基于改进YOLOv5的绿辣椒目标检测方法

doi:10.6048/j.issn.1001-4330.2024.12.018

新疆农业科学 ›› 2024, Vol. 61 ›› Issue (12): 3032-3041.DOI: 10.6048/j.issn.1001-4330.2024.12.018

• 土壤肥料·贮藏保鲜加工·园艺特产 • 上一篇下一篇

基于改进YOLOv5的绿辣椒目标检测方法

王震鲁¹(), 白涛¹^,²(), 李东亚¹, 戴硕¹, 陈珍¹

1.新疆农业大学计算机与信息工程学院,乌鲁木齐 830052
2.智能农业教育部工程研究中心/新疆农业信息化工程技术研究中心,乌鲁木齐 830052

收稿日期:2024-05-11 出版日期:2024-12-20 发布日期:2025-01-16
通信作者: 白涛(1979-),男,新疆乌鲁木齐人,副教授,硕士生导师,研究方向为农业大数据、数据挖掘,(E-mail)bt@xjau.edu.cn
作者简介:王震鲁(2000-),男,山东济宁人,硕士研究生,研究方向为计算机视觉、物联网,(E-mail)2425387367@qq.com
基金资助:
科技部科技创新2030重大项目“群体智能自主作业智慧农场”(2022ZD0115800);新疆维吾尔自治区重大科技专项“农场智能平台关键技术研究”(2022A02011-4);中央引导地方科技发展专项“智慧农业创新平台建设”(ZYYD2022B12);新疆维吾尔自治区高校基本科研业务费科研项目“农业大数据交换共享与可视化平台”(XJEDU2022J009)

Green chili pepper target detection method based on improved YOLOv5

WANG Zhenlu¹(), BAI Tao¹^,²(), LI Dongya¹, DAI Shuo¹, CHEN Zhen¹

1. College of Computer and Information Engineering, Xinjiang Agricultural University, Urumqi 830052, China
2. Engineering Research Center for Intelligent Agriculture of Ministry of Education/Xinjiang Research Center for Agricultural Information Technology, Urumqi 830052, China

Received:2024-05-11 Published:2024-12-20 Online:2025-01-16
Supported by:
S &T Innovation 2030 Major Project of Ministry of Science and Technology "Group Intelligent Independent Operation of Intelligent Farm"(2022ZD0115800);Major Scientific R & D Program Project of Xinjiang Uygur Autonomous Region "Research on Key Technologies of Farm Intelligent Platform"(2022A02011-4);Central Government Guiding the Local Science and Technology Development Special Fund Project(ZYYD2022B12);Basic Scientific Research Project for Universities in Xinjiang Uygur Autonomous Region "Agricultural Big Data Exchange, Sharing and Visualization Platform"(XJEDU2022J009)

摘要/Abstract

摘要：

【目的】 使用机器视觉对绿辣椒的精准识别是实现辣椒智能化采摘的重要前提,研究自然条件下辣椒遮挡情况、绿色辣椒及准确识别的方法,为辣椒智能化采摘机器人的精准识别提供技术支持。【方法】 提出一种基于改进YOLOv5辣椒目标检测模型,在YOLOv5主干网络加入CA(Coordinate Attention)注意力机制,以增强辣椒特征信息的提取,进一步增强对目标位置信息的提取;同时在特征融合网络中使用Bi-FPN结构,提高模型对遮挡辣椒的识别能力。【结果】 通过在自建辣椒数据集上进行训练,改进后的模型平均准确率达到91%,相比于研究其他所对比模型,改进模型的平均准确率更高。【结论】 基于改进YOLOv5的遮挡绿色辣椒的识别具有较高的准确性。

关键词: YOLOv5; CA注意力机制; Bi-FPN; 绿辣椒检测; 遮挡

Abstract:

【Objective】 Accurate recognition of green chili peppers using machine vision is an important prerequisite for realizing intelligent picking of chili peppers, so in view of the natural conditions of pepper occlusion, this study aims to accurately identify the problem.【Methods】 A chili pepper target detection model based on improved YOLOv5 was proposed, CA (Coordinate Attention) was added in YOLOv5 backbone network Attention mechanism in the YOLOv5 backbone network to enhance the extraction of chili pepper feature information and further enhance the extraction of target location information; meanwhile, a Bi-FPN structure was used in the feature fusion network to improve the model's ability to recognize occluded chili peppers.【Results】 By training on the self-constructed chili pepper dataset, the results showed that the improved model achieved an average accuracy of 91%, which was higher compared to the other models.【Conclusion】 The method proposed in this paper has high accuracy in recognizing occluded green chili peppers in natural environments, which can provide technical support for the accurate recognition of chili pepper intelligent picking robots.

