视线跟踪系统的SOC设计和验证
本文选题:SOC + 视线跟踪 ; 参考:《华南理工大学》2013年硕士论文
【摘要】:SOC(片上系统)将微处理器,数字、模拟功能单元和存储器等基本模块集成到单一芯片上,从而具有高速、低功耗和多功能等特点。因此,SOC技术及其设计方法学已成为集成电路未来发展和研究的主流,其中SOPC(System On Programmable Chip)技术,即可编程片上系统,提供了一种在性能、效率和成本上可灵活折衷的SOC解决方案。它通过片上总线和IP核(知识产权)复用的方式,将软硬件各功能模块集成到单一片上系统,实现SOC系统的集成。 本文依据片上总线和IP核复用的SOC设计方法,开展视线跟踪系统的SOC设计与验证。在分析系统需求和算法原理的基础下,比较各种可行方案,对系统进行设计和集成,并在SOPC实际系统中进行验证。本文的主要研究工作包括: (1)介绍SOC的概念、优势及其设计方法学的研究现状和视线跟踪系统的SOC研究进展,分析说明了视线跟踪SOC系统广阔的应用前景和市场前景。 (2)介绍基于普尔钦斑点法的视线跟踪技术的基本原理和工作流程,在普尔钦斑点法流程分析的基础上对SOC系统进行功能划为,确定SOC系统的框架。深入分析基于Adaboost的人眼定位算法的基本原理,对其进行数据流的分析和模块划分,提出了基于向量法的快速积分图生成器,并针对级联分类器通过率的差异设计串并混合架构的分类器。 (3)详细介绍视线向量提取的基本原理,针对瞳孔、亮斑定位算法的特点,采用软硬件协同设计的方法设计视线向量提取硬件模块。本文采用随机抽样一致性算法进行椭圆拟合以精确估计瞳孔中心,提高了系统的视线跟踪精度。通过分析随机抽样一致性拟合算法的数据流,对其进行模块划分,完成了改进型伪随机数生成器、矩阵快速逆运算和基于代数距离误差累计模块的设计和验证。 (4)对本文视线跟踪系统进行SOPC集成和验证,通过对样本集进行的测试实验说明SOC系统的运算结果与原型算法高度接近,保持了算法的原样性。同时,系统运行在50MHz的SOPC平台上,处理速度达到11帧/秒,,基本达到实时性的要求。通过对SOPC系统与其他平台系统的检测精度与运行效率比较证明:SOC系统能够在保证精度的前提下,以较低的运行频率实现高性能的运算,充分体现了SOC技术及其设计方法学的先进性。
[Abstract]:SOC (on-chip system) integrates the basic modules of microprocessor, digital, analog function unit and memory into a single chip, which has the characteristics of high speed, low power consumption and multifunction. Therefore, SOC technology and its design methodology have become the mainstream of IC development and research in the future. Among them, SOPC (system on Programmable Chip) technology, that is, system on Programmable Chip, provides a flexible SOC solution in terms of performance, efficiency and cost. By means of on-chip bus and IP core (IP) reuse, the software and hardware modules are integrated into a single chip system to realize the SOC system integration. According to SOC design method of on-chip bus and IP core reuse, the SOC design and verification of line of sight tracking system are carried out in this paper. Based on the analysis of the system requirements and the principle of the algorithm, this paper compares various feasible schemes, designs and integrates the system, and verifies it in the actual SOPC system. The main work of this paper is as follows: (1) introduce the concept, advantages and design methodology of SOC. The broad application prospect and market prospect of the sight tracking SOC system are analyzed. (2) the basic principle and work flow of the line of sight tracking technology based on the Pulchin speckle method are introduced. Based on the process analysis of Pohin's spot method, the function of SOC system is classified, and the frame of SOC system is determined. The basic principle of human eye location algorithm based on Adaboost is analyzed, and the data flow and module partition are analyzed. A fast integral graph generator based on vector