Design Of Visual Inertia Odometry Based On Nonlinear Optimization Algorithm

Visual Inertial Odometry is a SLAM technology that combines visual information and inertial measurement information.SLAM is an abbreviation of Simultaneous Localization and Mapping.The Chinese name is synchronous positioning and map construction.It is indispensable in the development and use of modern robots.A technology,with the continuous development of artificial intelligence and computer computing capabilities,the research of visual SLAM technology has become more and more popular in recent years.However,the scene where the visual SLAM works is still limited,and it only works in an environment with good lighting and obvious visual texture characteristics,and it is easy to produce tracking loss when the movement changes drastically.The IMU can measure the motion state of the target through the principle of inertia,and complements the visual information to a certain extent.By fusing the two sensor information together,a more accurate and robust SALM system can be obtained.Based on this,this paper designs a visual inertia mileage calculation method based on optimization algorithm,and applies it to the embedded development platform.(1)A visual odometer using a semi-direct method is designed.The key points of FAST are detected and extracted in continuous input image frames and tracked using the KLT optical flow method to restore the camera's posture and obtain the camera odometer data.And the algorithm process is analyzed,and a parallel acceleration method based on ARM processor NEON is proposed to improve the algorithm running speed.And the GRIC selection criteria are cited,which can correctly select the computer vision epipolar geometry model for camera pose recovery according to the characteristics of the key points correctly matched,which increases the robustness of the algorithm.(2)Analyze the related theories of SLAM system and deduce the related formulas,including: the derivation of the related theories of computer vision epipolar geometry,the derivation of the related formulas of IMU pre-integration,the derivation of the transformation between different navigation coordinate systems,the derivation of the objective function and constraint conditions in the optimization algorithm,And the derivation of Jacobian matrix when solving the objective function.(3)A complete set of visual inertial odometer system is designed,and a visually inertial tight coupling scheme based on optimization algorithm is proposed.The visual measurement constraints and IMU measurement constraints in VIO are discussed in detail.The objective function of linear optimization,and introduces the strategy of sliding window marginalization in the optimization algorithm.Finally,a loop detection method based on the bag of words model is adopted.By detecting the loop frame,the relocation function is realized,and the loop is constrained.It is added to the back-end optimization to reduce the cumulative error,eliminate trajectory drift,and perform a globally consistent pose optimization.(4)Finally,the VIO algorithm designed in this paper is experimentally verified,and the visual front-end acceleration method designed in this paper is valid and feasible.The public data set Eu Ro C data set is used to verify the overall algorithm.The experimental results show that the algorithm designed in this paper is effective and accurate In the traditional visual inertial SLAM system based on filtering algorithm OKVIS,it can meet the real-time requirements of intelligent mobile terminals.