[1]宋文韬,姜茹月,舒欣.基于零样本学习的枸杞虫害识别[J].江苏农业学报,2024,(02):320-330.[doi:doi:10.3969/j.issn.1000-4440.2024.02.014]
　SONG Wen-tao,JIANG Ru-yue,SHU Xin.Identification of Lycium barbarum pests based on zero-shot learning[J].,2024,(02):320-330.[doi:doi:10.3969/j.issn.1000-4440.2024.02.014]
点击复制

基于零样本学习的枸杞虫害识别()
分享到：

江苏农业学报[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2024年02期
页码:
320-330
栏目:
农业信息工程
出版日期:
2024-02-25

文章信息/Info

Title:
Identification of Lycium barbarum pests based on zero-shot learning
作者:
宋文韬; 姜茹月; 舒欣
(南京农业大学人工智能学院,江苏南京210095)
Author(s):
SONG Wen-tao; JIANG Ru-yue; SHU Xin
(College of Artificial Intelligence, Nanjing Agricultural University, Nanjing 210095, China)
关键词:
零样本学习; 矩阵分解; 枸杞病虫害识别; 哈希码
Keywords:
zero-shot learning; matrix factorization; Lycium barbarum pests detection; hashing code
分类号:
TP391
DOI:
doi:10.3969/j.issn.1000-4440.2024.02.014
摘要:
针对农业领域缺少有效的零样本虫害识别与检索方法,本研究提出一种基于零样本学习的枸杞虫害检索与识别方法。首先,通过对原始数据进行深层矩阵分解获得深层次结构特征,获取不同模态数据的特征表示,生成各模态的哈希码。然后结合类别属性信息对生成的哈希码引入线性约束,实现已知类别到新类别之间的知识迁移。最后,对所提出的模型通过直接学习离散哈希码避免了连续松弛方法带来的量化误差,提高了检索精度。在2020年宁夏枸杞虫害图文跨模态检索数据集及Wiki、Pascal VOC这3个公开数据集上的试验结果表明，与现有的基于协同矩阵分解的哈希方法（CMFH）、基于潜在语义的稀疏哈希方法（LSSH）、基于迁移监督知识的哈希方法（TSK）、基于属性的哈希方法（AH）、基于跨模态属性的哈希方法（CMAH）、基于正交投影的哈希方法（CHOP）、离散非对称零样本哈希方法（DAZSH）相比，本研究所提出的方法具有优越性。
Abstract:
In order to solve the problem of lack of effective zero-sample recognition and retrieval methods in agricultural field, a zero-sample learning-based retrieval and recognition method for Lycium barbarum pests was proposed in this study. Firstly, the deep structure features were obtained by deep matrix decomposition of the original data, and the characteristic representations of different modal data were obtained, and the hashing codes of each modality were generated. Then the linear constraint was introduced to the generated hashing code with the class attribute information to realize the knowledge transfer from the known class to the new class. Finally, the proposed model could avoid the quantization error caused by the continuous relaxation method and improve the retrieval precision by learning discrete hashing codes directly. The experimental results on the three public datasets, 2020 Ningxia Lycium barbarum pest image-text cross-modal retrieval dataset, Wiki, Pascal VOC, showed that the method proposed in this study was superior to the existing collective matrix factorization hashing (CMFH), latent semantic sparse hashing (LSSH), transferring supervised knowledge hashing (TSK), attribute hashing (AH), cross-modal attribute hashing (CMAH), cross-modal hashing with orthogonal projection (CHOP), and discrete asymmetric zero-shoot hashing (DAZSH).

参考文献/References:

