基于 DRN-YOLO 融合模型的奶牛乳头检测方法研究

Research on bovine teat detection method based on the DRN-YOLO fusion model

摘要

[目的]在自动化挤奶系统中,奶牛乳头精确识别极为关键。由于挤奶厅环境复杂,奶牛乳头检测存在精度差、效率低等问题,这将影响挤奶操作准确性和工作效率,降低养殖效益。本文旨在提出一种用于奶牛乳头检测的改进 DRN-YOLO 模型。[方法]该模型采用 DRNet 模块优化 YOLOv8 骨干网络,提升特征提取能力,并且在路径增强网络中增加检测特征尺度和空间金字塔池化(spatial pyramid pooling,SPP)结构提升特征图特征提取的方法,使模型可从更多维度分析、提取图像特征,增强奶牛乳头识别效果,优化参数配置。为测试 DRN-YOLO 性能,用 600 张奶牛乳头图像构成测试集进行测试。[结果]与常用目标检测模型(YOLOv9 模型和 Faster RCNN 模型)进行比较,DRN-YOLO 模型的精确率为 99.07%,召回率为 98.88%,平均精度均值(mean average precision,mAP)为 98.41%,F₁得分 98.97%。相较于 YOLOv8,DRN-YOLO 模型各种检测指标数值均提高2%左右;相较于 FasterRCNN,DRN-YOLO 模型检测时间缩短了542ms,且检测精度提高1%。[结论]改进的 DRN-YOLO 模型能够实现对奶牛乳头的准确检测。

Abstract

[Objectives]In automated milking systems, accurate identification of cow teats is crucial. Due to the complex environment of milking parlors, existing cow teat detection methods faced challenges such as poor accuracy and low efficiency, which negatively impacted milking operation precision, workflow efficiency, and farm profitability. To address these issues, an improved DRN-YOLO model for cow teat detection was proposed in this article. [Methods] The YOLOv8 backbone network was optimized using a DRNet module to enhance feature extraction capabilities. Additionally, the detection feature scale and a spatial pyramid pooling(SPP) structure were incorporated into the path enhancement network, enabling multi-dimensional analysis and feature extraction from images. This improved teat recognition performance and optimized parameter configuration. To evaluate the DRN-YOLO model, a test set composed of 600 cow teat images was utilized. [Results]Comparative experiments with widely-used object detection models(YOLOv9 and Faster R-CNN)demonstrated that the proposed model achieved a precision rate of 99.07%, recall rate of 98.88%, mean average precision(mAP)of 98.41%, and an F₁-score of 98.97%. Compared to YOLOv8, all detection metrics of the DRN-YOLO model improved by approximately 2%. Furthermore, the detection time was reduced by 542 ms compared to Faster R-CNN, while detection accuracy surpassed Faster R-CNN by 1%. [Conclusions]The improved DRN-YOLO model enabled accurate cow teat detection.

关键词

奶牛乳头检测 / 深度学习 / YOLOv8 / 改进的 DRN-YOLO 模型 / 目标检测

Key words

cow teat detection / deep learning / YOLOv8 / improved DRN-YOLO model / object detection

参考文献

[1] 高然. 挤奶机器人对奶牛场挤奶管理能力的提升[J]. 中国乳业, 2024(9): 26-29.

[2] Gao R. Milking robots enhance the milking management capabilities of dairy farms[J]. China Dairy, 2024(9): 26-29(in Chinese with English abstract).

[3] Múnera-Bedoya O D, Cassoli L D, Machado P F, et al. Influence of attitudes and behavior of milkers on the hygienic and sanitary quality of milk[J]. PLoS One, 2017, 12(9): e0184640.

[4] 马慧, 丛慧敏, 钟景田. 机器人挤奶系统对奶牛行为、健康状况及生产性能影响的研究进展[J]. 中国奶牛, 2017(6): 59-63.

