Relocation and loopback detection method based on depth map coding

Abstract

The invention discloses a relocation and loopback detection method based on depth map coding. The relocation and loopback detection method utilizes a depth map shot by an RGB-D camera to combine the estimated camera pose to assist in judging loopback detection and repositioning, that is to utilize a random fern encoding method to encode the depth map, compares the encoding similarity of the current frame depth map with the encoding similarity of the historical key frame depth map while comparing the spatial position similarity of the camera when the two frames of images are shot by utilizing the camera poses estimated by the two frames of images, and judges whether the camera returns to the previous area or not according to the encoding similarity and the spatial position similarity of thecurrent frame and the historical frame. On the basis of an original method based on a color image, by combining the depth information brought by a novel sensor, the relocation and loopback detectionmethod provides more stable repositioning and loopback detection effects, and encodes the image, so that the information of the whole depth map is prevented from being considered, and the calculationamount is greatly reduced. Meanwhile, the camera pose is also considered, so that the problem of false positive judgment result is avoided.

Description

Technical field [0001] The invention belongs to the field of computer vision, and is especially aimed at SLAM (Instant Localization and Map Construction), and specifically relates to a depth map coding-based relocation and loop detection method. Background technique [0002] The general process of real-time positioning and map construction is to take images according to the camera in the device during the movement, and process them frame by frame, compare them between frames, and then estimate the relative changes in the position of the camera, and then estimate the camera accordingly. The (equipment) movement trajectory, and finally the newly observed frame updates the map according to the estimated pose. However, when the camera is in motion, it may cause tracking failure. If this situation is not detected, the map is updated with the wrong pose, and the existing map may even be destroyed. And if the tracking failure can be detected, then the modification of the map can be pre...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to improve the problem of increased or even failed relocation and loopback detection errors caused by the degradation of color picture information

Images