AI-Based Environment Segmentation Using a Context-Aware Channel Sounder

摘要

We describe how the data acquired from the camera and Lidar systems of our context-aware radio-frequency (RF) channel sounder is used to reconstruct a 3D mesh of the surrounding environment, segmented and classified into discrete objects. First, the images captured by the camera are segmented into objects through an AI-based algorithm. Then the segmented images are projected onto the point cloud captured by the Lidar. Since the receiver end of the channel sounder is mounted on a mobile robot, the data is acquired in the local coordinate system and so must be transformed to a global coordinate system to synthesize a single, holistic point cloud of the environment. Finally, the synthesized point cloud is tessellated into a 3D mesh. The segmented mesh can be used for the automated - i.e., without human analysis - reduction of the data acquired by the RF system of the sounder into an object-specific channel model.