基于机器视觉的家禽机器人掏膛关键技术基础研究

Poultry eviscerating is the most difficult to realize mechanical automation in poultry slaughtering. At present, artificial auxiliary assembly line was often used in China, and the foreign eviscerating equipment was not suitable for poultry slaughtering of China, which was likely to cause the damage of vulnerable organs. In order to ensure the integrity of poultry visceral organs when the poultry robot work, this project integrated application of mathematics, biology, and machine vision technology, set up stereo vision system to acquire poultry carcass, proposed poultry carcass three-dimensional information display and stereo matching algorithm based on the image, set up three-dimensional carcass model, acquired poultry carcass body characteristic parameters change rule of different weight. The frozen milling imaging technology was applied to the preparation of poultry carcass cross section data set, and the 3d reconstruction of visceral organs in poultry abdominal cavity was carried out by combining image processing, registration and segmentation technology. A mathematical model was established to accurately reflect the relation between the morphological characteristics of poultry carcass size and the morphological characteristics of vulnerable internal organs in the abdominal cavity. The model of poultry robot cutting system was constructed, the special manipulator was designed, the cutting path was planned, the visceral damage evaluation system was established, and the optimal parameter combination of poultry cutting link was explored. The project results could provide necessary quantitative analysis methods for optimizing the effect of eviscerating, and had important practical significance for improving the automatic operation level of poultry slaughtering and improving the quality of food safety.

家禽内脏的掏出是家禽屠宰中最难实现机械自动化的环节，目前国内常采用人工辅助流水线，国外掏膛设备不适用于我国国情，且易造成内脏破损严重。本项目为确保家禽机器人掏膛时内脏的完整性，综合应用数学、生物学和机器视觉技术，搭建家禽胴体立体视觉系统，提出基于图像的胴体三维特征信息和立体匹配算法，建立胴体三维模型，获得不同体重的家禽胴体体尺特征参数的变化规律；采用冰冻铣削成像技术应用于家禽胴体断面数据集的制备，并结合图像处理、配准及分割技术进行家禽腹腔中内脏器官的三维重建；提出能够准确反映家禽胴体形态体尺特征与腹腔中脆弱内脏器官形态特征的约束关系，建立数学模型；基于家禽机器人掏膛系统构建，设计专用机械手，规划掏膛路径，建立内脏破损评价体系，探索家禽掏膛环节的参数最优组合。项目的完成可为优化掏膛效果提供必要的定量分析手段和方法，对于提高家禽屠宰自动化作业水平，改善食品安全质量也有着重要的实践意义。

家禽自动掏膛是制约家禽屠宰生产线的全自动化作业水平的瓶颈问题。目前主要以人工掏膛为主，大型企业采用的传统机械化掏膛方式不能适应家禽体型变化，且易造成家禽内脏的破损。目前为了确保机械手掏膛过程中内脏的完整性，减少脆弱器官的破损，利用机器视觉技术对家禽胴体进行了定位，并对内脏脆弱器官的位置进行了预测，为自动掏膛过程中机械手掏膛轨迹规划提供了新的理论指导。（1）实现了家禽胴体的三维重建，确定了家禽胴体在视觉空间中的位置及形态，并发现整体内脏相对胴体位置平均下移32mm。（2）建立了胴体和腹腔中内脏器官的位置关系，明确了家禽腹腔中脆弱内脏器官（心脏和肝脏）相对于外部胴体的位置及长度尺寸。（3）分析了家禽掏膛工艺需求，开展了家禽机器人掏膛试验，结果显示家禽掏膛效果有了显著提高，内脏残留率为6%左右，破损率也有了明显下降。项目成果有助于提高家禽加工作业的全自动化水平，为研发符合我国国情的家禽掏膛技术提供基础性资料。