Optimization and Path Planning of Simultaneous Localization and Mapping Construction Based on Binocular Stereo Vision
Neng‐Sheng Pai, Wei-Zhe Huang, Pi-Yun Chen, Shih-An Chen · Sensors and Materials · 2022
simultaneous localization and mapping (SLAM), random sample consensus (RANSAC), D*Lite, closed-loop detectionThe aim of this study is to help individuals easily reach their destinations independently whenever they are situated in an unfamiliar environment.After performing pre-processing optimization, a map constructed by simultaneous localization and mapping (SLAM) is inputted to a route planning algorithm to find the most suitable path.The proposed method utilizes a stereo camera sensor to take images of the environment, after which it conducts 2D/3D mapping through the SLAM framework before converting the constructed map into images.Then, obstacles are identified using an image segmentation method, and pseudo-obstacles are filtered out through optimization.Finally, route planning is conducted using the D*Lite algorithm.Experimental results revealed that most of the pseudo-obstacles can be filtered out through image optimization, thereby increasing the accuracy of the 2D map.