| As an important part of the energy system in our country,storage tank is widely used in the military,transportation,aviation and petrochemical industries.The regular detection of storage tank is a vital practical significance.The traditional storage tank inspection has high labor intensity,long inspection time,and the closed space is not allowed to work for a long time,so the inspection efficiency is low.Three-dimensional laser scanning technology,which has been developed rapidly in recent years,has been applied to the inspection of storage tanks with its characteristics of non-contact,high precision,digitalization and automation.However,the point cloud data obtained by 3D laser scanning has some problems,such as a large amount of data,a large number of redundant points and scattered data,which affect the subsequent 3D modeling and volume measurement.This paper carries out research from two aspects of tank point cloud simplification and tank point cloud modeling.The main work completed is as follows:1.A spherical model based RANSAC point cloud simplification method was proposed for large storage tanks through combining with uniform grid method and Random Sample Consensus(RANSAC)algorithm.In this method,the 3D point cloud data to be processed is divided into several small grids by using the uniform grid method;Then a ball model is established for point cloud data in each grid according to RANSAC algorithm,which is used as the standard for screening feature points and filtering redundant points;Finally,threshold values are selected to retain feature points and filter out redundant data points,thus achieving the purpose of simplifying point cloud.The experimental results show that the spherical model based RANSAC point cloud simplification method not only retains the high simplification rate characteristic of the uniform grid method,but also greatly improves the retention effect of the feature information of the point cloud data,avoiding the feature information loss and the appearance of data holes.2.Aiming at the problem of large randomness in initial value selection of MSAC(M-estimator SAmple Consensus)algorithm,a storage tank modeling method based on MSAC-LS is presented.In this method,the randomness of the initial value estimation of the MSAC algorithm is suppressed by setting threshold parameters,and the fitting result of the storage tank model is estimated by the least squares algorithm.The experimental results show that the storage tank modeling method based on MSAC-LS avoids the large fitting error caused by the large randomness of the MSAC algorithm while retaining the high accuracy of the MSAC algorithm in fitting the storage tank model;Moreover,the model still has high accuracy when modeling the simplified tank point cloud data and the tank point cloud data obtained under different sampling conditions. |
