您好，欢迎访问浙江省农业科学院 机构知识库！
浙江省农业科学院机构知识库

全部

  • 全部
  • 标题
  • 学者
  • 机构
  • 关键词

Outlier classification for microbiological open set recognition

文献类型： 外文期刊

作者： Pan, Yining 1 ; Ye, Wei 1 ; Xie, Dejin 1 ; Wang, Jiaoyu 2 ; Wang, Hongkai 3 ; Qiu, Haiping 1 ;

作者机构： 1.Zhejiang Univ, Coll Control Sci & Engn, Hangzhou 310027, Zhejiang, Peoples R China

2.Zhejiang Acad Agr Sci, Hangzhou 310021, Zhejiang, Peoples R China

3.Zhejiang Univ, Inst Biotechnol, Hangzhou 310058, Zhejiang, Peoples R China

4.Zhejiang Univ, Huzhou Inst, Huzhou 313000, Zhejiang, Peoples R China

关键词： Outlier classification; Open set recognition; Feature representation; Microorganism morphology recognition

期刊名称：COMPUTERS AND ELECTRONICS IN AGRICULTURE （ 影响因子：7.7； 五年影响因子：8.4 ）

ISSN： 0168-1699

年卷期： 2024 年 224 卷

页码：

收录情况： SCI

摘要： Precise detection and effective intervention of pathogenic microorganisms are pivotal to ensure crop yield. Traditional open set recognition methods distinguish unknown classes, i.e., those not involved in training, as outliers or separate classes, which is insufficient for real -world applications like recognizing airborne microorganisms and particles. In this paper, a new setting for the classification of outliers for open set recognition is proposed. The proposed method classifies unknown classes while achieving the classification of known classes, treating each category in the open set equally. The two cores of the method are hierarchical feature representation and feature gallery construction. To enhance the feature representation, this paper proposes a novel architectural module of spatial attention on region hierarchy, namely SPARC. This module extracts features from hierarchical levels and models spatial interaction through the attention mechanism. With the induction of local -to -global description, minute differences with ambiguous appearances between classes can be detected, and the negative effects of image quality degradation can be ameliorated. Furthermore, a feature gallery containing the feature representation from both known and unknown classes is developed for feature retrieval. To demonstrate the effectiveness of the proposed model, the first dataset composed of agricultural microorganisms, named SMAD, is constructed, and experiments are conducted on existing datasets and the newly proposed SMAD dataset. The experimental results outperform the prior art on these datasets under both the outlier rejection and outlier classification settings.

  • 相关文献
作者其他论文 更多>>
置顶
购物车
反馈

© 京ICP备09089781号-29 主办单位：浙江省农业科学院

学术资源： 中国农业科学院 农业专业知识服务系统 农科联盟资源共建共享平台 中国农业科技文献与信息服务平台 中国农业期刊集成服务平台