【作者】 李伟；

【导师】 毛恩荣；

【作者基本信息】 中国农业大学 ， 车辆工程， 2004， 博士

【摘要】 播种均匀度是衡量播种机性能与质量的重要指标。本文提出的适用于精密播种机、谷物条播机、穴播机，及排种器等多类型播种机具的播种精度检测方法，是集计算机视觉、模式识别、自动控制为一体的综合技术。其核心研究工作是计算机视觉技术在播种精度动态检测过程中的应用。本研究实现了播种精度检测的准确、可靠，实验台操作的自动控制。主要研究工作包括： 1 研究了基于计算机视觉技术的播种精度检测试验系统。该系统首次实现了对精密播种机、谷物条播机、穴播机，及排种器等多类型播种机具播种精度的检测，解决了通过直接获得种子粒间距而检测播种精度这一难点问题。检测结果具有数据准确、可靠，粒距测量结果可溯源。 2 研究了动态图像视觉采集系统构建方法。包括：为了有效地去除大量冗余图像信息，减少计算机存储量，而采用的逐场采集和隔帧存方法；为实现三行播种通道种子信息的并行采集，图像三分量独立采集的软硬件技术方法；对比实验了自定义阈值选取与基于迭代方式的最优阈值的优缺点，选用了自定义阈值进行图像分割，缩短了图像处理时间：研究了基于序贯算法的种子区域标记技术与种子目标识别技术，并进行了质心参数计算。 3 研究了长序列动态图像匹配与拼接技术。根据种子动态检测的实时性要求，对动态图像的拼接问题展开深入分析与研究。阐述了种子动态图像拼接原理，提出了动态图像亚拼接检测方法，该方法基于标记特征，以搜寻相邻两帧图像中重叠区域内相同标记为目标，利用相同标记在前后两帧图像中应具有相同特征向量的特点，通过特征向量的匹配，实现冗余图像信息的去除，达到相邻两帧图像的拼；提出了大小间隔的标记方案，以适应采集工况，实现准确图像亚拼接；研究了重叠区域宽度对拼接精度的影响，给出了重替区域宽度设置方式。 4 研究了摄像机标定方法，并分析了三摄像机联合采集的系统误差。针对视觉环境中光照系统的构建问题，确定了双列交叉空间光照视场，有效的解决了动态目标检测中，光照系统环境与摄像机图像采集的一致性问题；对三台摄像机采集信息输入过程中的数据精确性问题进行了深入的分析，确定了三台摄像机之间系统误差分析模型、及标定方法。 5 研究了穴播精度检测算法和重叠种子分离方法。针对穴播机检测中穴距测量的关键问题，研究了基于几何中心矩的穴播质心、穴间距离检测算法，获得较高的检测精度；研究了基于数学形态学的重叠种子图像检测算法，定义了串联、并联的种子分布形式，对重叠、挤挨种子的分割进行了有效的工作。 6 研究了试验台系统自动控制技术。设计了基于串行通讯的微机与PLC集成控制方案，控制内容包括：排种器驱动轴转速控制、变频器的控制、传送带的稳定性及速度控制、工况模拟过程中的试验平台倾斜位置的自动控制等。 论文对试验系统测量误差进行了全面分析，推导了排种器转速与传送带运动速度匹配关系公式，对综合研究动态检测中的误差规律具有指导意义。 文中给出的理论分析与方法，已经应用于播种机性能检测实验台中，并进行了大量应用实验，符合GB／T6973-1986，GB／T9478-1988标准。整机检测误差≤±2mm。论文工作期间发表相关学术论文6篇。更多还原

【Abstract】 Seeding uniformity is an important factor to evaluate performance and quality of seeders. The comprehensive seeding accuracy testing method proposed in this dissertation to test performance of precision seeder, grain drill seeder, hill-drop planter, seeding monomer, and seed sowing device is an integrative one with technology of computer vision, patter recognition, and automatic control. The core task is application of computer vision techniques in seeding accuracy dynamical testing. Accurate and reliable seeding accuracy test and automatic control of the test-bed are realized. Main study contains:1 A seeder performance test system based on computer vision is built. Precision seeder, grain drill seeder, hill-drop planter, seeding monomer, and seed sowing device for the first time are realize to be tested on one comprehensive test-bed, and at the same time, hard problem of evaluating seeding accuracy from seed kernel space interval is solved. Test results are accurate, reliable and can be traced to the source.2 Construction method of dynamical image acquisition is studied. Field by field and frame absent image acquisition method is used to remove redundant information and saving computer memory. The three primary color channels of color image are separately captured to the image acquisition board by using special hardware and software technology, therefore, three lines of seeds are parallel captured. Contrastive test is done to compare advantage and disadvantage of threshold chosen method, which can either be chosen from trying or iteration, and predefined threshold chosen is selected, which result in lessen processing time. Region labeling using sequential algorithm and seed object recognition are studied, and then the center of a region is calculated.3 Sequential images matching and splicing technology is studied. Begin with real-time requirement of dynamical testing, image splicing problem is further researched and principle of image splicing is expounded. A new sequential images sub-splicing method is proposed, which based on features of markers and aimed at looking for the same markers in overlapped area of two adjacent images to finish the splicing task. No real image coalescence is done, but the redundant information is diminished, so the two adjacent images are considered well spliced. Big marks are interlaced with small ones to realize accurate image sub-splicing in real situation. Influence of the width of overlapped area on image splicing is studied and its width setting mode is given.4 Construction of the illumination system and camera calibration means are researched. Illumination system construction of computer vision system is studied and a double line space crossing illumination system is built, which results in effective settlement of disparity between illumination system and image acquisition in moving object testing problem. Data accuracy when using three cameras to capture information is further researched, and three cameras system error module and calibration method are determined.5 Hole centers location and hole distance measuring technology are studied. Overlapped seeds are separated by certain algorithm. The key problem in hole-drill planter test is locating of holes centers,which is solved by using geometry moments of discrete masses. And then the holes distance is calculated from holes centers with fairly high precision. Sends scatter state is observed. Overlapped seeds are recognized based on morphological and series connection and parallel connection seeds are defined to effectively separate overlapped seed.6 Automatic control technology of the test-bed is studied. System is coordinated controlled based on serial communication between the computer and PLC. The followed variables are controlled: the rotational speed of the seeding device driven draft, the operation of the converter, stability and speed of the convey belt, and inclination of the test platform, which is done to simulate actual working condition of seeder.Measuring error is overall analyzed. Fo更多还原