[1]许盼盼. 枸杞抗盐种质资源筛选与抗盐基因的克隆鉴定[D]. 咸阳:西北农林科技大学,2018.
[2]徐峰. 宁夏枸杞产业竞争力研究[D]. 银川:宁夏大学,2017.
[3]范振军. 农作物病虫害图像检索方法研究与实现[D]. 绵阳:西南科技大学,2018.
[4]汪京京,张武,刘连忠,等. 农作物病虫害图像识别技术的研究综述[J]. 计算机工程与科学,2014,36(7):1363-1370.
[5]杭立,车进,宋培源,等. 基于机器学习和图像处理技术的病虫害预测[J]. 西南大学学报(自然科学版),2020,42(1):134-141.
[6]赵芸. 基于高光谱和图像处理技术的油菜病虫害早期监测方法和机理研究[D]. 杭州:浙江大学,2013.
[7]赵建敏,薛晓波,李琦. 基于机器视觉的马铃薯病害识别系统[J]. 江苏农业科学,2017,45(2):198-202.
[8]王佳. 计算机视觉在香芋病害检测中的应用研究[J]. 农机化研究,2020,42(8):241-244.
[9]NETTLETON D F, KATSANTONIS D, KALAITZIDIS A, et al. Predicting rice blast disease:machine learning versus process-based models[J]. BMC Bioinformatics,2019,20:1-16.
[10]王国伟,刘嘉欣. 基于卷积神经网络的玉米病害识别方法研究[J]. 中国农机化学报,2021,42(2):139-145.
[11]赵立新,侯发东,吕正超,等. 基于迁移学习的棉花叶部病虫害图像识别[J]. 农业工程学报,2020,36(7):184-191.
[12]鲍文霞,吴刚,胡根生,等. 基于改进卷积神经网络的苹果叶部病害识别[J]. 安徽大学学报(自然科学版),2021,45(1):53-59.
[13]冯晓,李丹丹,王文君,等. 基于轻量级卷积神经网络和迁移学习的小麦叶部病害图像识别[J]. 河南农业科学,2021,50(4):174-180.
[14]谢州益,冯亚枝,胡彦蓉,等. 基于ResNet18特征编码器的水稻病虫害图像描述生成[J]. 农业工程学报,2022,38(12):197-206.
[15]彭红星,徐慧明,刘华鼐. 融合双分支特征和注意力机制的葡萄病虫害识别模型[J]. 农业工程学报,2022,38(10):156-165.
[16]冀中,汪浩然,于云龙,等. 零样本图像分类综述:十年进展[J]. 中国科学(信息科学),2019,49(10):1299-1320.
[17]ZHONG F, CHEN Z, MIN G. An exploration of cross-modal retrieval for unseen concepts[C]//LI G L, YANG J, GAMA J, et al. Database systems for advanced applications:24th international conference, Proceedings, Part Ⅱ. Cham, Switzerland:Springer International Publishing,2019:20-35.
[18]JI Z, SUN Y, YU Y, et al. Attribute-guided network for cross-modal zero-shot hashing[J]. IEEE Transactions on Neural Networks and Learning Systems,2019,31(1):321-330.
[19]SHU Z, YONG K, YU J, et al. Discrete asymmetric zero-shot hashing with application to cross-modal retrieval[J]. Neurocomputing,2022,511:366-379.
[20]陈磊,刘立波,王晓丽. 2020 年宁夏枸杞虫害图文跨模态检索数据集[J]. 中国科学数据,2022,7(3):149-156.
[21]DING G, GUO Y, ZHOU J. Collective matrix factorization hashing for multimodal data[C]// IEEE. Proceedings of the IEEE conference on computer vision and pattern recognition. Los Alamitos, USA:IEEE,2014:2075-2082.
[22]ZHOU J, DING G, GUO Y. Latent semantic sparse hashing for cross-modal similarity search[C]//ACM. Proceedings of the 37th international ACM SIGIR Conference on research & development in information retrieval. New York:ACM,2014:415-424.
[23]XU Y, YANG Y, SHEN F, et al. Attribute hashing for zero-shot image retrieval[C]//IEEE. 2017 IEEE international conference on multimedia and expo (ICME). Hong Kong:IEEE,2017:133-138.
[24]YANG Y, LUO Y, CHEN W, et al. Zero-shot hashing via transferring supervised knowledge[C]//ACM. Proceedings of the 24th ACM international conference on multimedia. New York:ACM,2016:1286-1295.
[25]YUAN X, WANG G, CHEN Z, et al. CHOP:an orthogonal hashing method for zero-shot cross-modal retrieval[J]. Pattern Recognition Letters,2021,145:247-253.
[26]刘立波,赵斐斐. 融合注意力机制的枸杞虫害图文跨模态检索方法[J]. 农业机械学报,2022,53(2):299-308.

备注/Memo

备注/Memo:
收稿日期：2023-11-01基金项目：国家自然科学基金项目(61602248)；江苏省信息技术处理重点实验室开放课题项目(KJS2164)作者简介：宋文韬(1997-)，男，江苏涟水人，硕士，主要研究方向为计算机视觉、模式识别。(E-mail)374267655@qq.com。姜茹月为共同第一作者。通讯作者：舒欣，(E-mail)xinshu@njau.edu.cn
更新日期/Last Update: 2024-03-17