主办单位:中国气象局沈阳大气环境研究所
国际刊号:ISSN 1673-503X
国内刊号:CN 21-1531/P

气象与环境学报 ›› 2021, Vol. 37 ›› Issue (1): 106-112.doi: 10.3969/j.issn.1673-503X.2021.01.014

• 快报 • 上一篇    

基于机器学习的重庆长江航道雾图像特征识别研究

王远谋1(),李家启1,*(),陈施吉1,唐家萍1,夏佰成1,韩世刚1,2   

  1. 1. 重庆市气象服务中心, 重庆 401147
    2. 重庆市开州区气象局, 重庆 405400
  • 收稿日期:2019-10-28 出版日期:2021-02-28 发布日期:2021-01-21
  • 通讯作者: 李家启 E-mail:501242840@qq.com;475499221@qq.com
  • 作者简介:王远谋, 女, 1993年生, 工程师, 主要从事气象服务与应用气象技术研究, E-mail: 501242840@qq.com
  • 基金资助:
    重庆市气象局智慧气象技术创新团队项目(ZHCXTD—201917)

Identification of the fog image features on the Yangtze River waterways in Chongqing based on machine learning

Yuan-mou WANG1(),Jia-qi LI1,*(),Shi-ji CHEN1,Jia-ping TANG1,Bai-cheng XIA1,Shi-gang HAN1,2   

  1. 1. Chongqing Meteorological Service Center, Chongqing 401147, China
    2. Meteorological Service at Kaizhou District of Chongqing, Chongqing 405400, China
  • Received:2019-10-28 Online:2021-02-28 Published:2021-01-21
  • Contact: Jia-qi LI E-mail:501242840@qq.com;475499221@qq.com

摘要:

基于重庆市境内长江航道雷达站拍摄的雾天气过程影像资料,利用K最近邻、支持向量机、BP神经网络、随机森林等机器学习算法,对无雾和5类有雾天气个例进行图像识别训练,构建雾图像识别模型,并检验了识别准确率。结果表明:机器学习能够有效识别雾图像,随机森林算法的识别效果优于其余3种算法。对于能见度超过1500 m的无雾天气,模型的识别准确率为100%,对于能见度在1000—1500 m范围内的轻雾、能见度低于50 m的强浓雾,模型的识别准确率在90%以上,对于能见度在50—1000 m范围内的雾、大雾和浓雾,识别准确率超过70%。

关键词: 雾, 机器学习, 图像识别, 图形用户界面

Abstract:

Based on the image data of fog events taken by radar stations on the Yangtze River waterways in Chongqing, images without fog events and with five types of fog events were trained using algorithms including K-nearest neighbor, support vector machine, back propagation neural network, and random forest.According to the training results, a fog image identification model was built, and the identification accuracy was tested.The results show that machine learning can effectively identify fog images, and random forest performances better than the other three algorithms.The model has a recognition accuracy of 100% for non-fog recognition, over 90% for light fog and strong dense fog events, and over 70% for fog, heavy fog, and dense fog events.

Key words: Fog, Machine learning, Pattern recognition, Graphical user interface

中图分类号: