Adaptive Voxel Mapping Based on 3D Occupancy Grid Maps Using Object Detection

This paper proposes a voxel-based adaptive resolution mapping method to enable autonomous robots to efficiently perceive and map 3D spaces in various environments. Using the YOLO real-time object detection model, the system selectively switches the required areas and forms a consistent map by excluding overlapping points between high-resolution and low-resolution maps on an octomap. The study measures the number of point clouds in single-resolution and multi-resolution scenarios and analyzes map capacity by varying the ratio of high-resolution to low-resolution areas in adaptive resolution mapping. Experimental results demonstrate that the proposed mapping method maintains the accuracy of environmental perception while demonstrating efficiency in data capacity and computational performance, confirming its potential for resource optimization in real-time autonomous robot operations.