Robot multi-camera visual-inertial real-time positioning method and device

Abstract

The invention discloses a robot multi-camera visual-inertial real-time positioning method and device. The method comprises the steps that current multi-view images and inertial sensor data of a robotare obtained, according to the current images, image feature points are extracted, a current robot pose is estimated, according to the current robot pose, a 3D point cloud is rebuilt, historical and current point cloud data is stored to maintain a vision point cloud map, according to the inertial sensor data, initialization is completed, a sensor bias value is estimated, pre-integration is conducted to obtain the current speed and angle of the robot, according to the vision point cloud map and inertial sensor pre-integration, the current pose is optimized, and the like. Multi-view cameras mentioned in the robot multi-camera visual-inertial real-time positioning method provide a wider view by means of information of multiple view angles, the multiple cameras face differently, the situationthat all views are hindered is rare, and visual features provided by the multiple cameras are richer so that the requirements of features needed for achieving positioning can be almost met.

Description

technical field [0001] The invention relates to robot navigation technology, in particular to a multi-camera visual inertial real-time positioning method and device for a robot. Background technique [0002] At present, more and more different types of robots appear in all aspects of production and life. For the fields of warehousing and logistics, inspection and monitoring, etc., the work requires robots to be able to achieve long-term stable operation in a relatively fixed environment, and can realize Precise self-positioning. When the camera's field of view is blocked by obstacles, the visual features are not obvious, and the feature texture is highly repetitive and difficult to match, the event of positioning and tracking often occurs. In addition, when the robot moves so fast that the image is blurred, the existing pure vision positioning methods are also incompetent. Multi-eye cameras use information from multiple perspectives (overlapping or non-overlapping) to prov...

AI Technical Summary

Problems solved by technology

When the camera field of view is blocked by obstacles, the visual features are not obvious, and the feature texture is highly repetitive and difficult to match, the event of positioning and loss often occurs

In addition, when the robot moves so fast that the image is blurred, the existing pure visual positioning methods are also incompetent.

Images