基于数据增强深度学习的苹果花检测方法研究

doi:10.13733/j.jcam.issn.20955553.2022.05.022

摘要/Abstract

摘要： 为提高现代化果园化学疏花的工作效率以及促进疏花机械的研发，针对果园生产中传统识别方法检测开花强度不全面、效率低等问题，提出一种基于数据增强YOLOv4深度学习的苹果花检测方法。首先，搭建YOLOv4网络模型，以CSPDarknet53框架为主干特征提取网络，然后为了减少样本数据不均和数量不足的影响，进行离线和在线数据增强的YOLOv4方法研究。最后，利用人工标注的苹果花图像对模型进行微调和训练，将所提出的方法与YOLOv3、YOLOv4和Faster R-CNN的检测模型进行对比试验，并讨论了不同拍摄情况下苹果花的检测性能，验证了该方法的有效性。试验结果表明，所提出的数据增强的YOLOv4的方法检测苹果花的AP值为99.76%；分别比Faster R-CNN、YOLOv3和YOLOv4提高了2.53%、14.56%和5.08%。相较于其他检测方法，该方法具有更高的苹果花检测精准度。

关键词: 苹果花检测, 深度学习, 数据增强, YOLOv4, 开花强度, 现代化果园

Abstract: In order to improve the efficiency of chemical flower thinning in modern orchards, promote the development of flower thinning machinery, as well as to address the problems of incomplete and low efficiency of traditional identification methods for detecting flowering intensity in orchard production, a dataenhanced YOLOv4 deep learningbased apple flower detection method was proposed. First, a YOLOv4 network model was built with the CSPDarknet53 framework as the backbone feature extraction network, and then offline and online data augmented YOLOv4 methods were investigated in order to reduce the effects of uneven sample data and insufficient quantity. Finally, the artificially annotated apple flower images were used to finetune and train the model, and the proposed method was compared with the detection models of YOLOv3, YOLOv4 and Faster R-CNN, after which the detection performance of apple flower under different shooting situations was discussed. The experimental results showed that the AP value of the proposed dataenhanced YOLOv4 method for detecting apple flower was 99.76%. It improved 2.53%, 14.56% and 5.08% over Faster R-CNN, YOLOv3 and YOLOv4, respectively. Compared with other detection methods, this method has higher precision in apple flower detection.

Key words: apple flower detection, deep learning, data enhancement, YOLOv4, flowering intensity, modern orchard

中图分类号:

S24： TP391.4

陈国防, 陈兆英, 王玉亮, 王金星, 范国强, 李汉卿, . 基于数据增强深度学习的苹果花检测方法研究[J]. 中国农机化学报, 2022, 43(5): 148-155.

Chen Guofang, Chen Zhaoying, Wang Yuliang, Wang Jinxing, Fan Guoqiang, Li Hanqing.. Research on detection method of apple flower based on dataenhanced deep learning[J]. Journal of Chinese Agricultural Mechanization, 2022, 43(5): 148-155.

参考文献

［1］ Link H. Significance of flower and fruit thinning on fruit quality ［J］. Plant Growth Regulation, 2000, 31(1-2): 17-26.
［2］ Dennis F G. The history of fruit thinning ［J］. Plant Growth Regulation, 2000, 31(1-2):1-16.
［3］杨刚, 张杰, 田佶, 等. 3种疏除剂在梨树疏花疏果上的效应［J］. 北京农学院学报, 2017, 32(1): 18-23.
Yang Gang, Zhang Jie, Tian Ji, et al. Study on thinning effects of three thinning agents to pear flowers and fruits ［J］. Journal of Beijing University of Agriculture, 2017, 32(1): 18-23.
［4］ Kapach K, Barnea E, Mairon R, et al. Computer vision for fruit harvesting robotsstate of the art and challenges ahead ［J］. International Journal of Computational Vision And Robotics, 2012.
［5］张星, 高巧明, 潘栋, 等. 基于改进YOLOv3的田间复杂环境下菠萝拾捡识别研究［J］. 中国农机化学报, 2021, 42(1): 201-206.
Zhang Xing, Gao Qiaoming, Pan Dong, et al. Picking recognition research of pineapple in complex field environment based on improved YOLOv3 ［J］. Journal of Chinese Agricultural Mechanization, 2021, 42(1): 201-206.
［6］李龙, 彭彦昆, 李永玉, 等. 基于纹理和梯度特征的苹果伤痕与果梗/花萼在线识别［J］. 农业机械学报, 2018, 49(11): 328-335.
Li Long, Peng Yankun, Li Yongyu, et al. Online identification of apple scarring and stems/calyxes based on texture and edge gradient features ［J］. Transactions of the Chinese Society for Agricultural Machinery, 2018, 49(11): 328-335.
［7］ Dias P A, Tabb A, Medeiros H. Apple flower detection using deep convolutional networks ［J］. Computers in Industry, 2018, 99: 17-28.
［8］王丹丹, 何东健. 基于R-FCN深度卷积神经网络的机器人疏果前苹果目标的识别［J］. 农业工程学报, 2019, 35(3): 156-163.
Wang Dandan, He Dongjian.Recognition of apple targets before fruits thinning by robot based on R-FCN deep convolution neural network ［J］. Transactions of the Chinese Society of Agricultural Engineering. 2019, 35(3): 156-163.
［9］ Yu Y, Zhang K, Yang L, et al. Fruit detection for strawberry harvesting robot in nonstructural environment based on Mask R-CNN ［J］. Computers and Electronics in Agriculture, 2019, 163: 104846.
［10］熊俊涛, 郑镇辉, 梁嘉恩, 等. 基于改进YOLOv3网络的夜间环境柑橘识别方法［J］. 农业机械学报, 2020, 51(4): 199-206.
Xiong Juntao, Zheng Zhenhui, Liang Jiaen, et al. Citrus detection method in night environment based on improved YOLOv3 network ［J］. Transactions of the Chinese Society for Agricultural Machinery, 2020, 51(4): 199-206.
［11］ Redmon J, Farhadi A. YOLOv3: An incremental improvement ［J］. arXiv eprints, 2018.
［12］ Wang C Y, Liao H, Wu Y H, et al. CSPNet: A new backbone that can enhance learning capability of CNN ［C］. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). IEEE, 2020.
［13］吕铄, 蔡烜, 冯瑞. 基于改进损失函数的YOLOv3网络［J］. 计算机系统应用, 2019, 28(2): 1-7.
Lü Shuo, Cai Xuan, Feng Rui. YOLOv3 network based on improved loss function ［J］. Computer Systems & Applications, 2019, 28(2): 1-7.
［14］ Zheng Z, Wang P, Liu W, et al. DistanceIoU Loss: Faster and better learning for bounding box regression ［J］. arXiv, 2019.
［15］ H Rezatofighi, Tsoi N, JY Gwak, et al. Generalized intersection over union: A metric and a loss for bounding box regression ［C］. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2019.
［16］ Yun S, Han D, Chun S , et al. CutMix: Regularization strategy to train strong classifiers with localizable features ［C］. International Conference on Computer Vision, 2019.
［17］张宏鸣, 付振宇, 韩文霆, 等. 基于改进YOLO的玉米幼苗株数获取方法［J］. 农业机械学报, 2021, 52(4): 221-229.
Zhang Hongming, Fu Zhenyu, Han Wenting, et al. Detection method of maize seedlings number based on improved YOLO ［J］. Transactions of the Chinese Society for Agricultural Machinery, 2021, 52(4): 221-229.
［18］金耀, 何秀文, 万世主, 等. 基于YOLOv3的生猪个体识别方法［J］. 中国农机化学报, 2021, 42(2): 178-183.
Jin Yao, He Xiuwen, Wan Shizhu, et al. Individual pig identification method based on YOLOv3 ［J］. Journal of Chinese Agricultural Mechanization, 2021, 42(2): 178-183.

