节点文献
图像处理与伺服控制一体化系统设计研究
【作者】 袁文婷;
【导师】 刘建平;
【作者基本信息】 国防科学技术大学 , 控制科学与工程, 2006, 硕士
【摘要】 利用视觉系统获取信息作为反馈变量是现今较为有效的闭环控制手段之一,具有重大的科研价值和广泛的应用前景。视觉作为获取外界信息最重要的方法之一,已经成为各种控制智能化、自主化的前提条件。基于图像处理与伺服控制一体化系统的研究是这个领域的热点,对于智能机器人控制和红外成像制导的研究、发展和应用都具有重大的意义。我国在这个领域的研究起步比较晚。目前,基于视觉控制的智能机器人在应用实践方面尚处于探索阶段;成像制导武器在精确制导、抗干扰以及“发射后不管”的智能化等方面的技术水平尚还落后于美国、俄罗斯等国家。但我国的计算机科学技术、信息科学技术目前正处于快速蓬勃发展的阶段,这给了我们一个迎头赶上的机会。本文首先分析了机器人视觉伺服系统、红外成像制导系统和相关算法的最新发展情况,并对图像处理和伺服控制的关键技术做了深入研究,提出了以下问题:1.由于背景环境的干扰,简单算法的识别和跟踪效果并不令人满意。若要提高系统的自适应性和智能化,就需要以加大算法的复杂度作为代价,有时甚至需要形成一个庞大的算法库。而目前视觉伺服系统广泛采用的实现方案对运算资源的利用率较低,难以满足复杂背景下目标识别和跟踪的实时性要求。2.缺乏一体化的SoC设计,不适合对功能、可靠性、成本、体积、功耗等综合性严格要求的应用环境。3.系统通用性较差,进行移植、升级时软硬件架构的改动量都比较大,工作繁杂,开发周期长。4.单核数字信号处理器的方案所实现的系统在处理复杂算法时,难免频繁调用跳转指令,造成11级流水线经常中断,系统效率降低。围绕以上问题,本文研究了图像处理与伺服控制一体化的系统,提出了一种外围电路简单、软件结构灵活的通用设计方案。可广泛应用于机器人控制、自动焊接、自动分检和导弹制导等领域。具体研究内容包括以下几个方面:1.系统介绍了自动目标识别和跟踪的相关算法,重点介绍了卡尔曼滤波、差分法、空域滤波法、时域滤波法和门限法。并对各种经典算法的改进算法进行了归纳和比较,分析了算法实现的条件和难度。为硬件需求分析和系统的实际应用提供了理论依据。2.在研究了自动目标识别和跟踪算法的基础上,对现有的设计方案进行了定性、定量分析。在对各种方案比较分析后,针对目前实际应用中所面临的关键问题提出了一种集成度很高的双核设计方案。构建了总体框架,详细说明了实现过程。在具体研究中,分析了关键模块的应用及内部资源,对视频接口、程序存储器接口、数据存储器接口、电源电路、系统时钟和芯片工作模式方面的配置进行了设计,并完成了高速数字电路的PCB板设计。3.设计了基于模块化和ASOS体系层次的软件架构。根据ARM和DSP处理能力各自的特点,提出了面向双核处理器的分任务协作机制,进一步提高了运算资源的使用效率。最后利用Flash加载方法对DSP引导程序环节进行了设计。4.分析并给出了图像处理经典算法的实现方法,编写了C语言程序代码,并使用线性汇编对关键代码进行优化以提高运行速度,给出了CCS仿真器上运行后的处理结果。经分析后,提出一套测试方案。借助集成开发环境(Code Composer Studio Version)中包含的软件仿真器(Simulator),与单核处理器进行比较分析。通过对任务调度时间、数据通信时间和存储器占用情况等关键量进行反复测试和统计,部分验证了本文设计方案处理数据的并行度比较高,并且各处理器的存储空间都得到了比较合理的应用,部分数据的存取可以独立进行,避免了流水线冲突。从而提高了总体系统实时处理数据的能力。
【Abstract】 Nowadays, getting visual information as a feed back is an effective way for closed-loop control, witch has got an important science value and an abroad application future As one of the important ways to get information form outside world,vision has been the necessary condition for intelligent control.Visual servo system based on automatic target recognition and tracking is the hotspot in this field,which is important for the search,development and application of intelligent robot control and infrared missile.Our country has been dropped behind a little.The search of intelligent robot controlled by vision is just on the launch and the ability of precision striking,anti-jamming and intelligent control is all dragged behind American,Russia etc.However,the computer science and information technology is developing rapidly in our country recently,and this give us an opportunity to catch up.This thesis has analyzed the recent situation of the robot visual servo system,infrared imaging missile and the correlative arithmetic and raised such questions as following:1.Because of the interference from background,the effect of simple arithmetic is not satisfied.For improving the adaptability and intelligent,complex arithmetic or even a large arithmetic base is necessary.But now,the ratio of operation resource is a little low,witch causes the difficulty for processing when in the complex environment.2. The existing realization lacks of Soc design and is not suitable for the application that has strict limits in function, cost,volume,power consume etc.3.The existing systems are lack of universal hardware and this shortcoming will cause a lot of changes when the system is modified for another use.4.There is a problem for single core DSP that when to call jump instruct frequently,the efficiency of system will drop.So aiming at these problems,this thesis researched the integrated system of image process and servocontrol.Then it raised a design,whose hardware is simple and software structure is flexible.This design is universal enough so as to be used in robot control,automatic weld,automatic recognition,infrared missile etc. The research work contained as follows:1. Give an introduction of the correlative arithmetic of automatic target recognition and tracking,which focus on the Kalman filter,threshold method etc. And then,make a compare and analysis on these arithmetic in the aspects of realization condition and difficult,wich is the theory basis of the practice use.2. Based on the research of the automatic target recognize and tracking,quantilisly analyzed the existing systems.After comparison and analysis, a design based on the high integrated double core DSP was raised.In the thesis,the detail design about video port, program RAM,data RAM,PCB board,clock and mode was given.3. Design the software structure based on blocking and ASOS concept.Aiming at the differences in the processing ability of ARM and DSP, a task-distribute mechanism designed for double core DSP was raised in order to improve the ratio of the operating resource At the stone time, the thesis discussed how to boot program by flash.4. Analyze and give the realization method for classic image process.Then program with C and optimize the code for a fast operation speed.The results were given and analyzed.In order to test the performance of the system,a method was given in the thesis.With the help of Code Composer Studio Version and Simulator,we can make a compare with the single core system.After analyze the test results got with the CSC and Simulator,we reached the conclusion that the design of this double core system has improved the performance of the existing system.
【Key words】 Image process; Servocontrol; Integrated System; Automatic Target Recognition;
- 【网络出版投稿人】 国防科学技术大学 【网络出版年期】2007年 06期
- 【分类号】TP391.41;TM921.541;TP242
- 【被引频次】3
- 【下载频次】296