A Tightly Coupled Initialization Method for Underwater Visual Inertial Navigation Pressure Positioning

Abstract

The invention belongs to the field of robot positioning, and in particular relates to a tightly coupled initialization method for underwater visual inertial navigation pressure positioning. The sub-scale robot pose and map feature points are obtained through the traditional monocular SLAM method, and the positions between adjacent images are obtained. Pre-integrate the IMU data to establish the pre-integration residual between the images, and pre-integrate the IMU data between the adjacent image and the pressure measurement to establish the pre-integration residual between the image and the pressure measurement, and solve it by nonlinear optimization method The initialization parameters of the system, use the initialization parameters to update the map, optimize the updated system by beam adjustment, complete the initialization process, and obtain more accurate results, so as to promote the operation of the system. The invention utilizes high-frequency IMU information to couple image information and pressure information at different time steps, strengthens the coupling degree between initialization parameters, and improves the solution accuracy of initialization algorithms.

Description

technical field [0001] The invention belongs to the field of robot positioning, and in particular relates to a tightly coupled initialization method for underwater visual inertial navigation pressure positioning. Background technique [0002] In the positioning of underwater robots, multi-sensor fusion positioning based on graph optimization is a common method, and such methods often require proper initialization to promote the normal operation of the system. For the fusion of visual inertial navigation and pressure, it is necessary to determine the parameters such as the initial scale of the system, IMU bias and initial attitude. Since the time stamps of the camera, IMU and pressure sensor are often not synchronized, and the frequency of the camera and pressure sensor is low, a common method is to pre-integrate the IMU between two frames of images to establish a pre-integration residual factor, and use the pressure sensor information Associated with adjacent image frames t...

AI Technical Summary

Problems solved by technology

Obviously, due to the asynchronous time stamp, the pressure measurement associated with the image information is not the most accurate measurement value at that moment, so this method solves a suboptimal solution, resulting in a large error in the initialization parameters of the final solution, while The error of the initialization parameters will largely affect the subsequent positioning accuracy of the robot

For robot applications that need to rely on robot positioning for subsequent high-precision navigation, this kind of error is not allowed

Images