Toward real-time particle tracking using an event-based dynamic vision sensor
Toward real-time particle tracking using an event-based dynamic vision sensor
复制标题
DOI：
10.1007/s00348-011-1207-y
复制
发表时间：
2011-11-01
影响因子：
2.4
通讯作者：
Jensen, Atle
中科院分区：
文献类型：
--
作者：
Drazen, David;Lichtsteiner, Patrick;Jensen, Atle
高雷诺数流体流动中的基于光学的测量通常需要高速成像技术。这些相机通常在内部记录数据，因此受到板载可用内存量的限制。提出了一种用于粒子跟踪测速的新型摄像技术。该技术由一个动态视觉传感器组成，其中像素并行操作，仅当遇到约10%的强度相对变化时才传输异步事件，时间分辨率为1 μ s。这导致记录系统的数据存储和带宽要求比典型的高速图像传感器小约100倍。从该传感器收集的数据的后处理时间也增加到比真实的时间快约10倍。我们提出了一项概念验证研究，将这种新型传感器与能够以1，024 x 1，024像素记录高达2，000 fps的高速CMOS相机进行比较。比较了每种系统在固液两相管流中跟踪致密（ρ>1 g/cm（3））颗粒的能力。雷诺数的基础上的体积速度和管道直径高达100，000进行了研究。
Optically based measurements in high Reynolds number fluid flows often require high-speed imaging techniques. These cameras typically record data internally and thus are limited by the amount of onboard memory available. A novel camera technology for use in particle tracking velocimetry is presented in this paper. This technology consists of a dynamic vision sensor in which pixels operate in parallel, transmitting asynchronous events only when relative changes in intensity of approximately 10% are encountered with a temporal resolution of 1 mu s. This results in a recording system whose data storage and bandwidth requirements are about 100 times smaller than a typical high-speed image sensor. Post-processing times of data collected from this sensor also increase to about 10 times faster than real time. We present a proof-of-concept study comparing this novel sensor with a high-speed CMOS camera capable of recording up to 2,000 fps at 1,024 x 1,024 pixels. Comparisons are made in the ability of each system to track dense (rho >1 g/cm(3)) particles in a solid-liquid two-phase pipe flow. Reynolds numbers based on the bulk velocity and pipe diameter up to 100,000 are investigated.