【作者】 井利民；

【导师】 何东健； 张建锋；

【作者基本信息】 西北农林科技大学 ， 计算机应用技术， 2009， 硕士

【摘要】 苹果果实图像特征的实时精确提取、果实目标的快速识别与定位是提高苹果采摘机器人采摘速度的关键问题。论文针对目前苹果采摘机器人视觉系统中存在的系统资源消耗大、效率低等问题,结合嵌入式系统及实时数字图像处理技术等领域的最新研究成果,研究基于ARM的苹果图像果实识别与定位系统,以使处理速度满足实时要求,果实定位准确,为设计开发苹果采摘机器人提供技术依据。论文主要研究内容和结论如下:(1)在分析影响苹果采摘机器人作业效率原因的基础上,深入研究了机器视觉及实时数字图像处理技术,详细分析了嵌入式系统开发流程和设计准则及基于ARM的嵌入式系统实现方法,结合苹果图像果实识别与定位的应用需求,提出一种基于ARM7TDMI+μC/OS-Ⅱ架构的苹果图像果实识别与定位系统方案,并对系统的开放性进行评价。(2)在分析比较的基础上,对苹果果实图像识别与定位系统实现的硬件平台进行了设计和选择,构建了基于ARM7TDMI微处理器为基础的硬件方案,完成了以键盘输入、LCD显示等模块硬件电路的实现和调试。(3)对比多种嵌入式操作系统的优缺点,分析系统的软件体系架构,结合系统硬件平台的特性,建立以μC/OS-Ⅱ操作系统为基础的软件平台,实现Bootloader的移植和相关驱动程序的开发。(4)研究了自然条件下采集苹果图像的预处理方法,针对中值滤波法不能直接应用于彩色图像的问题,采用了一种改进的中值滤波方法对图像进行预处理,并对受遮挡果实的识别与定位算法进行研究,由于Hough变换技术对于图像中噪声及信息缺失的不敏感性,该算法可以实现遮挡果实的有效精确识别,同时对Hough变换中存在的检测速度问题进行了改进,并将该算法在嵌入式系统予以实现、移植。(5)测试结果表明,本文设计实现的苹果果实识别与定位系统可以实现苹果果实图像的快速准确定位,系统整体运行稳定,能够适应苹果采摘机器人视觉系统的要求,对今后苹果采摘机器人的实用化有着积极的意义。更多还原

【Abstract】 The speed that harvest robot picks apples is an important issue in apple harvest robot fields. To improve this speed, the key is extracting the apple fruit quickly and accurately, and then locating the apple rapidly from an image. Considering the shortcomings of apple fruit image recognition and location system, on the basis of analyzing and improving the algorithms of real-time image processing, this paper combines the latest achievements of embedded system in the technology of real-time image processing. Under the terms of utilizing software and hardware resources rationally, we regard ARM as the platform, research and implement an apple fruit image recognition and location system which obtains good performance and high speed. This provides technical sustain for designing quicker apple fruit harvest robot. The major research contents and conclusions of the paper are as follows:(1) We analyze the reasons why apple harvest robot is inefficient, study the technology of computer vision and real-time digital image processing deeply, and analyze the flow of embedded system development, the design principle and the way to implement the embedded system for ARM. Considering the application needs of apple fruit image recognition and location system, we propose a scheme based on the structure of ARM7TDMI+μC/OS-Ⅱ, and then evaluate the open characteristic of the system.(2) Through analysis and comparison, we design and select the hardware platform, construct the hardware scheme based on ARM7TDMI microprocessor, implement and debug keyboard and LCD.(3) We compare the advantages and disadvantages of many embedded operating systems, analyze the structure of software scheme. Considering the characters of hardware platform, we build a software platform based onμC/OS-Ⅱ, transplant Bootloader and implement device driver.(4) We study the pretreatment method of apple images which are collected under natural scenes, in order to overcome the problem that median filter can not be used in the color image directly, we improve it to suitable for the pretreatment.For the advantage of its insensitivity to the noise, Hough transform can recognize and locate the apple fruits rapidly and effectively. Then we improve the Hough transform to overcome the disadvantage of comparative low speed. In the end, we transplant all algorithms into the embedded system. (5) The test results show that the apple fruit image recognition and location system can locate the apple fruit rapidly and accurately, this system is stable, reliable, and suitable for the vision system of apple harvest robot, it is beneficial to accelerating industrialization of apple harvest robot.更多还原