基于实例白化与特征恢复的草莓病害识别领域泛化方法  

Domain Generalization Method of Strawberry Disease Recognition Based on Instance Whitening and Restitution

作　　者：胡晓波 许桃胜[2] 王成军[1] 朱洪波 甘雷 HU Xiaobo;XU Taosheng;WANG Chengjun;ZHU Hongbo;GAN Lei(School of Artificial Intelligence,Anhui University of Science and Technology,Huainan 232001,China;Hefei Institutes of Physical Science,Chinese Academy of Science,Hefei 230031,China)

机构地区：[1]安徽理工大学人工智能学院,安徽淮南232001 [2]中国科学院合肥物质科学研究院,安徽合肥230031

出　　处：《智慧农业(中英文)》2025年第1期124-135,共12页Smart Agriculture

基　　金：安徽理工大学高层次引进人才科研启动基金(2024yjrc05);安徽理工大学校级重点项目(XCZX2021-01);国家自然科学基金(62003001)。

摘　　要：[目的/意义]基于深度神经网络的草莓病害识别模型通常假设训练集(源域)和测试集(目标域)满足独立同分布。然而,在实际应用中由于光照、背景环境和草莓品种等多种因素的影响,测试集与训练集存在领域差异,造成模型在应用过程中的识别精度出现明显下降。针对这一问题,本研究提出一种基于实例白化与特征恢复的领域泛化方法,用于提升草莓病害识别模型的泛化性能。[方法]该方法首先利用实例白化技术消除源域和目标域间的风格差异,再从滤除的风格特征中提取任务相关特征,最后将任务相关特征恢复到白化后的特征中,以减轻实例白化对特征类别区分度的影响。为增强从风格特征中分离任务相关特征的效果,设计了两个特征提取器分别提取任务相关和任务无关特征,并采用双段恢复损失约束两特征提取器所提取特征与任务的相关性,引入互信息损失确保特征的相互独立,进一步增强特征分类效果。[结果和讨论]该方法可以在不降低源域识别精度的前提下,有效提升各病害识别模型在目标域上的泛化性能,如AlexNet加入该算法后,其不同风格目标域上的识别精度可分别提升3.97个百分点和2.79个百分点。相较于IBN-Net(Instance Batch Normalization Net)、可切换白化(Switchable Whitening,SW)、样式归一化和恢复模块(Style Normalization and Restitution,SNR)等其他领域泛化方法,该算法在测试数据集上的泛化性能可分别提高2.63%、2.35%和1.14%。[结论]本方法可有效提升基于深度学习的草莓病害识别模型在目标域中的泛化性能,可为草莓病害精准识别提供可靠的技术支撑。[Objective]Strawberry disease recognition models based on deep neural networks generally assume that the training(source domain)and the test(target domain)datasets are identically and independently distributed.However,in practical applications,due to the influence of illumination,background and strawberry variety,the target domain often exhibits significant domain shift from the source domain.The domain shift result in accuracy decline of the models in target domain.To address this problem,a domain generalization method based on instant whitening and restitution(IWR)was proposed to improve the generalization performance of strawberry disease identification models in this research.[Methods]Samples from different source often exhibit great domain shift due to variations in strawberry varieties,regional climate,and photography methods.Therefore,a dataset was constructed for domain generalization research on strawberry disease using two distinct approaches.The first dataset was acquired using a Nikon D810 camera at multiple strawberry farms in Changfeng county,An‐hui province,with a fixed sampling schedule and fixed camera distance.In contrast,the second dataset was an open-source collection,primarily comprising images captured using smartphones in multiple strawberry greenhouses in Korea,with varied and random shoot‐ing distances and angles.The IWR module mitigated style variations(e.g.,illumination,color)through instance whitening,where fea‐tures were normalized to reduce domain discrepancies between the datasets.However,such operation was task-ignorant and inevitable removed some task-relevant information,which may be harmful to classification performance of the models.To remedy this,the re‐moved task-relevant features were attempted to recover.Specifically,two modules were designed to extract task-relevant and task-ir‐relevant feature from the filtered style features,respectively.A dual restitution loss was utilized to constraint the modules'feature cor‐relation between the task and a mutual loss was

关 键 词：深度神经网络 草莓病害识别 实例白化 特征恢复 领域泛化 

分 类 号：S431.9[农业科学—农业昆虫与害虫防治] TP391.4[农业科学—植物保护]