| Self-navigation mobile robot has a wide range of applications.Web is an open,cross-platform and widely-used computing platform.If there is some project to make mobile robot SLAM applications running in the Web platform,that will greatly promote the development and promotion of robot SLAM applications.But now there is no such solution yet.In response to this problem,this paper aims to propose a complete SLAM scheme of mobile robot based on Web technologies.This paper firstly analyzed the development of SLAM algorithms and Web technologies in recent years,and selected ORB-SLAM2 as the prototype.This paper designed a set of programs based on Node.js to control mobile robot by JavaScript.Bound the data types and functions between C++ and JavaScript by using Node.js C++ plug-i.Implemented the multithreading sigslots mechanism in native driver by using libuv asynchronous I/O model.This paper optimized and expanded OpenCV.js.Optimized its compilation and function interface design.Optimized some of the memory management methods for custom data types,and designed the read and display interfaces for web images and videos by using HTML5 canvas and video elements and the WebRTC technology.Finally,this paper made OpenCV.js a complete,easy-to-use solution for image processing and visual computing in the web.Then this paper successfully transplanted ORB-SLAM2 into the Web based on OpenCV.js.Designed the method of loading and using the BoW model based on the Emscripten virtual file system,and optimized the loading speed of the file by using the binary format.Implemented the multi-threaded computing model in the Web by using Web Worker.This paper also tested and evaluated the proposed methods in this paper,and they are complete,cross-platform,easy-to-develop and with good performance.This paper brings a new development mode for SLAM developers and Web developers to develop SLAM applications and other visual computing applications on the Web platform,and has certain practical value. |
PDF Full Text Request