摘要
为提升花椒图像分类识别准确率,借助自适应多级小波分解的时间序列分类(Adaptive Multi-level Wavelet Decomposition based neural network,AMWDNet)模型,着重关注此分类模型中的频域信息,克服从时域出发对目标序列进行建模造成频域信息缺失的瓶颈。本研究借鉴小波分解技术,结合长、短期时间模式提取方法,构建出更加精确、更加实时的时间序列花椒识别模型。经过对比试验研究发现,在UCR数据库中的4个数据集测试上,AMWDNet模型表现出优异的分类性能和强大的泛化能力,超过其他3个基准模型,由此提升花椒图像分类识别准确率。
To improve the accuracy of zanthoxylum bungeanum maxim image classification and recognition,the Adaptive Multilevel Wavelet Decomposition based neural network(AMWDNet)model is utilized to focus on the frequency domain information in this classification model,overcoming the bottleneck of missing frequency domain information caused by modeling the target sequence from the time domain.This study draws on wavelet decomposition technology and combines long and short-term time pattern extraction methods to construct the more accurate and real-time time series zanthoxylum bungeanum maxim recognition model.After comparative experimental research,it is found that the AMWDNet model demonstrates excellent classification performance and strong generalization ability on four datasets tested in the UCR database,surpassing the other three benchmark models.This research improves the accuracy of zanthoxylum bungeanum maxim image classification and recognition.
作者
李论
徐杨
王义
王天一
蒋宁
LI Lun;XU Yang;WANG Yi;WANG Tianyi;JIANG Ning(College of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China;China Telecom Guizhou Branch,Guiyang 550001,China;Guizhou Xuande Zanthoxylum Bungeanum Maxim Industry Development Co.,Ltd.,Guiyang 550018,China)
出处
《智能计算机与应用》
2024年第5期235-240,共6页
Intelligent Computer and Applications
基金
贵州省科技计划项目(黔科合支撑[2021]一般176)。
关键词
花椒图像分类
时间序列分类
小波分解
时频信息
准确率
zanthoxylum bungeanum maxim image classification
time series classification
wavelet decomposition
time-frequency information
accuracy rate
