种子纯度与活力快速测定的人工智能图像识别与处理系统

【作者】 任东；

【导师】 袁洪印；

【作者基本信息】 吉林农业大学 ， 农业机械化工程， 2003， 硕士

【摘要】 今天，计算机图像识别与处理技术已经成为发展现代农业不可缺少的组成部分。随着计算机硬件成本的下降和CPU运算速度的提高以及图像识别与处理技术本身的专业化发展，图像识别与处理技术在农业上的应用越来越广泛。 种子的纯度与活力对种子的销售者和使用者来说都是非常重要的指标。对杂交的玉米种子来说，种子纯度提高1％，作物产量可提高10％。近年来全国各地因种子纯度和活力等造成作物减产，绝收的情况时有发生，经济损失巨大。蛋白电泳、四唑染色、发芽率实验和田间实验等常规方法虽能测定种子纯度和活力，但在数据处理的速度、精度和结果分析等方面远远不能满足种子生产的实际需要。 本文旨在以我省的玉米品种为研究对象，以计算机、摄像机、图像采集卡、电泳仪等设备为工具，综合运用图像处理、统计学理论和对比试验等理论和技术，将计算机图像处理技术与玉米种子纯度与活力的生物学检测方法相结合，研制出种子纯度与活力快速测定的图像识别与处理系统。本研究分三大部分： (1) 检测方法研究 对试验采用的种子纯度与活力的生物学检测方法进行分析，进而选取最适合的检测方法。改进纯度检测方法，提高电泳谱带的清晰度，利用选取的纯度与活力检测方法，进行人工检测，得出可靠的人工检测结果。 (2) 系统研制 根据选取的检测方法，确定图像识别与处理系统的组成。对人工检测得到的图像进行分析，确定图像处理方法。对处理后的图像进行识别理解，编写软件程序包。最终研制出玉米种子纯度与活力快速测定的人工智能图像识别与处理系统。 (3) 系统可靠性研究利用对比试验来验证和修正图像识别与处理系统。 一、检测方法研究 研究了目前种子纯度与活力的检测方法。在玉米种子纯度检测的方法中，蛋白质凝胶电泳法具有快速、准确、稳定、经济、不受环境条件影响等特点，因此，选取此方法作为种子纯度检测方法。对国家试行的玉米蛋白凝胶电泳法的反映体系进行了改进，使得到的图像更清晰。在种子活力检测方法中，主要采用种子生吉林农业大学硕士学位论文玉米种子纯度与活力快速侧定的图像识别与处理系统活力、发芽率和幼苗评定来评价种子活力高低。种子生活力、发芽率、幼苗长势高低是种子活力高低的反应。四吐染色测定种子生活力是目前国际上承认的一种准确可靠的浏定方法，且易于图像处理，因此选取此方法作为种子活力的检浏方法。 二、系统研制 根据以上选取的玉米种子测定方法，设计了图像识别与处理系统的硬件组成。IBM一Pnl徽机、ccD行间变换摄像机、图像采集卡和光照室组成图像识别与处理系统。 将人工检测得到的图像进行图像预处理，将32位真彩色图像转化成灰度图像。降低了数据童，加快了图像处理速度。 应用图像的直方图均衡、窗口变换、平滑去噪、滤波等处理技术对图像进行去噪和增强，得到本研究所需要的图像特征。 种子活力的人工检浏是通过计算胚芽中染色的面积与未染色的面积的比值来定量的研究种子的活力。如果染色部分与整个种胚的面积比大于2/3，就认为此拉种子是有活力的。人工检浏方法是大概的估计染色的面积比。在大量的检浏中，由于疲劳等因素可能产生人工误判等情况。计算机图像识别和理解是以像素的个数表示区域的面积，保证了检测精度。对于种子活力检测，需要利用边缘检测对图像进行分割，得到活力检测所需要的种胚部分，为图像的理解作准备。对种胚部分选取特定阂值确定为染色像素，染色比例就是染色与未染色像素个数之比，从而得出活力检浏结果。 对于种子纯度的检浏，利用图像的模板匹配能很好地解决问题。在图像的模板匹配中，用下式来衡量模板少和被匹配图像酬的相似程度 MM艺艺S“(m，。)*T(m，。)砖，j)=.二l月二l、/戈戈s“(m，·，·觉觉:“(m，·，.=l月司用司月二l式中数时二二为模板上的坐标。当0<R(工，j)<1，并且仅在比值s“(m，n)lT(m，n)为常取极大值(等于1)o定义一个不变的阅值兀。当相似性浏度R(i，j)之兀时，吉林农业大学硕士学位论文玉米种子纯度与活力快速测定的图像识别与处理系统就认为模板的匹配是成功的。通过实验，当爪二0.“时，纯度检浏最成功。 将以上的各个方法设计成软件模块，加以集成，便形成图像识别与处理系统的软件流程。利用界面友好、功能强大的vc++开发语言，将该流程形成软件包。 三、系统可靠性研究 将计算机图像识别结果和人工检浏结果进行对比，验证和修正图像识别与处理系统。在种子纯度检浏中，图像识别与处理系统检浏结果与人工检浏结果的吻合率达到了95%以上，达到了项目任务书的要求。试验中发现，摄像机和数码相机的光学失真以及种子品种本身的质量问题对生成的电泳谙带以及图像识别的结果有很大影响。在种子活力检浏中，图像识别与处理系统检浏结果与人工检浏结果的平均吻合率达到9 6 .7%以上，同样达到了项目任务书的要求。更多还原

