Robot positioning method and system fusing monocular vision and encoder information

Abstract

The invention relates to a robot positioning method and a system with monocular vision and encoder data fused, which comprise real-time calibration on external parameters and visual scale factors based on a sliding window method and pose loose coupling fusion based on the sliding window method. According to the real-time calibration on the external parameters and the visual scale factors based on the sliding window method, the monocular vision and the pose information outputted by the encoder are used to build a sliding window pose diagram, the relative pose between a vision sensor and the encoder is automatically estimated, and the scale factor for the monocular vision is calculated. The built sliding window pose diagram is used to fuse the monocular vision positioning information and the encoder positioning information, problems of poor precision of long-time positioning of the encoder and poor robustness of vision positioning are overcome, the method based on the sliding window pose diagram can effectively ensure the calculation amount of the fusion method is not added along with time, and the algorithm is thus applicable to mobile robot embedded equipment.

Description

technical field [0001] The invention relates to the technical field of mobile robot positioning, in particular to a robot positioning method and system that integrates monocular vision and encoder information. Background technique [0002] With the continuous development of robot technology, people have more and more requirements for service robots to be able to locate and navigate autonomously in the natural environment. Using the encoder to calculate the odometer according to the robot kinematics model is a commonly used method in the field of robotics, but there are problems that the cumulative error cannot be eliminated and the positioning accuracy is poor for a long time. The use of monocular vision for positioning has developed rapidly recently, and visual odometry calculation methods such as svo (semi direct visual odometry) and dso (direct sparse odometry) have appeared one after another. It is characterized by small cumulative error and high positioning accuracy, b...

AI Technical Summary

Problems solved by technology

However, the computational complexity of the pose graph will increase as the number of nodes added increases, which is not conducive to real-time calculations on robot embedded devices.

Images