Mobile cleaning robot SLAM device and algorithm based on depth vision

Abstract

The present invention provides a mobile cleaning robot SLAM (Simultaneously Localization And Mapping) device and algorithm based on depth vision. The device comprises: an RGB-D sensor configured to collect RGB-D information comprising a RGB color image and a Depth information in an environment; and a rotation mechanism configured to perform 360-degree movement of the RGB-D sensor. The SLAM algorithm is improved compared to a traditional SLAM algorithm, the RGB-D sensor is swung at a fixed position, the multi-frame RGB-D information at a current position is fused, the cleaning robot is moved, athree-dimensional image of the whole environment is generated while optimization of the movement trace of the cleaning robot through SURF features to be used for automatic navigation of the cleaningrobot so as to achieve unmanned cruising work of the cleaning robot.

Description

technical field [0001] The invention relates to the field of image processing and establishment of three-dimensional maps, in particular to a SLAM device and algorithm for a mobile sweeping robot based on depth vision. Background technique [0002] In order to navigate in an unknown environment, the mobile sweeping robot needs to build an environmental map and simultaneously locate its position in the map. The process of solving these two problems at the same time is called simultaneous localization and map construction (Simultaneously Localization And Mapping, SLAM) ). When the sweeping robot is in an outdoor environment, this problem can be solved by high-precision GPS. However, when the sweeping robot is in an indoor environment, or when the GPS is not accurate enough to meet the high-precision requirements, or when the environment of the sweeping robot is confidential, people must use other methods to accurately estimate the position of the sweeping robot and simultaneo...

AI Technical Summary

Problems solved by technology

Firstly, the RGB-D sensor is used to obtain the color information and depth information of the environment, and the ORB feature is used to extract the image features; then, the pose estimation of the sweeping robot is realized through the RANSAC-ICP inter-frame registration algorithm, and the initial pose graph is constructed; Finally, BoVW (Bag of Visual Words) is constructed by extracting the ORB features in the KeyFrame key frame. The current key frame is compared with the words in BoVW to achieve closed-loop key frame detection, and the pose graph is added through key frame inter-frame registration detection. Constraints to get the globally optimal sweeping robot pose, but it does not solve the navigation problem of the sweeping robot, and how to generate a panoramic three-dimensional map after the sweeping robot moves

Images