面向边缘应用的脉冲视觉传感器

Pulse vision sensor for edge applications

图像传感器是人类获取视觉信息的重要途径,在边缘感知和计算中有着重要的应用。在边缘端感存算一体化的发展趋势下,传统CMOS图像传感器由于大量的实数型数据、冗余的像素数据以及较低的时间分辨率,阻碍了边缘视觉系统的进一步发展。借鉴人眼视觉原理,人们发现了生物成像系统的电信号脉冲特性、高效的信息预处理机制以及丰富的信息获取种类,并从中获得了改进视觉成像的设计灵感。从上世纪80年代起,研究者们试图通过模仿人类视觉构建新型脉冲视觉传感器,包括高动态脉冲调制成像、高动态高灵敏度边缘检测动态视觉成像和二维三维融合的单光子成像。在未来的趋势中,脉冲传感器将有助于构建低延迟、多功能、低功耗的感存算一体化边缘视觉系统。

Image sensors are an important way for humans to obtain visual information,and have important applications in edge perception and computing.Under the development trend of the integration of sensing,storage and computing at the edge,traditional CMOS(complementary metal oxide semiconductor)image sensors hinder the further development of edge vision systems due to a large amount of real data,redundant pixel data,and low time resolution.Drawing on the principles of human vision,people have discovered the electrical signal pulse characteristics of biological imaging systems,efficient information preprocessing mechanisms,and rich types of information acquisition,and obtained design inspiration for improving visual imaging.Since the 1980s,researchers have attempted to construct new pulse vision sensors by imitating human vision,including high dynamic pulse modulation imaging,high dynamic and high sensitivity edge detection dynamic vision imaging,and single-photon imaging of two-dimensional and three-dimensional fusion.In the future trend,the pulse sensor will help to build a low-latency,multi-functional,and low-power sensor-memory-computing integrated edge vision system.