基于深度学习图像识别的气象站净空环境量化方法研究

Research on Quantitative Methods for Meteorological Station Clear-zone Environments Based on Deep Learning Image Recognition

  • 摘要: 净空环境情况分析是气象站址建站的关键工作,对站点代表性意义重大。传统净空环境数据获取主要依靠全站仪或经纬仪测量站址周边地形、建筑物及植被的遮挡情况,而人工测量过程易受主观因素影响,误差较大且效率较低。为满足气象观测站网快速、高效、准确开展此项工作的需求,本文提出一种基于机器学习的气象站址净空情况分析方法,借助DeeplabV3+与Resnet18组合模型处理气象站址全景照片,计算站址周围净空环境的遮挡角度参数并自动识别遮挡物,进而对气象站址净空情况进行分析。实验结果表明:该方法能有效生成不同训练模型,加载模型后可快速从站址全景照片中提取净空环境数据;与传统方法相比,其在效率和便捷性方面优势显著;此外,该方法通过图像加工,可兼容地面气象站点的全天空成像图片,在气象站选址建站及探测环境保护方面具有重要的应用价值。

     

    Abstract: Clear-zone environment analysis is a critical task for meteorological station sites, which is of great significance for the representativeness of the stations. Traditionally, the acquisition of clear-zone environment data mainly relies on total stations or theodolites to measure occlusion of terrain, buildings, and vegetation around the station site. However, the manual measurement process is easily affected by subjective factors, with large errors and low efficiency. To meet the demand for rapid, efficient, and accurate clear-zone assessment in meteorological observation networks, this study proposes a method for analyzing the clear-zone situation of meteorological station sites based on machine learning, which can quickly provide calculation results. This method uses the combined model of DeeplabV3+ and Resnet18 to process the panoramic photos of meteorological station sites, calculates the occlusion angle parameters of the clear-zone environment around the station sites, automatically identifies the occluding objects, and then analyzes the clear-zone situation of the meteorological station sites. The experimental results show that this method can effectively generate different training models. After loading the model, it can quickly extract the clear-zone environment data from station panoramic photos. Compared with traditional methods, it has significant advantages in efficiency and convenience. In addition, through image processing, this method can be compatible with all-sky imaging pictures of ground meteorological stations, and has important application value for site selection and construction of meteorological stations as well as the protection of meteorological observation environments.

     

/

返回文章
返回