Low-Resolution LiDAR Upsampling Using Weighted Median Filter

Abstract

This paper presents a 3D LiDAR data upsampling method to obtain dense 3D depth data from a low-resolution LiDAR, a vision camera, and the Weighted Median Filter (WMF) algorithm. Recently, LiDAR is widely used in the field of Computer Vision since it can obtain accurate 3D data. However, data acquisition from the LiDAR is expensive due to the high cost of the LiDAR. We address how to obtain large amounts of 3D data from a low-channel LiDAR. In this paper, we acquire LiDAR data and color images from a calibrated multi-sensor platform. We first begin the upsampling steps from linear interpolation of a depth image. And then, we use an WMF algorithm to complement the first interpolation image. We use the upsampling algorithm to create dense 3D depth map from the existing sparse LiDAR data. And we generated a high-density 3D map using the ICP matching of multiple depth data acquired by a moving robot system.