【作者】 王树才；

【导师】 文友先；

【作者基本信息】 华中农业大学 ， 农产品加工及贮藏工程， 2006， 博士

【摘要】 无论是为了提高鲜蛋出口，还是进行禽蛋的纵深加工，都必须对禽蛋进行清洗、消毒、检测和分级。而目前在完成这些工作方面，我国多数企业仍沿用人工作业，无论是在加工质量方面，还是在加工效率方面，我们与发达国家都存在较大差距。在美国，禽蛋的采集、保鲜、分级、包装均采用机械操作，自动化程度高，成本低。 本系统的工作原理是首先由场景摄像机采集蛋的群体图像，经过图像处理，提取每枚蛋的中心坐标和长轴方向等特征值，将该数据经过坐标变换和机器人运动逆解，获得机器人关节运动量，机器人末端执行器运动到蛋的中心位置，调整位姿，发送信号给末端执行器吸盘单片机控制器，启动真空吸附气路，将蛋吸取。机器人根据运动规划，将蛋搬运到蛋壳敲击装置上方，蛋壳被敲击发出声音，麦克风采集该声音信号，送DSP系统处理，识别破损蛋，将识别结果传递给机器人运动控制器，如果是破损蛋，就放到破损蛋蛋箱；如果不是破损蛋，机械手运动到灯箱和彩色摄像头之间，获得蛋内容物彩色图像，经过图像处理，按反映蛋新鲜度的哈夫值与蛋芯颜色HSI之间的关系模型进行模式识别，并按新鲜度分级，将分级信息传递给机器人运动控制器，控制机械手将蛋放入相应等级的蛋箱中。 本文使用图像一阶矩与像素总面积的比作为每枚蛋中心在图像坐标系下的坐标，采用线性标定方法转换到绝对坐标系下，使用二维平面上与最小惯量轴同方向的最小二阶矩轴为蛋的长轴。使用最佳阈值变换和二维零均值离散高斯函数平滑处理，通过多次图像的腐蚀与膨胀，进行图像分割，并采用四连通成分序贯算法，对图像中的连通成分(同一个蛋的区域)进行标记。使用边界跟踪计算蛋的周长，从而得到区域的密集度，用于修正蛋的大小计算。对以上内容在理论研究的基础上，得出的相应算法，能够满足引导机器人准确定位每枚蛋的工作要求。 为了让机械手准确将蛋吸取，通过大量的矩阵变换，将蛋中心的绝对坐标和长轴方向解算到机械人各关节转角，即运动逆解；为了在一次吸放蛋运动过程中完成全部检测和分级工作，对机器人进行运动轨迹规划，将检测和分级工作中机器人运动始点和终点都是变化的这一复杂过程简化为始点变化终点固定(从吸起到破损检测)、始点终点都固定(从破损检测到品质检测)和始点固定终点变化(从品质检测到放下)三段，这种更多还原

【Abstract】 Whether for the export of fresh eggs or for egg processing, it is necessary to clean out, antisepticise, detect and grade eggs. At present, most processing plants in China are still employing manual eyes in fulfilling these tasks; as a result, there is a remarkable gap between China and developed countries both in processing quality and in efficiency. In the US, mechanical operation is applied to the collection, preservation, grading and packing of eggs, which is highly automatized and less costly.The principles of the system are as follows. First, a scene video camera captures the image of the massive members of eggs. Through image processing, the character extraction of the central coordinate, the major axis direction, etc. of each egg is used to obtain motion angles of the robot joints by coordinate transformation and kinematic inverse computation of the data. A robot end-effector then moves to the center of the egg, coordinates its pose, transmits signals to SCM controller of the suction cup, starts the suction circuit, and then adsorbs the egg. According to the motion planning, the robot carries the egg to the top of the knocking device. The eggshell is knocked and the sound is collected by a microphone and sent to DSP for processing and diagnosing eggs’ dilapidation, and the signals of the result are sent to a robot motion controller. If it is cracked, the egg will be put in the eggcrate for cracked eggs. Otherwise, the manipulator will carry the egg to the place between the light room and color camera. The interior color image of the egg is captured and processed. The depth of greenness of eggs is judged and graded by the formula of connection between Haugh, which reflects the depth of greenness of eggs, and HIS of color in yolk. According to the grading information the robot motion controller controls the manipulator to put the eggs in corresponding eggcrate.In this paper, the ratio between one rank moments in image and total area更多还原