基于改进FPN模型的西瓜幼苗智能识别方法

doi:10.13733/j.jcam.issn.20955553.2024.12.022

摘要/Abstract

摘要：

为提高对不同时期西瓜幼苗智能识别的准确度和运行效率，采用深度学习技术提出改进特征金字塔模型（FPN）的智能识别方法。首先结合特征金字塔网络模型和Res2Net模型设计网络模型，利用有效通道注意力机制（ECA)赋予空间特征不同权重，采用通道参数共享的方式，降低模型的计算复杂度；然后采用残差结构对模型进行优化改进，在不增加训练参数的情况下，解决网络深度不断提升时出现的网络退化问题；最后在全连接层使用深度可分离卷积替换传统卷积，从而大幅减少计算量，实现轻量化的设计。对不同生长期西瓜幼苗叶片进行试验。结果表明：与几种较为先进的识别算法相比，提出的识别方法具有更高的识别准确度和最短的运算耗时，识别率达到96.84%，等误率仅为0.54%，平均精度mAP达到91.68%，运算耗时低至112 ms，为推动智慧农业的发展和实现智能化的农业管理决策提供技术保障。

关键词: 农作物表型识别, 深度学习, 特征金字塔, 残差网络, 多尺度特征, 可分离卷积

Abstract:

In order to improve the accuracy and efficiency of intelligent identification of watermelon seedlings in different periods, an intelligent recognition method with improved Feature Pyramid Network (FPN) is proposed by using deep learning technology. Firstly, a network model was designed by combining the feature pyramid network model and the Res2Net model, and the Efficient Channel Attention (ECA) mechanism was used to assign different weights to spatial features, and the channel parameter sharing was adopted to reduce the computational complexity of the model. Then, the residual structure was used to optimize and improve the model, which solved the problem of network degradation that occurred when the network depth continued to increase without increasing training parameters. Finally, the deep separable convolutions were used in the fully connected layer to replace traditional convolutions, significantly reducing computational complexity and achieving a lightweight design. The experimental results on watermelon seedling leaves at different growth stages show that the proposed recognition method has higher recognition accuracy and the shortest computational time, compared with several more advanced recognition algorithms, the recognition rate is up to 96.84%, an equal error rate is only 0.54% and the mAP is up to 91.68%, and the computational time is as low as 112 ms, providing technical support for promoting the development of intelligent agriculture and achieving intelligent agricultural management decision-making.

Key words: crop phenotype recognition, deep learning, feature pyramid, residual network, multi scale features, separable convolution

中图分类号:

S651
TP391.4

李彦勤, , 王晓婷. 基于改进FPN模型的西瓜幼苗智能识别方法[J]. 中国农机化学报, 2024, 45(12): 148-153.

Li Yanqin, , Wang Xiaoting. Intelligent identification method for watermelon seedlings based on improved FPN model[J]. Journal of Chinese Agricultural Mechanization, 2024, 45(12): 148-153.