method is proposed. A series-parallel hybrid classifier is designed for the difference of pass rate of cascaded classifiers. (3) the basic principle of line of sight vector extraction is introduced in detail, and the characteristics of pupil and bright spot location algorithm are discussed in detail. The hardware module of line of sight vector extraction is designed by hardware and software co-design. In this paper, the random sampling consistency algorithm is used for ellipse fitting to estimate pupil center accurately, which improves the tracking accuracy of the system. By analyzing the data flow of random sampling consistency fitting algorithm and dividing it into modules, the improved pseudorandom number generator is completed. The fast inversion of matrix and the design and verification of the module based on algebraic distance error accumulation. (4) the SOPC integration and verification of the line of sight tracking system in this paper. The test results of the sample set show that the calculation result of SOC system is close to that of the prototype algorithm, and the algorithm remains the same as the original algorithm. At the same time, the system runs on a 50MHz SOPC platform, and the processing speed reaches 11 frames / sec, which basically meets the requirement of real time. By comparing the detection accuracy and operation efficiency of SOPC system with that of other platforms, it is proved that the system can achieve high performance operation with low operating frequency under the premise of ensuring precision. It fully embodies the advancement of SOC technology and its design methodology.
【学位授予单位】:华南理工大学
【学位级别】:硕士
【学位授予年份】:2013
【分类号】:TN47;TP391.41
【参考文献】
相关期刊论文 前10条
1 徐宁仪,周祖成;Avalon总线与SOPC系统架构实例[J];半导体技术;2003年02期
2 李乙成,周祖成,陈尚松;SoC片上总线技术的研究[J];半导体技术;2003年02期
3 赵雪峰;一种伪随机数生成算法的研究与实现[J];电脑学习;2005年06期
4 王朔,李刚;USB接口器件PDIUSBD12的接口应用设计[J];单片机与嵌入式系统应用;2002年04期
5 魏少军;;未来SoC技术发展的几个特点[J];电子产品世界;2008年10期
6 刘丁,毛德柱,王云飞;USB在数据采集系统中的应用[J];电子技术应用;2000年04期
7 冯毅;易江芳;刘丹;佟冬;程旭;;面向SoC系统芯片中跨时钟域设计的模型检验方法[J];电子学报;2008年05期
8 王前,吴淑泉,李韬;逆变器消谐方程实时求解中矩阵求逆与ASIC实现[J];重庆大学学报(自然科学版);2005年03期
9 徐江丰;张涌;汤心溢;王世勇;;基于Wishbone-PCI Bridge核的红外图像高速采集系统[J];红外与激光工程;2006年06期
10 ;集成电路产业“十一五”专项规划[J];中国集成电路;2008年02期
相关重要报纸文章 前1条
1 记者 谢耕发;[N];科技日报;2000年
相关博士学位论文 前1条
1 刘瑞安;单摄像机视线跟踪技术研究[D];天津大学;2007年
相关硕士学位论文 前8条
1 宋卫艳;RANSAC算法及其在遥感图像处理中的应用[D];华北电力大学(北京);2011年
2 梁铭炬;自由空间视线跟踪算法研究[D];华南理工大学;2011年
3 夏菁;椭圆拟合方法的比较研究[D];暨南大学;2007年
4 林皓;基于FPGA的矩阵运算实现[D];南京理工大学;2007年
5 胡海朋;一种新的伪随机数产生方法及其统计性能分析[D];国防科学技术大学;2007年
6 谭阳;关于随机数生成算法的研究[D];湖南师范大学;2008年
7 居然;Adaboost算法并行硬件架构研究与FPGA验证[D];上海交通大学;2008年
8 曾宇森;视线跟踪SoC的系统建模及验证[D];华南理工大学;2012年
本文编号:2117257
本文链接:https://www.wllwen.com/falvlunwen/zhishichanquanfa/2117257.html