Positioning method and device adopting visual inertial data deep fusion

Abstract

The invention discloses a positioning method and device adopting visual inertial data deep fusion, wherein the method comprises the following steps of: S1: enabling a system to be in a static state atan initial moment, and acquiring a measured value of an accelerometer and a measured value of a gyroscope in an initialization time period so as to estimate an initial state of the system; S2: afterobtaining the initial state, carrying out propagation on a system state according to the measured value of the accelerometer and the measured value of the gyroscope, and carrying out updating on a covariance matrix of the system; S3: after acquiring an image, tracking feature points by utilizing an IMU (Inertial Measurement Unit) assisted external point elimination method; and S4: for feature points failed in tracking, constructing visual measurement according to visual measurement information, and carrying out updating on the system state. The method sufficiently utilizes various sensor datain a visual inertial fusion system, can effectively utilize IMU data to promote a tracking effect and efficiency, and can accurate estimate a pose of the system in real time.

Description

technical field [0001] The invention relates to the technical field of navigation and positioning of unmanned aerial vehicles, in particular to a positioning method and device for deep fusion of visual-inertial data. Background technique [0002] Due to its small size, high maneuverability and convenient hovering, UAV is becoming an ideal platform for monitoring, exploration, rescue and other tasks. [0003] To solve the problem of autonomous navigation of UAVs without GPS (Global Positioning System, Global Positioning System), the main problems that need to be faced include: environment perception, state estimation, mission planning, flight control, etc. In this series of problems, the state estimation function, which is often called the positioning function, is the most basic and arguably the most important. On the one hand, due to the high maneuverability of the UAV and the unknown environment, we must determine the The position and attitude of the aircraft, on the other...

AI Technical Summary

Problems solved by technology

[0004] Inertial navigation equipment has a high status update frequency, is sensitive to attitude angles, and can handle large rotations well. It is accurate enough in a short period of time and is less affected by the environment, and can restore absolute scale information. However, its position information is obtained through acceleration integration. , there is cumulative error and cannot be eliminated by itself; camera-based visual estimation is more suitable for long linear motion scenes and can eliminate cumulative error through visual association, but its estimation effect for rotating scenes is poor, there is scale uncertainty and Susceptible to the influence of light, texture, etc. in the environment

Images