A dynamic compression technique for streaming kinect-based Point Cloud data

Abstract

With the relative inexpensiveness of 3D sensors today, it has become easier to collect 3D spatial data. This opens up the usability for these data, called point clouds, in areas such as robotics, telemedicine, and entertainment. However, a primary issue is the data rate of point clouds being produced by sensors such as the Microsoft Kinect. When transmitted, the size of a single point cloud coupled with the high frame rate results in an impractical bandwidth requirement. Therefore, compression is required. However, current compression techniques offered by the Point Cloud Library (PCL) are static in terms of available network bandwidth, meaning the techniques do not adjust as the network changes. In this paper, we propose using a dynamic compression technique that adjusts the compression ratio in response to the network throughput. Experimental results under both static and dynamic network traffic conditions show that dynamic compression is promising in achieving higher and smoother frame rates compared with the built-in compression algorithm in PCL.