[5] Ma H, Cong H M, Zhong J T. Research progress on the effect of automatic milking system on behavior, health status and performance of dairy cows[J]. China Dairy Cattle, 2017(6): 59-63(in Chinese with English abstract).

[6] 李昊原, 付彤, 李转见, 等. 规模化奶牛养殖场动物福利水平评价[J]. 家畜生态学报, 2023, 44(12): 54-59.

[7] Li H Y, Fu T, Li Z J, et al. Assessment of animal welfare level in large-scaledairy farms[J]. Journal of Domestic Animal Ecology, 2023, 44(12): 54-59(in Chinese with English abstract).

[8] 夏事成, 王磊, 王成军, 等. 基于改进 YOLOv5 的奶牛乳头目标检测与数据集构建[J]. 安徽科技学院学报, 2023, 37(5): 64-70.

[9] Xia S C, Wang L, Wang C J, et al. Nipple subject detection and dataset construction based on improved YOLOv5[J]. Journal of Anhui Science and Technology University, 2023, 37(5): 64-70(in Chinese with English abstract).

[10] Rodriguez F A N, Lopes M A, Lima A L R, et al. Comparative analysis of milking and behavior characteristics of multiparous and primiparous cows in robotic systems[J]. Anais da Academia Brasileira de Ciencias, 2024, 96(3): e20221078.

[11] Rastogi A, Ryuh B S. Teat detection algorithm: YOLO vs. Haar-cascade[J]. Journal of Mechanical Science and Technology, 2019, 33(4): 1869-1874.

[12] 王政, 宋怀波, 王云飞, 等. 奶牛运动行为智能监测研究进展与技术趋势[J]. 智慧农业(中英文), 2022, 4(2): 36-52.

[13] Wang Z, Song H B, Wang Y F, et al. Research progress and technology trend of intelligent monitoring of dairy cow motion behavior[J]. Smart Agriculture, 2022, 4(2): 36-52(in Chinese with English abstract).

[14] 王彦超, 康熙, 李孟飞, 等. 基于改进 YOLOv3-tiny 的奶牛乳房炎自动检测方法[J]. 农业机械学报, 2021, 52(增刊 1): 276-283.

[15] Wang Y C, Kang X, Li M F, et al. Automatic detection method for dairy cow mastitis based on improved YOLOv3-tiny[J]. Transactions of the Chinese Society for Agricultural Machinery, 2021, 52(Supp1): 276-283(in Chinese with English abstract).

[16] Lu Z H, Zhao M F, Luo J, et al. Automatic teat detection for rotary milking system based on deep learning algorithms[J]. Computers and Electronics in Agriculture, 2021, 189: 106391.

[17] Porter I R, Wieland M, Basran P S. Feasibility of the use of deep learning classification of teat-end condition in Holstein cattle[J]. Journal of Dairy Science, 2021, 104(4): 4529-4536.

[18] 刘亚楠, 沈明霞, 刘龙申, 等. 基于机器视觉的母猪哺乳行为监测方法研究[J]. 南京农业大学学报, 2022, 45(2): 404-411. DOI: 10.7685/jnau.202106027.

[19] Liu Y N, Shen M X, Liu L S, et al. Research on monitoring method of sow's nursing behaviour based on machine vision[J]. Journal of Nanjing Agricultural University, 2022, 45(2): 404-411(in Chinese with English abstract).

[20] Jia Q X, Yang J C, Han S J, et al. CAMLLA-YOLOv8n: cow behavior recognition based on improved YOLOv8n[J]. Animals, 2024, 14(20): 3033.

[21] 陆舟, 沈明霞, 刘龙申, 等. 基于轻量化网络与注意力机制的育肥猪采食行为识别方法研究[J]. 南京农业大学学报, 2023, 46(4): 802-812. DOI: 10.7685/jnau.202208023.

[22] Lu Z, Shen M X, Liu L S, et al. Study on feeding behavior recognition method of fattening pigs based on lightweight networks and attention mechanism[J]. Journal of Nanjing Agricultural University, 2023, 46(4): 802-812(in Chinese with English abstract).