Key words: YOLOv5; CA attention mechanism; Bi-FPN; green chili pepper detection; shading

中图分类号:

王震鲁, 白涛, 李东亚, 戴硕, 陈珍. 基于改进YOLOv5的绿辣椒目标检测方法[J]. 新疆农业科学, 2024, 61(12): 3032-3041.

WANG Zhenlu, BAI Tao, LI Dongya, DAI Shuo, CHEN Zhen. Green chili pepper target detection method based on improved YOLOv5[J]. Xinjiang Agricultural Sciences, 2024, 61(12): 3032-3041.

图/表 12

图1 辣椒数据集(部分)

Fig.1 Chili pepper dataset (partial)

图2 YOLOv5网络结构

Fig.2 YOLOv5 network structure diagram

图3 改进YOLOv5网络结构

Fig.3 Improvement of YOLOv5 network structure diagram

图4 坐标注意力模块图

Fig.4 Diagram of the coordinate attention module

图5 不同特征金字塔结构

Fig.5 Structure of different characteristic pyramids

表1 训练参数

Tab.1 Training parameters

参数项 Parameter items	数值 Numerical value
Epochs	150
Batch-size	64
Img-size	640
Learn_rate	0.01
优化器Optimizer	SGD

表2 不同模型对比结果

Tab.2 Comparison results of different models

模型 Model	平均精确率 Average accuracy (%)	参数量 Parameter quantity (M)	权重大小 Weight size (M)
Faster-RCNN	89.6	137.1	108.2
YOLOv5s	89.5	7.01	13.7
YOLOv7	90.4	37.1	71.3
YOLOv8s	90.3	11.1	21.5
Ours	91.0	7.1	14.0

图6 检测结果对比

Fig.6 Comparison of test results

表3 注意力模块对比

Tab.3 Comparison results of attention modules

模型 Model	平均精确率 Average accuracy (%)	精确率 Accuracy (%)	召回率 Recall (%)
YOLOv5s+GAM	89.2	87.3	87.7
YOLOv5s+EMA	89.6	87.8	87.7
YOLOv5s+CA	89.9	86.9	89.7

图7 不同注意力模块训练

Fig.7 Training chart for different attention modules

表4 消融试验对比

Tab.4 Comparative results of ablation experiments

模型 Model	平均精确率 Average accuracy (%)	召回率 Recall (%)	平均精确率 Average accuracy (%)
YOLOv5s	88.5	88.2	89.5
YOLOv5s+CA	86.9	89.7	89.9
YOLOv5s+BiFPN	87.2	88.8	90.7
Ours	89.4	87.5	91.0

