the development of economy and infrastructure construction,a large number of slopes have been produced in recent years.Some slopes have risk s of landslide,which threaten people’s life and property safety.Therefore,it is particularly important to realize effective health monitoring of slope.The existing methods mainly include remote sensing,GPS,ground penetrating radar,total station,manual inspection,etc.These methods have some problems,such as costly equipment,diseconomy,too much manual work required,and uncertain monitoring interval.The computer vision method effectively solves problems above.The computer vision method has the advantages of non-contact,less manual work,higher level of automation,higher monitoring efficiency,higher monitoring frequency and lower cost.Based on the above reasons,this paper focuses on the computer vision based slope deformation monitoring and crack identification methods,and proposes several slope monitoring methods that are able to monitor a variety of indicators via different kinds of hardware.The main work of this paper is as follows:(1)A new vertical target is designed for existed target-based computer vison slope deformation methods.It overcomes the magnification of feature extraction error which is caused by the rotation of the target in the process of slope 3d deformation.In this way,it improves the fault tolerance rate of feature extraction algorithm in the process of monitoring.The effect of six common template matching algorithms in this method is compared.An experiment considering three-dimensional deformation of slope is designed to verify the effect of the proposed method.(2)Target free slope deformation and displacement monitoring methods based on binocular vision and 3D laser scanning are proposed.The limitation of the position and number of monitoring aim-points is eliminated,and the global slope monitoring at any point in the view can be realized.It also eliminates the process of manual target installation and improves monitoring efficiency.The slope surface deformation prediction method based on LSTM is also applied,which has high prediction accuracy on the data set obtained in practical engineering.(3)The slope crack monitoring method based on binocular camera,3D laser scanner and UAV equipped with binocular camera is proposed to fill the research gap in this specific direction.A drone with binocular camera was modified.The identification of slope cracks using only single frame data is realized and verified by experiments.