文章摘要
陈思玮,贾克斌,王聪聪,刘钧.深度学习在多天气分类算法中的研究与应用[J].高技术通讯(中文),2020,30(10):1010~1017
深度学习在多天气分类算法中的研究与应用
  
DOI:doi:10.3772/j.issn.1002-0470.2020.10.003
中文关键词: 天气分类; 深度学习; 卷积神经网络(CNN); DenseNet; 迁移学习
英文关键词: multiple weather classification, deep learning, convolutional neural network (CNN), DenseNet, transfer learning
基金项目:
作者单位
陈思玮  
贾克斌  
王聪聪  
刘钧  
摘要点击次数: 325
全文下载次数: 247
中文摘要:
      针对目前多天气识别分类问题,提出了一种基于深度学习和计算机视觉的天气现象自动分类算法。采集并建立了一个包括雾霾、沙尘、雨、雪、霜、露6类天气的适用于任意场景的多天气现象数据集,改善了目前已见报数据集规模小、种类单一、只面向特定场景的情况;同时采用密集连接和池化均衡的结构搭建深度卷积神经网络(CNN)模型,训练并挖掘天气数据的特征与内在规律,用深度学习方法实现天气现象的自动分类。实验结果表明:相比传统计算机视觉算法,该算法解决了严重依靠特征提取、适用场景单一问题;且比大多数深度网络模型参数更少、识别准确性更高,算法泛化性能大幅提升。
英文摘要:
      Aiming at the problem of multi weather recognition and classification, an automatic weather classification algorithm based on deep learning and computer vision is proposed. A multi weather phenomenon dataset is collected and established, which includes six kinds of weather: haze, dust, rain, snow, frost and dew. It is suitable for any scene, and improves the problem of small scale, single type and specific scene only. The deep convolutional neural network (CNN) model is built with the structure of intensive connection and pool equilibrium. The characteristics and internal laws of weather data are trained and mined, and the automatic classification of weather phenomena is realized with the method of deep learning. The experimental results show that compared with traditional computer vision algorithm, this paper solves the problem of relying heavily on feature extraction and single applicable scene; at the same time, compared with most depth network models, it has fewer parameters, higher recognition accuracy, and greatly improves the generalization performance.
查看全文   查看/发表评论  下载PDF阅读器
关闭

分享按钮