参考文献

［1］　岑海燕, 朱月明, 孙大伟. 深度学习在植物表型研究中的应用现状与展望［J］. 农业工程学报, 2020, 26(9): 1-16.
Cen Haiyan, Zhu Yueming, Sun Dawei, et al. Current status and future perspective of the application of deep learning in plant phenotype research ［J］. Transactions of the Chinese Society of Agricultural Engineering, 2020, 26(9): 1-16.
［2］　姚仁秀, 陈燕, 吕晓琴, 等. 海拔及环境因子影响杜鹃属植物的表型特征和化学性状［J］. 生物多样性, 2023, 31(2): 23-37.
Yao Renxiu, Chen Yan, Lü Xiaoqin, et al. Altituderelated environmental factors shape the phenotypic characteristics and chemical profile of rhododendron ［J］. Biodiversity Science, 2023, 31(2): 23-37.
［3］　朱润军, 杨巧, 李仕杰, 等. 植物表型可塑性对环境因子的响应研究进展［J］. 西南林业大学学报(自然科学), 2021, 41(1): 183-187.
Zhu Runjun, Yang Qiao, Li Shijie, et al. Advances at phenotypic plasticity in plant responses to environmental factors ［J］. Journal of Southwest Forestry University (Natural Science), 2021, 41(1): 183-187.
［4］　卢鹏, 金静静, 曹培健, 等. 植物及烟草表型组学大数据研究进展［J］. 烟草科技, 2021, 54(3): 90-100, 112.
Lu Peng, Jin Jingjing, Cao Peijian, et al. Progress on plant phenomics big data and their application in tobacco research ［J］. Tobacco Science & Technology, 2021, 54(3): 90-100, 112.
［5］　王春颖, 泮玮婷, 李祥, 等. 基于STLSTM的植物生长发育预测模型［J］. 农业机械学报, 2022, 53(6): 250-258.
Wang Chunying, Pan Weiting, Li Xiang, et al. Plant growth and development prediction model based on STLSTM ［J］. Transactions of the Chinese Society for Agricultural Machinery, 2022, 53(6): 250-258.
［6］　王映龙, 肖欢, 殷华, 等. 基于RGBD相机的黑皮鸡枞菌子实体形态视觉测量［J］. 农业工程学报, 2022, 28(20): 140-148.
Wang Yinglong, Xiao Huan, Yin Hua, et al. Measurement of morphology of oudemansiella raphanipies based on RGBD camera ［J］. Transactions of the Chinese Society of Agricultural Engineering, 2022, 28(20): 140-148.
［7］　陈旭波, 章雨希, 张亚芬, 等. 入侵植物藿香蓟表型可塑性对种间竞争的响应［J］. 植物科学学报, 2023, 41(1): 37-43.
Chen Xubo, Zhang Yuxi, Zhang Yafen, et al. Response of phenotypic plasticity of invasive Ageratum conyzoides L. to interspecific competition ［J］. Plant Science Journal, 2023, 41(1): 37-43.
［8］　王红梅, 路宁娜, 陈小爽, 等. 两种同域分布的马先蒿属植物花特征与报酬间的相关性及表型选择［J］. 兰州大学学报(自然科学版), 2022, 58(3): 292-297, 305.
［9］　娄路, 吕惠, 宋然. 基于多视角时间序列图像的植物叶片分割与特征提取［J］. 农业机械学报, 2022, 53(1): 253-260.
Lou Lu, Lü Hui, Song Ran. Segmentation of plant leaves and features extraction based on mutiview and timeseries image ［J］. Transactions of the Chinese Society for Agricultural Machinery, 2022, 53(1): 253-260.
［10］　袁山, 汤浩, 郭亚. 基于改进Mask R-CNN模型的植物叶片分割方法［J］. 农业工程学报, 2022, 28(1): 212-220.
Yuan Shan, Tang Hao, Guo Ya. Segmentation method for plant leaves using an improved Mask R-CNN model ［J］. Transactions of the Chinese Society of Agricultural Engineering, 2022, 28(1): 212-220.
［11］　王晓婷, 赵展, 王阳, 等. 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.
Wang Xiaoting, Zhao Zhan, Wang Yang, et al. Intelligent detection algorithm of plant phenotype based on improved Mask R-CNN［J］. Journal of Chinese Agricultural Mechanization, 2022, 43(8): 151-157.
［12］　邹龙, 王以松, 魏昆鹏, 等. LU-ReNet植物叶片分割与计数模型［J］. 计算机工程与设计, 2022, 43(2): 553-559.
Zou Long, Wang Yisong, Wei Kunpeng, et al. LU-ReNet plant leaf segmentation and counting model ［J］. Computer Engineering and Design, 2022, 43(2): 553-559.
［13］　Zhao D, Ma H, Yang Z, et al. Finger vein recognition based on lightweight CNN combining center loss and dynamic regularization ［J］. Infrared Physics & Technology, 2020, 105(8): 103221.
［14］　杨万里, 段凌凤, 杨万能. 基于深度学习的水稻表型特征提取和穗质量预测研究［J］. 华中农业大学学报, 2021, 40(1): 227-235.
Yang Wanli, Duan Lingfeng, Yang Wanneng, et al. Deep learningbased extraction of rice phenotypic characteristics and prediction of rice panicle weight ［J］. Journal of Huazhong Agricultural University, 2021, 40(1): 227-235.
［15］　姜妍, 王琳, 杨月, 等. 无人机高光谱成像技术在作物生长信息监测中的应用［J］. 东北农业大学学报, 2022, 53(3): 88-96.
Jiang Yan, Wang Lin, Yang Yue, et al. Application of UAV hyperspectral imaging technology in crop growth information monitoring ［J］. Journal of Northeast Agricultural University, 2022, 53(3): 88-96.
［16］　张卫正, 李旭光, 万瀚文, 等. 基于骨架提取和二叉树分析的玉米植株图像茎叶分割方法［J］.河南农业科学, 2020, 49(9): 166-172.
Zhang Weizheng, Li Xuguang, Wan Hanwen, et al. Stem and leaf segmentation of maize plant image based on skeleton extraction and binary tree analysis ［J］. Journal of Henan Agricultural Sciences, 2020, 49(9): 166-172.
［17］　王鹤树, 曹丽英. 基于全卷积神经网络的植物叶片自动分割及表型解析［J］. 中国农机化学报, 2021, 42(8): 161-168.
Wang Heshu, Cao Liying. Automatic segmentation and phenotypic analysis of plant leaves based on fully convolutional networks ［J］. Journal of Chinese Agricultural Mechanization, 2021, 42(8): 161-168.
［18］　胡玲艳, 许巍, 秦山, 等. 基于分时重叠算法的欧洲甜樱桃表型关键特征区域图像分割方法［J］. 江苏农业科学, 2023, 51(1): 195-202.
Hu Lingyan, Xu Wei, Qin Shan, et al. Image segmentation of key feature regions of European sweet cherry phenotype based on timesharing overlap algorithm ［J］. Jiangsu Agricultural Sciences, 2023, 51(1): 195-202.
［19］　李少辰, 张爱武, 张希珍, 等. 叶片尺度的玉米幼苗三维表型信息提取方法［J］. 激光与光电子学进展, 2023, 60(2): 61-69.
Li Shaochen, Zhang Aiwu, Zhang Xizhen, et al. 3D phenotypic information extraction method of maize seedlings at leaf scale ［J］. Laser & Optoelectronics Progress, 2023, 60(2): 61-69.

