Dictionary optimization method of mobile robot synchronous positioning and map construction system

Abstract

The invention discloses a dictionary optimization method of a mobile robot synchronous positioning and map construction system, comprising the steps of changing a dictionary format and training a small-scale dictionary, and characterized by changing the dictionary format by firstly reading a Text dictionary, restoring the original data type, and then calling a function to be saved in a binary format. The training of the small-scale dictionary comprises that at first, that ORB feature point are extracted, and then all the features are processed with K Means + + clustering, when the number of clustering layers of dictionary tree reaches the requirement, create leaf nodes are created and weights are set to complete the construction of dictionary. The dictionary format is changed and the small-scale dictionary is trained, so that the method of the invention can accelerate the start-up speed of the mobile robot system and keep the performance without loss on the basis of the lightweight system.

Description

technical field [0001] The invention relates to the field of mobile robots, in particular to a dictionary optimization method for a mobile robot synchronous positioning and map construction system. Background technique [0002] With the advancement of technology in the fields of machinery, information, materials, control, and medicine, home service robot simultaneous positioning and map construction (Simultaneous Localization and Mapping, referred to as SLAM) technology helps robots perform real-time positioning and build environmental maps in unknown environments. The basis for mobile robots to realize autonomous positioning and navigation. Visual SLAM has gradually emerged and developed rapidly in the past ten years. According to different types of visual sensors, visual SLAM is mainly divided into three types: monocular, binocular and RGB-D SLAM. Mur-Artal et al. proposed the ORB-SLAM algorithm in 2015. Based on the PTAM framework, the author improved most of the compone...

AI Technical Summary

Problems solved by technology

In the prior art, ORB-SLAM2 provides a large-scale data training dictionary based on ORB features, and in order to facilitate the system to directly read the dictionary, the storage method of the Text format dictionary is added, but the text file needs to be converted during the reading process Data type and line break, when the data size is large, the reading work will be very time-consuming, which leads to slow startup speed when applied on the robot platform

In addition, the dictionary provided by ORB-SLAM2 is trained based on a huge database, which helps the system to maintain good accuracy and robustness in different environments, but for robots used in specific environments, large-scale Dictionary not fully functional, too much invalid data

Images