Research On Key Technologies Of Virtual Assembly Software Platform

Currently,the assembly of industrial products mainly relies on manual guidance,and the assembly of various electromechanical products depends on the assembly experience of technicians.Operators need to undergo professional technical training before they can guide the assembly.This manual training and guidance method for assembly consumes a lot of labor,has low work efficiency,low standardization level,long work cycle,and high assembly cost.Virtual assembly technology fully utilizes three-dimensional visualization design software to repeatedly verify the position relationship of components and parts during the virtual assembly process.With the continuous development and application of computer technology,virtual assembly technology based on augmented reality has gradually become the mainstream of research and development.This thesis uses the ORB-SLAM system as the framework and studies a virtual assembly system based on monocular vision,and develops a virtual assembly software platform based on augmented reality technology.The main research contents are as follows:1.Research on key technologies of virtual assembly software platform.The key technologies of the ORB-SLAM system for virtual assembly are analyzed.To solve the problem of distortion in the image acquisition of monocular cameras,the internal parameter matrix and distortion coefficient of the camera are processed using Open CV library functions.In response to possible interference in image acquisition devices,image enhancement processing is proposed to make the image easy to acquire,easy to transmit,and facilitate subsequent image processing through filtering and grayscale transformation.2.Improved feature extraction and matching algorithm.The principles and main problems of traditional ORB algorithm are expounded and improved.To address the fixed threshold problem in the o FAST algorithm,this thesis proposes to extract feature points by calculating image contrast to set dynamic thresholds and using a method that combines local optimal feature points with Harris response values to filter feature points.The matching algorithm is analyzed,and matching based on FLANN is used instead of the matching method in the SLAM system.RANSAC algorithm is used to screen out mis-matching,and the improved algorithm is verified to improve accuracy while satisfying real-time performance.3.Improvement of tracking and positioning and map construction technology in ORB-SLAM system.According to the application scenario of this thesis,the essential matrix is selected to solve the camera pose change,and depth information is restored through triangulation.A new keyframe extraction strategy is proposed to extract features only from keyframes,greatly reducing the operating time.Li algebra Euclidean distance is used to prevent tracking failure due to excessive translation or rotation.The key technologies of Open GL model rendering and three-dimensional animation are explained,laying a foundation for the construction of virtual assembly software platform.4.Taking pipeline cleaning robot as the assembly object,the virtual model is designed and modeled,and a virtual assembly training software platform based on augmented reality technology is constructed,which realizes functions such as three-dimensional visualization,human-computer interaction,and multi-angle assembly,verifying the feasibility and effectiveness of the virtual assembly software platform constructed based on visual SLAM.