图8 不同模块训练结果

Fig.8 Graph of training results of different modules

参考文献 26

[1]	张子峰. 我国辣椒产业发展现状、主要挑战与应对之策[J]. 北方园艺, 2023,(14): 153-158.
	ZHANG Zifeng. Current development situation, main challenges and its countermeasures of the pepper industry in China[J]. Northern Horticulture, 2023,(14): 153-158.
[2]	段洁利, 王昭锐, 叶磊, 等. 水果采摘机械臂运动规划研究进展与发展趋势[J]. Journal of Intelligent Agricultural Mechanization(in Chinese and English), 2021, (2): 7-17.
	DUAN Jieli, WANG Zhaorui, YE Lei, et al. Research progress and development trend of motion planning of fruit picking robot arm[J]. Journal of Intelligent Agricultural Mechanization, 2021, (2): 7-17.
[3]	金晶, 张小明, 付浩. 贵州省辣椒产业发展现状及建议[J]. 北方园艺, 2021,(21): 152-156.
	JIN Jing, ZHANG Xiaoming, FU Hao. Status and suggestions of chili industry development in Guizhou Province[J]. Northern Horticulture, 2021,(21): 152-156.
[4]	Kang D, Benipal S S, Gopal D L, et al. Hybrid pixel-level concrete crack segmentation and quantification across complex backgrounds using deep learning[J]. Automation in Construction, 2020, 118: 103291.
[5]	韩文霆, 崔家伟, 崔欣, 等. 基于特征优选与机器学习的农田土壤含盐量估算研究[J]. 农业机械学报, 2023, 54(3): 328-337.
	HAN Wenting, CUI Jiawei, CUI Xin, et al. Estimation of farmland soil salinity content based on feature optimization and machine learning algorithms[J]. Transactions of the Chinese Society for Agricultural Machinery, 2023, 54(3): 328-337.
[6]	陈小毛, 王立成, 张健, 等. 融合YOLOv5与ASFF算法的海产品目标检测算法研究[J]. 无线电工程, 2023, 53(4): 824-830.
	CHEN Xiaomao, WANG Licheng, ZHANG Jian, et al. Research on seafood object detection algorithm integrating YOLOv5 and ASFF algorithm[J]. Radio Engineering, 2023, 53(4): 824-830.
[7]	周岳淮, 李震, 左嘉明, 等. 基于改进YOLOv5s模型的山地果园单轨运输机搭载柑橘的检测[J]. 湖南农业大学学报(自然科学版), 2023, 49(4): 491-496.
	ZHOU Yuehuai, LI Zhen, ZUO Jiaming, et al. Detection of Citrus carried by mountainous orchard monorail transporter based on improved YOLOv5s[J]. Journal of Hunan Agricultural University (Natural Sciences), 2023, 49(4): 491-496.
[8]	Tian Y N, Yang G D, Wang Z, et al. Apple detection during different growth stages in orchards using the improved YOLO-V3 model[J]. Computers and Electronics in Agriculture, 2019, 157: 417-426.
[9]	段洁利, 王昭锐, 邹湘军, 等. 采用改进YOLOv5的蕉穗识别及其底部果轴定位[J]. 农业工程学报, 2022, 38(19): 122-130.
	DUAN Jieli, WANG Zhaorui, ZOU Xiangjun, et al. Recognition of bananas to locate bottom fruit axis using improved YOLOv5[J]. Transactions of the Chinese Society of Agricultural Engineering, 2022, 38(19): 122-130.
[10]	Wang F H, Sun Z X, Chen Y, et al. Xiaomila green pepper target detection method under complex environment based on improved YOLOv5s[J]. Agronomy, 2022, 12(6): 1477.
[11]	王圆圆, 林建, 王姗. 基于YOLOv4-tiny模型的水稻早期病害识别方法[J]. 江苏农业科学, 2023, 51(16): 147-154.
	WANG Yuanyuan, LIN Jian, WANG Shan. An early rice disease recognition method based on YOLOv4-tiny model[J]. Jiangsu Agricultural Sciences, 2023, 51(16): 147-154.
[12]	Tang Z, Zhou L, Qi F, et al. An improved lightweight and real-time YOLOv5 network for detection of surface defects on Indocalamus leaves[J]. Journal of Real-Time Image Processing, 2023, 20(1): 14.
[13]	高芳征, 汤文俊, 陈光明, 等. 基于改进YOLOv3的复杂环境下西红柿成熟果实快速识别[J]. 中国农机化学报, 2023, 44(8): 174-183.
	GAO Fangzheng, TANG Wenjun, CHEN Guangming, et al. Fast recognition of ripe tomato fruits in complex environment based on improved YOLOv3[J]. Journal of Chinese Agricultural Mechanization, 2023, 44(8): 174-183. DOI
[14]	朱立成, 韩振浩, 赵博, 等. 基于VINS-MONO和改进YOLO v4-Tiny的果园自主寻筐方法[J]. 农业机械学报, 2023, 54(8): 97-109.
