Semantic Map Construction with Camera and LiDAR for Real-Time Object Detection

Real-time semantic mapping is crucial for enhancing robotic perception in dynamic environments. This paper presents a novel system that integrates deep learning-based object detection with Simultaneous Localization and Mapping (SLAM) to create a detailed semantic map for indoor navigation. The proposed approach leverages camera and Light Detection and Ranging (LiDAR) data, utilizing You Only Look Once version 3 (YOLOv3) for object detection and Gmapping for SLAM. The system generates both geometric maps and an object-level understanding essential for effective robot navigation and interaction. Implemented on a low-cost computing platform, the system was tested in indoor environment, showcasing its ability to detect and localize objects in real time. The results indicate that the integration of these technologies facilitates improved navigation and decision-making capabilities in complex environments. This research contributes to the advancement of real-time semantic mapping and highlights the potential for future enhancements, such as scene classification and 3D mapping.