Method and device for improving robustness of visual inertial navigation system and robot equipment

Abstract

The invention relates to the field of simultaneous positioning and map construction integrating vision and inertial navigation, in particular to a method and a device for improving the robustness of avision inertial navigation system and robot equipment, and the method comprises the following steps: determining a reference frame; correcting the reference frame; obtaining a reference frame according to the oldest frame and the corrected reference frame; performing nonlinear optimization on the frames except the reference frame and the oldest frame according to the oldest frame and the corrected reference frame to introduce a new visual constraint into the system in a static or constant-speed state of the graph, so that the existing visual constraint and inertial constraint mechanism is optimized, the problem of track drifting is well solved, and the robustness of the system is remarkably improved.

Description

technical field [0001] Embodiments of the present invention relate to the field of simultaneous positioning and map construction of fusion of vision and inertial navigation, and in particular, to a method, device and robot equipment for improving the robustness of a visual-inertial navigation system. Background technique [0002] Simultaneous localization and mapping system (SLAM, Simultaneous localization and mapping, hereinafter referred to as SLAM) refers to the main body equipped with specific sensors, such as robots, drones, AR equipment, etc., in the absence of environmental prior information, in the movement process In the process of building a model of the environment and estimating the process of its own movement at the same time. The sensors in SLAM can be visual cameras, IMUs, GPS, lidar, etc. The sensors discussed in this article are mainly visual cameras and IMUs, which are usually called SLAM systems based on the fusion of visual and stargazing measurement unit...

AI Technical Summary

Problems solved by technology

However, when situations such as fast rotation, fast movement, camera occlusion, or no texture occur, visual constraints will be lost. This situation often occurs in AR, robots, etc. Take AR devices such as mobile phones as an example. When fast movement When , because of motion blur and other reasons, the image quality will be blurred. At this time, it will become unrealistic to perform visual feature point detection and tracking on such an image.

[0004] When the visual constraint is lost, only the IMU constraint works in the overall objective function of the VIO system, but because of the cumulative error of the IMU, the calculated state vector will be wrong, and when the vision is lost, the camera trajectory will still be speed up in a direction

Images