【Abstract】 today, image processing and identifying is the indispensable component of modern agricultural development. Along with specialization of image processing technology, descending of hardware cost and increasing of CPU speed. Image processing is more and more important in the way of agricultural applicationvigor and purity of seed is very important target for the seller and user. For cross corn seed, the purity of seed increase 1%, the crop output will increase 10%. In the recent years, Because of vigour and purity of seed, it is universal that output decreased and sterilized. Economic loss is enormous. The common method, such as protein electrophoresis, tetrazole staining burgeon experiment and farm experiment can measure vigour and purity of seed, but it can not be satisfied for actual request of seed producing in the way of speed and precision of data processing and result analyse.We aim for com breed of our province, using computer, pick-up camera, image gathering card, electrophoresis equipment ect as means, synthetically apply image processing technology, statistic theory and comparison experiment method, will combine the staining theory and and electrophoresis atlas character theory witch express character of seed vigor and purity with the image processing; Then develop a method and system of rapid measuring corn seed purity and vigor. The thesis can be divided into three parts:(1) research of measure method we study recent measuring method of seed purity and vigor, according to the choosed method we assure the component of image processing system.(2) system developing analyzing the processed image, confirm the method of imaging processing; comprehend and identify the processed image, design and test the program.(3) research of reliability Apply comparison experiment to verify and adjust the image processing system.1 research of measure methodwe study recent measuring method of seed purity and vigor. In recent measuring method, protein electrophoresis is rapid, accurate, steady and economic, not affected by environment condition, We improve reactive system of our national tentative com protein gelatinous electrophoresis, and that which is more adapted to imaging process. In the method of measuring vigor, we mainly use vital capacity, germination percentage and seedling to assess seed vigor grade, vital capacity, germination percentage and seedling of seed reflect seed vigor. Tetrazole staining method for measuring seed vigor that is accurate and steady which is admited by international word. And that witch is easy for imaging processing. 2, system developingAccording to our choosing, we design hardware component of imaging processing system, and preprocess it, Image processing system is as follows: IBM-PIIIPC, CCD camera, braket, lighting room.because quantity of real color data is very huge, we change it to grayness image.In the process of gathering image, it must produce noise, otherwise, we focus on the image character that we are interested in when we do it; so we should use the technology of histogram equalization of image, windows conversion, removing noise by smoothness and filtering to enhance the image.And then we part the image, get the destination image we are interested in, which create a condition for image comprehension. For measuring of seed vigor, we study seed vigor quantitatively by means of computing specific value of stained and unstained area in plumule. If specific value of stained and all of seed plumule is greater than 2/3,we take it to be a vital capacity the seed. As we know, manual decision method estimate specific value of stained area, and so in a large quantity of measuring, because of human factor of fatigue and so on, maybe generate misjudgement condition, computer can probably reduce productive probability of the mistake. In the image process, using pixel to express area is more accurate than eyes.In the method of purity measuring, we find that utilizing image template match can perfectly resolve the prob更多还原