A Progressive Transmission Technique for the Streaming of Point Cloud Data Using the Kinect

Abstract

Sensors for collecting 3D spatial data from the real world are becoming more important in various applications in consumer markets, such as medical, entertainment and robotics. However, a primary concern with collecting and delivering data from these sensors is the vast amount of information being generated, which, in turn, needs to be processed and transmitted. To address the issue, we propose the use of a progressive transmission scheme. To collect the 3D spatial data, called point clouds, we used the Microsoft Kinect sensor. In addition, we utilized the Point Cloud Library to process the data and the Boost library for transmission. A client/server application was developed to evaluate our approach against two others approaches: adaptive resolution transmission and dynamic compression. These three ideas were tested in combination with additional techniques, including filtering and frame skipping. Moreover, we conducted these experiments in three different network environments: normal, static, and dynamic. We found that, overall, the use of progressive transmission outperforms the other solutions in terms of non-visual quality of service metrics. However, in a visual context, the dynamic compression technique tended to outperform the others.