Research On Indoor Pose Measurement Technology Of Multi-camera System Based On Circumferential Vision

Indoor pose measurement technology develops rapidly in industrial manufacturing and our daily life.Among them,multi-camera system based on circumferential vision has obvious advantages in operation difficulty,environmental adaptability in complex scenes,research and manufacturing costs and measurement accuracy.However,in the process of pose measurement of multi-camera system based on circumferential vision,the observation information of different cameras is difficult to be used simultaneously,and it is difficult to construct the optimization model of multi-camera structure.In this paper,the problem of multi-camera pose measurement based on circumferential vision and the problem of pose optimization based on known constraints are studied.Feasible algorithms are proposed and implemented,and multi-camera parallel acquisition system is built to verify the algorithm.In the actual measurement process,the multi-camera system based on circumferential vision obtains the position and attitude information by using the multi-camera observation information at the same time,and carries on the non-linear optimization to the pose,realizes the high precision and high robustness position and attitude measurement.The main contents of this paper are as follows:1.Starting with the measurement model of the multi-camera system,the non-central circumferential imaging measurement model which can simultaneously utilize the multi-camera observation information is analyzed in depth.On this basis,the algorithm principle of the gPnP algorithm is expounded,the principle of Random Sample Consistency is introduced,and the multi-camera spatial localization algorithm based on Opengv is realized.2.The extended representation method of coordinate system using Lie group and Lie algebra is proposed,the algorithm principle of nonlinear optimization is expounded,the objective function of nonlinear optimization based on re-projection error is designed,the descending Jacobian matrix based on left disturbance model is deduced,and finally the descending Jacobian matrix is established.A graph optimization model based on g2 o is implemented to realize the nonlinear optimization algorithm.3.To meet the requirements of the multi-camera system,a circumferential image acquisition system based on raspberry pie is built,which realizes the acquisition of image and the parallel transmission based on Socket interface.4.The feasibility,measurement accuracy and consumption time of the space positioning algorithm plus pose optimization algorithm are evaluated by simulation system.The V-STARS system is used to build the three-dimensional control field in the laboratory environment.The comparative experiments are carried out with the help of the multi-camera acquisition system and the laser tracker T-mac six-degree-of-freedom measurement system.The results show that the proposed algorithm is effective.Effectiveness and measurement accuracy,and further improvement proposals are proposed for the experimental results.