High Speed Laser and Camera for in Situ 4D Visualization
Abstract
This proposal requests funds in the amount of $363.4k for the purchase of a high speed camera and a high-speed laser system for the research of four dimensional (4D) in situ visualization. The requested equipment will be used in an ongoing research project funded by the US Army Research Office (Contract number: W911NF1820192, Project Title: In situ 4D visualization, duration: 2018- 2022). The focus of the research is to enable in situ visualization of 4D (all three spatial dimensions plus time) datasets and higher dimensional datasets through collaborative efforts among the University of Virginia (Charlottesville, VA), TARDC (Warren MI), and ARL (Aberdeen Proving Ground, MD). Visualization of multidimensional datasets based on post-processing is relatively mature. However, in many applications, it is of great interest and benefit to Army to be able to visualize high dimensional datasets in situ, and furthermore to guide the subsequent postprocessing based on the in situ visualization. The project proposes several research topics that can potentially enable such in situ aspect and also proposes an experimental demonstration. In this project, we propose to tackle in situ visualization in three steps. In the first step, we propose to research algorithms and hardware that can rapidly pre-process large amount (on the order of 50GB per second) of high dimensional data (3D and higher). Two specific directions are proposed: sparse representation and dynamics meshing. For the first direction, we propose to investigate multidimensional POD (proper orthogonal decomposition) and compressive sensing to extract the sparse presentation of the data. For the second direction, we propose a few ideas to attempt dynamic meshing based on the raw information in the data. In contrast, existing dynamic meshing algorithms predominately depend on post-processed information. In the second step, we propose to research in situ visualization engine (VE), and our efforts will concentrate again on two specific directions: asynchronous rendering and superspatiotemporal resolution. Recognizing the fact that humans cannot capture changes in details of a moving object and can only do so of a static object, the asynchronous rendering technique attempts to exploit and embed this fact in the VE by using different resolutions depending on the targetÕs motion. When it is indeed important to accurately analyze and display transient motions in situ, we propose to investigate a super-resolution technique based on an intensity variation function of grids both in spatial and temporal domain. Both directions are of particular value to visualize large and high-dimensional datasets with mismatching spatial or temporal scales. In the third step, we propose to test and demonstrate the algorithms and hardware developed above in collaboration with TARDEC and ARL under the context of fire suppression, an ideal testbed to generate high dimensional data and an ideal application where in situ visualization is of critical importance for the Army. These collaborations will leverage two complementary expertise: the PIÕs expertise in multidimensional visualization and the TARDEC teamÕs expertise in large-scale computation and fire suppression experimentation. Lastly, the above algorithms and hardware are expected to be applicable to higher dimensional data beyond 4D (e.g., spatiotemporally resolved velocity vectors), and exploratory research is proposed to investigate such applications and possible limitations.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 09, 2020
- Source ID
- W911NF2010134
Entities
People
- Lin Ma
Organizations
- Army Contracting Command
- United States Army
- University of Virginia