[1]方国文,何超,王鑫泽.基于YOLOv8n的轻量级巴旦木果实识别方法[J].江苏农业学报,2024,(09):1662-1670.[doi:doi:10.3969/j.issn.1000-4440.2024.09.010]
　FANG Guowen,HE Chao,WANG Xinze.Lightweight almond fruit recognition method based on YOLOv8n[J].,2024,(09):1662-1670.[doi:doi:10.3969/j.issn.1000-4440.2024.09.010]
点击复制

基于YOLOv8n的轻量级巴旦木果实识别方法()
分享到：

江苏农业学报[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2024年09期
页码:
1662-1670
栏目:
农业信息工程
出版日期:
2024-09-30

文章信息/Info

Title:
Lightweight almond fruit recognition method based on YOLOv8n
作者:
方国文; 何超; 王鑫泽
(西南林业大学机械与交通学院,云南昆明650000)
Author(s):
FANG Guowen; HE Chao; WANG Xinze
(College of Mechanical and Transportation Engineering, Southwest Forestry University, Kunming 650000, China)
关键词:
巴旦木; 果实识别; BiFPN; ContextGuide; MPDIoU损失函数; YOLOv8n
Keywords:
almond; fruit recognition; BiFPN; ContextGuide; MPDIoU loss function; YOLOv8n
分类号:
TP391.4
DOI:
doi:10.3969/j.issn.1000-4440.2024.09.010
文献标志码:
A
摘要:
在果园环境下,快速精准识别巴旦木果实对提升巴旦木采摘机器人的作业精度和效率至关重要。为减少果园场景中因树叶遮挡或果实重叠导致的巴旦木果实漏检现象,降低计算量和参数量,提高果实识别模型的性能和准确度,本研究在YOLOv8n模型的基础上,利用ContextGuide模块替换原模型中主干网络（Backbone）部分基本构成单元C2f中的Bottleneck模块,利用BiFPN模块替代原模型中颈部网络（Neck）部分中的PANet模块,同时引入MPDIoU损失函数替换原模型中的CIoU损失函数,提出了一种改进的轻量级巴旦木果实检测模型(YOLOv8n-BCG)。并利用公开的巴旦木影像数据集对优化后的模型性能进行比较分析。结果表明,改进后模型参数量仅为1.528 M,平均精度值(mAP0.50∶0.95)为69.7%,相比于原YOLOv8n模型提升0.5个百分点。与YOLOv5s、YOLOv5n、YOLOv7-tiny、Faster R-CNN等模型相比,YOLOv8n-BCG模型具有更低的浮点计算量和更高的检测精度值。本研究结果可为高效的巴旦木果实采摘机器人自动化作业提供技术支持。
Abstract:
In the orchard environment, rapid and accurate identification of almond fruits is very important to improve the operation accuracy and efficiency of almond picking robots. In order to reduce the missed detection of almond fruits caused by leaf occlusion or fruit overlap in the orchard scene, reduce the amount of calculation and parameters, and improve the performance and accuracy of the fruit recognition model, based on the YOLOv8n model, this study used the ContextGuide module to replace the Bottleneck module in the basic component unit C2f of the Backbone part of the original model, used the BiFPN module to replace the PANet module in the Neck part of the original model, and introduced the MPDIoU loss function to replace the CIoU loss function in the original model. An improved lightweight almond fruit detection model (YOLOv8n-BCG) was proposed. The performance of the optimized model was compared and analyzed by using the public almond image data set. The results showed that the number of parameters of the improved model was only 1.528 M, and the mean average precision (mAP0.50∶0.95) was 69.7%, which was 0.5 percentage points higher than that of the original YOLOv8n model. Compared with YOLOv5s, YOLOv5n, YOLOv7-tiny and Faster R-CNN models, YOLOv8n-BCG model had lower floating-point calculation and higher detection accuracy. The results of this study can provide technical support for efficient automatic operation of almond picking robots.

参考文献/References:

