隧道掌子面节理点云识别及微服务模块开发

Point Cloud Identification of Joints in Tunnel Faces and Its Implementation in a Microservice

快速识别岩体隧道掌子面不连续面的点云信息并解译其几何与力学参数是实现隧道远程诊断的重要基础。目前基于数码相片的三维点云识别严重依赖商业程序,本文基于计算机视觉开源框架,自主开发了隧道掌子面不连续面点云信息识别的微服务模块。该微服务模块基于Django框架封装,可灵活部署于任一具有微服务架构的隧道安全诊断平台中,根据用户在线输入的不同视角下隧道掌子面岩体相片,可自动识别三维点云并实现三维重构。该微服务模块已部署于同济大学基础设施智慧服务系统(iS3)中,点云识别结果与其他程序进行了对比,结果表明,该微服务模块能满足基本的三维重构功能要求,在点云识别速度方面具有优势,未来在计算精度方面仍存在提升空间。

Quickly identifying discontinuity of the rock in a tunnel face and interpreting its geometric and mechanical parameters are considered as important basis for realizing remote tunnel diagnosis.However,3D point cloud recognition based on digital photos often relies on commercial programs.Based on the open source framework of AliceVision,this paper developed a microservice module for point cloud information recognition of discontinuity of rocks in a tunnel face.Encapsulated with the Django framework,the module can be flexibly deployed in a platform that supports the microservice architecture.With photos of the rock mass on the tunnel face taken from different perspectives and uploaded remotely by users,the microservice helps automatically identify 3D point cloud and achieve 3D reconstruction.Currently,the microservice module has been deployed infrastructure Smart Service System(iS3)developed by Tongji University.The results compared with those obtained from other software show that the microservice module can meet the basic need for 3D construction with advantage in the computational speed although effort is still needed to further improve the accuracy.