Mobile robot positioning method and system based on instance segmentation and multi-sensor fusion, and medium

Abstract

The invention discloses a mobile robot positioning method and system based on instance segmentation and multi-sensor fusion, and a medium. The method comprises the following steps: carrying out target instance segmentation on a camera image; extracting and tracking feature points of the segmented image; calculating a depth value of the feature point and carrying out pose calculation by using an EPnP algorithm; carrying out pose calculation based on IMU pre-integration; selecting an updating mode of the pose initial value according to the instance segmentation information and the feature point extraction condition; and enabling the visual sensor and the IMU to be subjected to tight coupling pose optimization, and obtaining an accurate pose. According to the method, the instance segmentation information of the image is acquired, the static feature points in the complex environment are selected as the reference, the pose state of the robot is estimated, and the IMU can be switched to update the positioning initial value information when the camera positioning is not ideal, so that the positioning of the robot also has accuracy and robustness in a highly dynamic environment.

Description

Technical field[0001]The present invention belongs to the field of intelligent mobile robots, and in particular to a positioning method, system, and medium based on an example segmentation and a multi-sensor fusion.Background technique[0002]The original intention of the robot is to help humans complete some trivial, service, mechanical, and even certain dangerous tasks, making people's lives easier. Since the century, robot technology has developed rapidly. The robots of various types of various types begin to appear in people's lives, such as patrol robots, sweeping robots, survey robots, express distribution robots, etc. The mobile robot acts as a large class in the robot, and many cases need to complete its own tasks during the move. This requires that they need to identify the surrounding environments like human beings, and autonomously navigate according to environmental information. Among them, it is the basis for solving the autonomous navigation problem of mobile robots in t...

AI Technical Summary

Problems solved by technology

However, most of the visual positioning methods assume that the environment is a rigid body that does not change. When a moving object appears in the image, mistakenly using the features on the object as the positioning reference will cause serious damage to the positioning of the robot. Significant errors, even lead to positioning failure in highly dynamic environments

Images