AER时域视觉传感器总线仲裁建模与仿真分析  

Modeling and emulation of address-event representation vision sensor bus arbitration

作　　者：杨帅[1] 胡燕翔[2] 徐江涛[3] 

机构地区：[1]天津工业大学计算机科学与软件学院,天津300387 [2]天津师范大学计算机与信息工程学院,天津300387 [3]天津大学电子信息工程学院,天津300072

出　　处：《天津工业大学学报》2016年第6期84-88,共5页Journal of Tiangong University

基　　金：国家自然科学基金资助项目(61274021);天津市科技重大专项与工程计划资助项目(15ZXHLGX00300)

摘　　要：以视觉传感器设计为背景,对AER时域视觉传感器的总线仲裁进行建模,并对事件取样、AER仲裁和目标跟踪进行仿真分析.实验结果表明：AE事件数量仅占全部像素数目的5%-10%,事件集中优先仲裁方式比特定区域优先仲裁方式和轮权仲裁方式在高速目标跟踪应用上准确率高10%-20%.因此可知：基于AE数据的目标跟踪更加高效,并且事件集中优先仲裁方式更适合于高速目标跟踪应用.An emulator of AER vision sensor bus arbitration is implemented and is used to emulate the process of event sampling, AER artribution and object tracking. Experimental result indicates that the visual information generat-ed by AER vision sensor is only 5%-10% of the total number of image sensor pixels, and the object tracking ac-curacy of event-aggregation arbitration is higher than that of specific region arbitration and round-robin arbitra-tion in high-speed object tracking applications by 10%-20%. So object tracking based on AE data is more effi-cient ,and event-aggregation artribution has more efficiency for the applications of high speed object tracking.

关 键 词：地址-事件表示(AER) 视觉传感器 模拟器 仲裁 目标跟踪 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]