Research On The Integrated Above-water And Underwater 3D Reconstruction Technology Of Large Karst Caves With Fully Water-covered Botto

In recent years,with the innovative development of measurement equipment and technology,the research focus of surveying and mapping work has gradually shifted to accurate measurement of complex scenes such as mountains,forests,and underwater.As the country proposes the development strategy of "green waters and green mountains are golden mountains and silver mountains",protecting the ecological environment of karst caves and developing karst cave tourism resources have become particularly important.In tunnel construction,excavation operations often encounter disasters such as water and mud inrush caused by karst caves,rock mass collapse,and other disasters,which directly affect the safety of construction personnel.The cost of grouting materials is also difficult to estimate.Therefore,whether it is ecological protection and development of karst caves,or tunnel construction safety and grouting quantity estimation,there is an urgent need for surveying and mapping work to provide accurate information about the location,size,structure,and other aspects of karst caves.Since the layout lines and heat generation of lighting,signal base stations and other equipment will damage the stalactite,stalagmite and other natural resources inside the cave,the measurement must be carried out without light and signal.At the same time,because of the changeable internal structure of the naturally formed cave and the complex water system,the traditional measuring equipment such as total station and RTK cannot work normally in the cave,and the measurement of the karst cave has become one of the difficulties in the surveying and mapping field.In response to this issue,the following research work has been carried out in this article:(1)According to the self-designed data acquisition scheme,the collection of point cloud data above and under water in karst caves is completed using 3D laser scanners,multibeam echosounder,RTK,and other measuring equipment.Use a three-dimensional laser scanner to collect laser point cloud data from the water part,and use a multi beam echosounder to collect underwater structure point cloud data.Use RTK to arrange ground control points at the karst cave outlet and inlet for accuracy verification of point cloud data registration.(2)Aiming at the problem of partially reflecting point clouds in the laser point cloud collected by the 3D laser scanner due to reflected laser light from the water surface,a neighborhood search algorithm is used to calculate the k-neighborhood characteristics of each point for matching calculations.Points with low Y-axis values in point pairs with the same neighborhood characteristics are eliminated,and this point is called a reflection point.(3)Aiming at the problems of low accuracy and slow speed in registration between laser point cloud data from various stations on water due to the irregular surface of karst caves and weak geometric features,this paper uses the NARF+SIFT algorithm to extract key points with stronger expression ability,and extends the multi view convolutional neural network(MVCNN)to point cloud registration.A novel target disk iteration form is designed to solve the problem of limited number of MVCNN viewpoints.The proposed point cloud multi view convolutional neural network(PC-MVCNN)registration model improves the accuracy of point cloud data registration in karst caves,and invokes GPU to accelerate the calculation of matrix,improving the efficiency of point cloud registration.(4)In order to establish an integrated three-dimensional model,it is necessary to register the above water and underwater point clouds to obtain the overall point cloud data of the karst cave.However,GNSS cannot be used to obtain real-time location information inside the cave,and the point cloud data collected by the multibeam echosounder is still stored in the local coordinate system of the instrument.The point cloud data collected by a three-dimensional laser scanner can be converted to the CGCS2000 coordinate system using ground control points arranged at both ends of the cave.Therefore,it is necessary to register the underwater point cloud to the coordinate system of the underwater point cloud.In this paper,a shape context algorithm is used to extract the above water and underwater point clouds close to the water surface as template images and input images,respectively,and a similarity metric matrix is calculated for point cloud matching.This solves the problem of low overlap between the above water and underwater point cloud data that makes it impossible to register.(5)Aiming at the problem of holes in the 3D model created due to the sparse and missing point clouds in some regions,this paper uses the triangulation method to find the smallest area triangles on the edge points of the holes,and adds corresponding triangular patches to the model to gradually complete the hole repair work.Design experiments to verify the repair accuracy,and the results show that the error of the repair method is controlled within the required range.