饲喂辅助机器人的智能推料方法与试验研究  

Experimental Study on Intelligent Pushing Method of Feeding Assistant Robot

作　　者：张勤[1] 胡嘉辉 任海林 ZHANG Qin;HU Jiahui;REN Hailin(School of Mechanical and Automotive Engineering,South China University of Technology,Guangzhou 510640,Guangdong,China)

机构地区：[1]华南理工大学机械与汽车工程学院,广东广州510640

出　　处：《华南理工大学学报（自然科学版）》2022年第6期111-120,共10页Journal of South China University of Technology(Natural Science Edition)

基　　金：广东省现代农业产业共性关键技术研发创新团队项目(2019KJ129)。

摘　　要：在中、大型牧场的奶牛饲喂中,现有的推料机器人功能单一,只能完成均匀推料功能,无法满足牛只的个性化采食需求。针对这个问题,文中提出了一种饲喂辅助机器人的智能推料方法。首先,引入二维码标牌作为牛颈枷的定位标签,基于YOLOv4深度学习模型得到二维码标牌和牛头的检测框区域;然后,通过预处理和预测算法对二维码标牌的检测框区域进行实时识别与跟踪,并将二维码标牌与牛头进行匹配,确定觅食奶牛所在的采食颈枷位置;最后,根据牛-码位置匹配信息、余料分布信息控制机器人推板改变推料角度,实现个性化推料,以满足奶牛个体自由采食需求。实验结果表明:所提出的智能推料方法对二维码的识别率为96.0%;在二维码标牌连续丢失60帧的情况下,对二维码的跟踪预测精度在±2.85%以内;每帧图像在图形处理器中处理的时间为34.4 ms;智能送料的准确率为100%,满足牛舍复杂环境下机器人智能推料的实时性要求。In the feeding of dairy cows in medium and large-scale pastures,the existing pushing robot has a single function,which can only complete the uniform feeding function,and can not meet the personalized feeding needs of each cow.To solve this problem,this paper proposed an intelligent pushing method of feeding assistant robot.Firstly,the two-dimensional code label is introduced as the positioning label of cow neck rail,and the detection frame area of the two-dimensional code label and the cow's head is obtained based on the YOLOv4 deep learning model.Then,the detection frame area of the two-dimensional code label is identified and tracked in real time through the preprocessing and prediction algorithm,and the two-dimensional code label is matched with the cow head to determine the feeding neck rail position of the foraging cow.Finally,based on the cow-rail position mat-ching information and residual material distribution information,the robot push plate is controlled to change the push angle to realize personalized push,so as to meet the individual feeding needs of dairy cows.The experimental results show that the recognition rate of the proposed intelligent pushing method is 96.0%.When the two-dimensional code label loses 60 frames continuously,the tracking and prediction accuracy of the two-dimensional code is within±2.85%;the processing time of each frame image in the graphics processor is 34.4ms;the accuracy of intelligent pushing is 100%,which meets the real-time requirement of robot intelligent pushing in complex environment of cowshed.

关 键 词：机器人 智能推料 奶牛饲喂 二维码 深度学习 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]