Research On Combined Navigation And Positioning Of Vision And Other Sensors

With the rapid development of computer science and sensor technology,indoor service robots occupy more and more positions in people's life and work.Multi-sensor fusion navigation and positioning technology has become a research hotspot in the field of indoor robot navigation and positioning in recent years due to its excellent performance and widely used scenarios.It has great research value and practical significance in the field of indoor robot navigation and positioning.This paper studies several key technologies based on fusion navigation of vision and other sensors,proposes two main sensor fusion technologies and designs and implements related systems.One is the fusion navigation and positioning technology of binocular vision and inertial measurement unit(IMU)based on tightly coupled optimization.This method first obtains the real-time system by establishing inertial prediction constraints and image observation constraints,and then iteratively optimizing them.State;the other is a three-sensor fusion navigation method of visual/inertial measurement unit/3D lidar based on federal filtering.By combining the inertial measurement unit with 3D lidar and camera,two sub-filters are formed.The local optimal results of the filter are fused,and the optimal combination result is finally obtained.The main research results of this article are summarized as follows:First,a fusion navigation technology based on tightly coupled binocular vision and inertial measurement units is studied,and an improved initialization method of fusion navigation system is proposed.Based on this method,a binocular vision and inertial measurement unit fusion navigation and positioning system based on this method was implemented.An experiment was designed and verified in an indoor environment.The experimental results show that the system can achieve high positioning accuracy.For indoor scenes with high accuracy requirements,the system can be fully applied.Secondly,a three-sensor integrated navigation and positioning technology based on federated filter is studied.It is mainly divided into three steps:(1)According to the needs of indoor experimental scenes,the appropriate visual SLAM algorithm is selected as the front end of the sensor fusion algorithm;(2)In the scale-invariant feature transform(SIFT)algorithm The technology is improved,and a three-dimensional point cloud feature extraction method based on SIFT algorithm is proposed.On this basis,the initial value is quickly obtained by using feature points to participate in the initial registration,and the initial value is used to fine-tune the source point cloud,and then combined with the Iterative Closest Point(ICP)fine registration to improve the registration accuracy and speed(3)Design and implement a three-sensor integrated navigation and positioning framework based on federal filtering,and design experiments to verify.