Research on Positioning and Mapping for Embedded System

With the rapid development of artificial intelligence in society, the use of robots is becoming more and more extensive. The method of synchronous positioning and mapping with camera as sensor is an important one in SLAM technology. In this paper, combining vision sensor with embedded development board, the ORB-SLAM3 algorithm based on vision is studied and implemented. This paper introduces and describes each module in visual synchronous positioning and mapping, and presents an improved error matching elimination algorithm. Quality factor is introduced into feature point classification, and high-quality feature points are arranged in descending order according to quality level, which improves the efficiency and accuracy of the algorithm. ORB-SLAM3 algorithm will be adopted as the implementation scheme, and the overall system flow of the algorithm will be described and analyzed. The raster map is used to replace the original point cloud map representation, so that the map display can be more intuitive and clear and reduce the map storage space, improve the algorithm running time consumption and the point cloud map occupies a large memory space and other problems. After that, an optimized algorithm of synchronous location and map building is implemented on iTop4412.