The Analysis of Task and Data Characteristic and the Collaborative Processing Method in Real-Time Visualization Pipeline of Urban 3DGIS

Abstract

Parallel processing in the real-time visualization of three-dimensional Geographic Information Systems (3DGIS) has tended to concentrate on algorithm levels in recent years, and most of the existing methods employ multiple threads in a Central Processing Unit (CPU) or kernel in a Graphics Processing Unit (GPU) to improve efficiency in the computation of the Level of Details (LODs) for three-dimensional (3D) Models and in the display of Digital Elevation Models (DEMs) and Digital Orthphoto Maps (DOMs). The systematic analysis of the task and data characteristics of parallelism in the real-time visualization of 3DGIS continues to fall behind the development of hardware. In this paper, the basic procedures of real-time visualization of urban 3DGIS are first reviewed, and then the real-time visualization pipeline is analyzed. Further, the pipeline is decomposed into different task stages based on the task order and the input-output dependency. Based on the analysis of task parallelism in different pipeline stages, the data parallelism characteristics in each task are summarized by studying the involved algorithms. Finally, this paper proposes a parallel co-processing mode and a collaborative strategy for real-time visualization of urban 3DGIS. It also provides a fundamental basis for developing parallel algorithms and strategies in 3DGIS.