Robot pose estimation method, device and equipment and storage medium

Abstract

The invention discloses a robot pose estimation method. The method comprises the steps that an environment image collected by a target robot is acquired; instance segmentation is carried out on the environment image by using a Deep Snake instance segmentation network, and a moving object in the environment image is removed to obtain a static environment image; and pose estimation is carried out according to the static environment image, and the estimated pose of the target robot is determined. According to the method, the moving object can be removed more accurately, so that the accuracy of pose estimation of the robot is improved; and moreover, the Deep Snake instance segmentation network processes a molded line through a ring convolution kernel, so that the calculation amount can be reduced, and the real-time performance of pose estimation of the robot is improved. The invention further discloses a robot pose estimation device and equipment and a computer readable storage medium which are provided with the beneficial effects.

Description

technical field [0001] The present invention relates to the field of robot navigation and mapping, and in particular, to a method, device, device and computer-readable storage medium for estimating robot pose. Background technique [0002] SLAM (simultaneous localization and mapping, real-time localization and mapping) technology is one of the core technologies of robots. The robot can instantly determine its position and posture in the environment through the input data of sensors in the environment, and establish a kind of environment map. technology. Specifically, the robot collects the environment image through the camera during the movement, performs feature point extraction and inter-frame matching on the environment image, obtains the observation data of the coordinate changes of the key points, and establishes the functional relationship between the observation information and the robot pose, Use the optimization method to solve the extreme value of the function, an...

AI Technical Summary

Problems solved by technology

However, the main workflow of the instance segmentation technology in the prior art is: first detect the rectangular frame of the moving object, and then perform pixel-by-pixel segmentation along the boundary of the moving object in the rectangular frame; in this way, the division of the rectangular frame is inaccurate , the problem of rough object outlines makes the process of removing moving objects inaccurate, which affects the accuracy of pose estimation; in addition, the computational complexity of pixel-by-pixel segmentation in the prior art method is very large, making the position of the robot There is a delay in pose estimation, that is, the pose estimation of the robot cannot be performed in real time

Images