[1]	韦锦, 李正强, 许恩永, 蒙艳玫, 韦和钧, 武豪. 基于DA2-YOLOv4算法绿篱识别研究[J]. 中国农机化学报, 2022, 43(9): 122-130.
[2]	李东升, 胡文泽, 兰玉彬, 范明洪, 李翠云, 朱俊科. 深度学习在杂草识别领域的研究现状与展望[J]. 中国农机化学报, 2022, 43(9): 137-144.
[3]	王晓婷, 赵展, 王阳, 李林, . 基于改进Mask R-CNN的植物表型智能检测算法[J]. 中国农机化学报, 2022, 43(8): 151-157.
[4]	孙逸飞, 柳平增, 张艳, 宋长青, 张大雷, 马学文. 基于Sentinel-2A遥感影像的潍坊市冬小麦种植面积提取研究[J]. 中国农机化学报, 2022, 43(7): 98-105.
[5]	张三林, 张立萍, 郑威强, 郭壮, 付子强. 基于YOLOv5的核桃品种识别与定位[J]. 中国农机化学报, 2022, 43(7): 167-172.
[6]	史方青, 王虎林, 黄华. 基于卷积神经网络的马铃薯芽眼检测识别研究[J]. 中国农机化学报, 2022, 43(6): 159-165.
[7]	曹金凤, 沈大港, 郭继鸿, 刘鹏, 李策, 兰添贺. 基于CNN和SVM分类优化的大蒜鳞芽朝向识别研究[J]. 中国农机化学报, 2022, 43(5): 134-139.
[8]	雷雪梅, 张光强, 姚旗, 刘伟渭, 邱帅. 基于卷积神经网络的农机图像自动识别研究[J]. 中国农机化学报, 2022, 43(5): 140-147.
[9]	张燕斌, , 杜健民, 王圆, , 皮伟强, 高新超. 基于无人机高光谱遥感和3D-ResNet的荒漠草原地物分类[J]. 中国农机化学报, 2022, 43(4): 66-73.
[10]	何颖, 陈丁号, 彭琳. 基于改进YOLOv5模型的经济林木虫害目标检测算法研究[J]. 中国农机化学报, 2022, 43(4): 106-115.
[11]	项新建, 宋晓敏, 郑永平, 王海波, 方政洋. 基于MobileNet-YOLO的嵌入式人脸检测研究#br# #br# [J]. 中国农机化学报, 2022, 43(4): 124-130.
[12]	董萍, 卫梦华, 时雷, 郭伟. 迁移学习在玉米叶片病害识别中的研究与应用[J]. 中国农机化学报, 2022, 43(3): 146-152.
[13]	张继成, 李德顺. 基于深度残差学习的成熟草莓识别方法[J]. 中国农机化学报, 2022, 43(2): 136-142.
[14]	王权顺, 吕蕾, 黄德丰, 付思琴, 余华云. 基于改进YOLOv4算法的苹果叶部病害缺陷检测研究[J]. 中国农机化学报, 2022, 43(11): 182-187.
[15]	陈继清, 韦德鹏, 龙腾, 罗天, 王桦彬. 基于卷积神经网络的害虫分类[J]. 中国农机化学报, 2022, 43(11): 188-194.