[1]	韩鹏飞, 宋其江, 贾梦实. 基于改进轻量化 EfficientNet-V2模型的小麦种子分类[J]. 中国农机化学报, 2024, 45(9): 111-117.
[2]	李志臣, 凌秀军, 李鸿秋. 基于声音深度学习的发动机失火故障诊断[J]. 中国农机化学报, 2024, 45(9): 134-140.
[3]	沈漫林, 刘姣娣, 许洪振, 何捷, 段玉龙. 基于改进 YOLOv4的蔗种坏芽识别方法研究[J]. 中国农机化学报, 2024, 45(9): 190-195.
[4]	桑一男 , 徐增莱 , 汪琼 , 葛海涛 , 王殿广 . 一种轮式采花机器人设计——以黄蜀葵为例[J]. 中国农机化学报, 2024, 45(9): 202-208.
[5]	李邦国, 王辉, 宋杨, 任志伟, 刘跃华, 徐乐程. 基于无人驾驶小麦收割机立体视觉感知系统[J]. 中国农机化学报, 2024, 45(9): 244-249.
[6]	宋立航, 张屹, 时寅豪. 基于改进 YOLOv5s的自然环境下油桃成熟度检测方法[J]. 中国农机化学报, 2024, 45(9): 250-257.
[7]	李洁, 高尚兵, 余骥远, 陈新, 李士丛, 袁星星. 基于PMMS-Net和叶绿素荧光成像的绿豆叶斑病抗性鉴定方法[J]. 中国农机化学报, 2024, 45(8): 210-216.
[8]	许贞辉, 李晓娟. 果园树干检测与导航线拟合算法研究[J]. 中国农机化学报, 2024, 45(8): 217-222.
[9]	胡国玉, 刘广, 周星光, 董娅兰, 周建平, . 基于Swin-TDL算法的果园环境下葡萄病害检测方法[J]. 中国农机化学报, 2024, 45(8): 234-239.
[10]	王克涛, 陈世锋, 陈贵, 韦锦, 蒙丽雯, 陈泉成. 基于YOLOv5s的园林球形绿篱检测方法[J]. 中国农机化学报, 2024, 45(8): 262-269.
[11]	黄晓宇, 张聪, 陈晓玲. 融合多尺度特征的遮挡番茄病害图像识别研究[J]. 中国农机化学报, 2024, 45(7): 194-200.
[12]	李运良, 赵明岩, 王鑫, 闫泓硕, 李育蝉. 基于改进ResNet的带孔蛋胚裂纹检测方法#br#[J]. 中国农机化学报, 2024, 45(7): 201-208.
[13]	吴利刚, 吕媛媛, 周倩, 陈乐, 张梁, 史建华. 基于改进YOLOv5的轻量级黄花成熟检测方法[J]. 中国农机化学报, 2024, 45(7): 235-242.
[14]	张书贵, 陈书理, 赵展. 改进YOLOv8的农作物叶片病虫害识别算法[J]. 中国农机化学报, 2024, 45(7): 255-260.
[15]	张继成, 黄向党. 基于改进YOLOv3的玉米病害识别方法[J]. 中国农机化学报, 2024, 45(7): 269-275.