摘要
黄瓜病害识别存在准确率不高、计算耗时较长等问题,本文提出一种用于黄瓜叶片病害识别的改进框架。首先在特征提取阶段,对VGG19和Inception V3模型使用迁移学习进行训练,采用并行最大协方差(PMC)进行信息融合,其次利用鲸鱼优化算法进行特征优化,最后使用监督学习算法对选定的最佳特征进行分类。利用本文构建的数据集,算法准确率提高至96.5%,耗时45.28 s,优于传统算法。
Cucumber disease identification faces challenges such as low accuracy and long computation time.This paper proposes a novel framework for cucumber leaf disease recognition.Firstly,in the feature extraction stage,VGG19 and Inception V3 models are trained using transfer learning,and Parallel Maximum Covariance(PMC)is employed for information fusion.Secondly,the Whale Optimization Algorithm(WOA)is utilized for feature optimization.Finally,supervised learning algorithms are used to classify the selected optimal features.Utilizing the dataset constructed in this paper,the algorithm achieves an accuracy of 96.5%with a computation time of 45.28 s,outperforming traditional algorithms.
作者
吴洪昊
孙娟
WU Honghao;SUN Juan(Yancheng Agricultural College,Yancheng 224051,Jiangsu,China)
出处
《智能计算机与应用》
2024年第10期176-181,共6页
Intelligent Computer and Applications
基金
2024年度江苏省青年科技人才托举工程(盐城市科协资助)
2024年盐城市重点研发计划。
关键词
迁移学习
鲸鱼优化算法
病害识别
transfer learning
Whale Optimization Algorithm
disease recognition