[23] 张哲鼎, 王超柱, 陈骏, 等. 基于 YOLOX 和 DeepSORT 的淡水鱼行为检测算法研究[J]. 智能化农业装备学报(中英文), 2025, 6(3): 88-97.

[24] Zhang Z D, Wang C Z, Chen J, et al. Research on freshwater fish behavior detection algorithm based on YOLOX and DeepSORT[J]. Journal of Intelligent Agricultural Mechanization(in Chinese and English), 2025, 6(3): 88-97(in Chinese with English abstract).

[25] 何东健, 刘建敏, 熊虹婷, 等. 基于改进 YOLOv3 模型的挤奶奶牛个体识别方法[J]. 农业机械学报, 2020, 51(4): 250-260.

[26] He D J, Liu J M, Xiong H T, et al. Individual identification of dairy cows based on improved YOLOv3[J]. Transactions of the Chinese Society for Agricultural Machinery, 2020, 51(4): 250-260(in Chinese with English abstract).

[27] Xu B B, Wang W S, Falzon G, et al. Automated cattle counting using Mask R-CNN in quadcopter vision system[J]. Computers and Electronics in Agriculture, 2020, 171: 105300.

[28] 薛月菊, 朱勋沐, 郑婵, 等. 基于改进 FasterR-CNN 识别深度视频图像哺乳母猪姿态[J]. 农业工程学报, 2018, 34(9): 189-196.

[29] Xue Y J, Zhu X M, Zheng C, et al. Lactating sow postures recognition from depth image of videos based on improved FasterR-CNN[J]. Transactions of the Chinese Society of Agricultural Engineering, 2018, 34(9): 189-196(in Chinese with English abstract).

[30] Zheng C, Zhu X M, Yang X F, et al. Automatic recognition of lactating sow postures from depth images by deep learning detector[J]. Computers and Electronics in Agriculture, 2018, 147: 51-63.

[31] 徐彦威, 李军, 董元方, 等. YOLO 系列目标检测算法综述[J]. 计算机科学与探索, 2024, 18(9): 2221-2238.

[32] Xu Y W, Li J, Dong Y F, et al. Survey of development of YOLO object detection algorithms[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(9): 2221-2238(in Chinese with English abstract).

[33] Huang Z C, Wang J L, Fu X S, et al. DC-SPP-YOLO: dense connection and spatial pyramid pooling based YOLO for object detection[J]. Information Sciences, 2020, 522: 241-258.

[34] Singh D, Kumar V, Kaur M. Densely connected convolutional networks-based COVID-19 screening model[J]. Applied Intelligence, 2021, 51(5): 3044-3051.

[35] 庄前伟, 王志明, 吴龙贻, 等. 基于改进 SOLOv2 的穴盘幼苗图像分割方法[J]. 南京农业大学学报, 2023, 46(1): 200-209. DOI: 10.7685/jnau.202203024.

[36] Zhuang Q W, Wang Z M, Wu L Y, et al. Image segmentation method of plug seedlings based on improved SOLOv2[J]. Journal of Nanjing Agricultural University, 2023, 46(1): 200-209(in Chinese with English abstract).

[37] 邱天衡, 王玲, 王鹏, 等. 基于改进 YOLOv5 的目标检测算法研究[J]. 计算机工程与应用, 2022, 58(13): 63-73.

[38] Qiu T H, Wang L, Wang P, et al. Research on object detection algorithm based on improved YOLOv5[J]. Computer Engineering and Applications, 2022, 58(13): 63-73(in Chinese with English abstract).

[39] Ji S J, Ling Q H, Han F. An improved algorithm for small object detection based on YOLOv4 and multi-scale contextual information[J]. Computers and Electrical Engineering, 2023, 105: 108490.

基金资助

山东省科技型中小企业创新能力提升工程项目(2023TSGC059901)

日喀则市科技计划项目课题(RKZ2024ZY-03)