Estimation of Spatial Features in 3-D-Sensor Network Using Multiple LiDARs for Indoor Monitoring

Abstract

In recent years, smart city has been attracting attention as a sustainable urban development. Smart monitoring is one of the key components of a smart city, and Light-Detection-and-Ranging (LiDAR) sensor is used as one of the sensors for smart monitoring. A 3D sensor network using multiple LiDAR sensors can be constructed to eliminate blind spots. Prior work presented a system that prioritizes transmission in important regions to prevent data loss in the case of bandwidth limitation. However, prior work has only presented transmission methods and has not shown how to estimate important regions in the point cloud space. This paper proposes a system to estimate important regions using spatial features created based on multiple spatial metrics. The important regions vary from task to task, and in particular, for the task of detecting moving objects, the important regions are the spaces where the moving objects may be located. The accuracy of estimation can be improved by creating spatial features based on more spatial metrics. This paper addresses two kinds of spatial metrics, temporal metric and statistical metric. In this work, we collect point-cloud data using multiple LiDAR sensors and produce datasets with labels for evaluation. Using datasets, we validate the proposed system in terms of the accuracy of spatial feature estimation.