[1]石鑫,杨豫新,牛长河,等. 滚筒刷式巴旦木脱青皮机的设计与试验[J]. 西北农林科技大学学报(自然科学版),2023,51(12):143-154.
[2]金寿祥,周宏平,姜洪喆,等. 采摘机器人视觉系统研究进展[J]. 江苏农业学报,2023,39(2):582-595.
[3]刘冬. 数字化与智能化在农业机械工程设计中的应用[J]. 中国农业资源与区划,2023,44(7):48,65.
[4]陈礼鹏,穆龙涛,刘浩洲,等. 基于猕猴桃果萼图像的多目标果实识别方法[J]. 计算机工程与设计,2018,39(6):1738-1744.
[5]ABDULLAH M S, HASAN R M A, RAHMAN A N, et al. Versatile recognition of graphene layers from optical images under controlled illumination through green channel correlation method[J]. Nanotechnology,2023,34(44):1899-1904.
[6]贾伟宽,赵德安,阮承治,等. 苹果采摘机器人夜间图像降噪算法[J]. 农业工程学报,2015,31(10):219-226.
[7]李斌,韩昭洋,王秋,等. 基于高光谱成像技术的枇杷碰伤等级检测研究[J]. 光谱学与光谱分析,2023,43(6):1792-1799.
[8]MAR A, HILMY B, SERGIO V, et al. Object detection and tracking on UAV RGB videos for early extraction of grape phenotypic traits[J]. Computers and Electronics in Agriculture,2023,211(20):347-353.
[9]CHEN B, SHI S, CHEN B, et al. True color 3D imaging optimization with missing spectral bands based on hyperspectral LiDAR[J]. Optics Express,2021,29(13):20406-20422.
[10]李昕,李立君,高自成,等. 改进类圆随机Hough变换及其在油茶果实遮挡识别中的应用[J]. 农业工程学报,2013,29(1):164-170.
[11]张彦斐,刘茗洋,宫金良,等. 基于两级分割与区域标记梯度Hough圆变换的苹果识别[J]. 农业工程学报,2022,38(19):110-121.
[12]刘安旭,黎向锋,刘晋川,等. 改进卷积空间传播网络的单目图像深度估计[J]. 电子测量技术,2021,44(23):78-85.
[13]吕伟,宋轩,杨欢. 基于深度学习和多源遥感数据的玉米种植面积提取[J]. 江苏农业科学,2023,51(23):171-178.
[14]鲍彤,罗瑞,郭婷,等. 基于BERT字向量和TextCNN的农业问句分类模型分析[J]. 南方农业学报,2022,53(7):2068-2076.
[15]翟先一,魏鸿磊,韩美奇,等. 基于改进YOLO卷积神经网络的水下海参检测[J]. 江苏农业学报,2023,39(7):1543-1553.
[16]向俊,严恩萍,姜镓伟,等. 基于全卷积神经网络和低分辨率标签的森林变化检测研究[J]. 南京林业大学学报(自然科学版),2024,48(1):187-195.
[17]阮子行,黄勇,王梦,等. 基于改进卷积神经网络的番茄品质分级方法[J]. 江苏农业学报,2023,39(4):1005-1014.
[18]DEBANIRANJAN M, NIVA D, KUMAR K M. Deep neural network based fruit identification and grading system for precision agriculture[J]. Proceedings of the Indian National Science Academy,2022,88(2):228-239.
[19]TANG Z, LIU J J, CHEN Z Y, et al. Improved Pest-YOLO:real-time pest detection based on efficient channel attention mechanism and transformer encoder[J]. Ecological Informatics,2023,78(14):891-898.
[20]郭明月,刘雅晨,李伟夫,等. 基于视频跟踪算法的果园猕猴桃产量实时预估[J]. 农业机械学报,2023,54(6):178-185.
[21]王大方,刘磊,曹江,等. 基于空洞空间池化金字塔的自动驾驶图像语义分割方法[J]. 汽车工程,2022,44(12):1818-1824.
[22]李恒,南新元,高丙朋,等. 一种基于GhostNet的绿色类圆果实识别方法[J]. 江苏农业学报,2023,39(3):724-731.
[23]SUE R. GhostNets:go global,and local[J]. Artlink,2013,33(2):70-71.
[24]赵升,赵黎. 基于双向特征金字塔和深度学习的图像识别方法[J]. 哈尔滨理工大学学报,2021,26(2):44-50.
[25]储鑫,李祥,罗斌,等. 基于改进YOLOv4算法的番茄叶部病害识别方法[J]. 江苏农业学报,2023,39(5):1199-1208.
[26]李运寰,闻继伟,彭力. 高帧率的轻量级孪生网络目标跟踪[J]. 计算机科学与探索,2022,16(6):1405-1416.
[27]王琛,林威,胡良鹏,等. 分体式飞行汽车全自主对接导引系统设计与验证[J]. 浙江大学学报(工学版),2023,57(12):2345-2355.
[28]孔维刚,李文婧,王秋艳,等. 基于改进YOLOv4算法的轻量化网络设计与实现[J]. 计算机工程,2022,48(3):181-188.
[29]LIN C L, WANG T Y, DONG S Y, et al. Hybrid convolutional network combining 3D depthwise separable convolution and receptive field control for hyperspectral image classification[J]. Electronics,2022,11(23):3992-3997.
[30]代云,卢明,何婷,等. 基于改进型YOLO的密集环境下槟榔果实的快速识别方法[J]. 食品与机械,2023,39(4):83-88.
[31]YANG S Z, WANG W, GAO S, et al. Strawberry ripeness detection based on YOLOv8 algorithm fused with LW-Swin Transformer[J]. Computers and Electronics in Agriculture,2023,215(5):1234-1240.
[32]TAN M, PANG R, LE Q. EfficientDet:scalable and efficient object detection[C]//IEEE. 2020 IEEE/CVF conference on computer vision and pattern recognition(CVPR). Washington:IEEE Computer Society,2020:10778-10787.
[33]WU T Y, TANG S, ZHANG R, et al. CGNet:a light-weight context guided network for semantic segmentation[J]. IEEE Transactions on Image Processing，2021,30:1169-1179.
[34]HENDERSON P, FERRARI V. End-to-end training of object class detectors for mean average precision[C]//LAI S H, LEPETIT V, NISHIMO K, et al. Computer Vision-ACCV2016. Cham,SW:Springer,2016:198-213.

备注/Memo

备注/Memo:
收稿日期：2023-12-04基金项目：云南省高层次人才项目(YNWR-QNBJ-2018-066、YNQR-CYRC-2019-001)作者简介：方国文(2001-),女,安徽合肥人,硕士研究生,主要研究方向为图像识别。(E-mail)fgw09182023@163.com通讯作者：何超,(Tel)15887130986；(E-mail)hcsmile@163.com
更新日期/Last Update: 2024-11-17