Method and device for real-time global optimization and error loopback judgment in three-dimensional reconstruction

Abstract

The invention discloses a method and a device for real-time global optimization and error loopback judgment in three-dimensional reconstruction, and the method comprises the following steps: obtainingvisual information and IMU information, and constructing an optimization equation through the secondary statistics of the visual information and the pre-score of the IMU information, so as to carry out the real-time online global optimization of a CPU; coupling a transformation matrix in the global optimization function, and providing a state estimator in the global optimization process, so thatmapping continues to be carried out when rapid movement is carried out and the number of feature points is small; and obtaining an estimator of IMU zero offset, and identifying an error loop and a correct loop according to the estimator of IMU zero offset. According to the method, high-efficiency global optimization of vision and IMU information can be realized on the premise of not losing precision, the problem of accumulative errors caused by inaccurate calibration is solved, the loop-back correctness can be accurately judged, and the method is simple and easy to realize.

Description

technical field [0001] The invention relates to the technical field of camera positioning and mapping, in particular to a method and device for real-time global optimization and error loopback judgment in three-dimensional reconstruction. Background technique [0002] With the development and wide application of robot indoor navigation, virtual reality and augmented reality technology, 3D reconstruction requires higher and higher precision and efficiency, and needs to deal with various environments and motion states to achieve a more robust reconstruction process. What 3D reconstruction needs most is SLAM (Simultaneous Localization And Mapping, Simultaneous Localization and Mapping) technology based on visual positioning, which can accurately calculate the pose of the camera when the feature point information is rich, and achieve global consistency through loop detection. When there are few moving and feature points, the visual tracking of the camera tends to fail. At this t...

AI Technical Summary

Problems solved by technology

The transformation matrix obtained in this way is not accurate enough, and calibration errors will accumulate;

[0006] 3. For loop detection: loop detection only through visual information is prone to misjudgment, damages global consistency, and makes the results of 3D reconstruction confusing

Images