	ZHU Licheng, HAN Zhenhao, ZHAO Bo, et al. Autonomous basket searching method for orchards transporter based on VINS-MONO and improved YOLO v4-tiny[J]. Transactions of the Chinese Society for Agricultural Machinery, 2023, 54(8): 97-109.
[15]	He K M, Zhang X Y, Ren S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916. DOI PMID
[16]	汤文虎, 吴龙, 黎尧, 等. 基于改进Faster RCNN的钢线圈头部小目标检测算法[J]. 现代制造工程, 2023, (8): 127-133, 147.
	TANG Wenhu, WU Long, LI Yao, et al. Improved faster RCNN based steel coil head detection algorithm for small objects[J]. Modern Manufacturing Engineering, 2023, (8): 127-133, 147.
[17]	吕宗旺, 邱帅欣, 孙福艳, 等. 基于改进YOLOv5s的轻量化储粮害虫检测方法[J]. 中国粮油学报, 2023, 38(8): 221-228.
	LYU Zongwang, QIU Shuaixin, SUN Fuyan, et al. Lightweight grain storage pest detection method based on improved YOLOv5s[J]. Journal of the Chinese Cereals and Oils Association, 2023, 38(8): 221-228.
[18]	李淑菲, 李凯雨, 乔岩, 等. 基于可见光光谱和改进YOLOv5的自然场景下黄瓜病害检测方法[J]. 光谱学与光谱分析, 2023, 43(8): 2596-2600.
	LI Shufei, LI Kaiyu, QIAO Yan, et al. Cucumber disease detection method based on visible light spectrum and improved YOLOv5 in natural scenes[J]. Spectroscopy and Spectral Analysis, 2023, 43(8): 2596-2600.
[19]	Hong W W, Ma Z H, Ye B L, et al. Detection of green Asparagus in complex environments based on the improved YOLOv5 algorithm[J]. Sensors, 2023, 23(3): 1562.
[20]	杨国亮, 王吉祥, 聂子玲. 基于改进型YOLOv5s的番茄实时识别方法[J]. 江苏农业科学, 2023, 51(15): 187-193.
	YANG Guoliang, WANG Jixiang, NIE Ziling. A real-time tomato recognition method based on improved YOLOv5s[J]. Jiangsu Agricultural Sciences, 2023, 51(15): 187-193.
[21]	李尚平, 郑创锐, 文春明, 等. 基于改进YOLO v5s的甘蔗切种茎节特征识别定位技术[J]. 农业机械学报, 2023, 54(10): 234-245,293.
	LI Shangping, ZHENG Chuangrui, WEN Chunming, et al. Stem node feature recognition and positioning technology for transverse cutting of sugarcane based on improved YOLO v5s[J]. Transactions of the Chinese Society for Agricultural Machinery, 2023, 54(10): 234-245,293.
[22]	姜国权, 杨正元, 霍占强, 等. 基于改进YOLOv5网络的疏果前苹果检测方法[J]. 江苏农业科学, 2023, 51(14): 205-215.
	JIANG Guoquan, YANG Zhengyuan, HUO Zhanqiang, et al. Apple detection method before thinning fruit based on improved YOLOv5 model[J]. Jiangsu Agricultural Sciences, 2023, 51(14): 205-215.
[23]	Zhe T, Lin Z, Fang Q, et al. An improved lightweight and real-time YOLOv5 network for detection of surface defects on indocalamus leaves[J]. Journal of Real-Time Image Processing, 2023, 20(1).
[24]	杨永闯, 王昊, 王新良. 基于改进SSD的食物浪费行为识别方法[J]. 计算机工程与设计, 2023, 44(8): 2523-2530.
	YANG Yongchuang, WANG Hao, WANG Xinliang. Food waste behavior recognition method based on improved SSD[J]. Computer Engineering and Design, 2023, 44(8): 2523-2530.
[25]	郭殿鹏, 柯海森, 李孝禄, 等. 基于改进YOLOv7的织物疵点检测算法[J]. 棉纺织技术, 2023, 51(12): 5-11.
	GUO Dianpeng, KE Haisen, LI Xiaolu, et al. Fabric defect detection algorithm based on improved YOLOv7[J]. Cotton Textile Technology, 2023, 51(12): 5-11.
[26]	崔海彬, 蒲东兵, 陆云凤, 等. 基于CA-YOLO的安全帽佩戴检测[J]. 东北师大学报(自然科学版), 2023, 55(3): 94-100.
	CUI Haibin, PU Dongbing, LU Yunfeng, et al. Helmet wearing detection based on CA-YOLO[J]. Journal of Northeast Normal University (Natural Science Edition), 2023, 55(3): 94-100.

编辑推荐 0

Metrics

阅读次数

全文

HTML			PDF

最新录用	在线预览	正式出版	最新录用	在线预览	正式出版
0	0	6	0	0	16

来源	本网站	其他网站

次数	18	4
比例	82%	18%

摘要

最新录用	在线预览	正式出版

0	0	94

来源	本网站	其他网站

次数	46	48
比例	49%	51%

基于改进YOLOv5的绿辣椒目标检测方法

Green chili pepper target detection method based on improved YOLOv5

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 26

相关文章 2

编辑推荐 0

Metrics

[1]	董兆森, 张佳喜, 周欣, 罗文杰, 蒋永新, 郭刚. 基于无线通讯技术的松土装置应力应变监测系统设计与试验[J]. 新疆农业科学, 2024, 61(4): 1029-1040.
[2]	方杰, 张杰, 马娟, 田翔, 于秀针, 冯斌. 基于神经网络PID算法的优化饲料配制系统[J]. 新疆农业科学, 2023, 60(4): 1003-1010.