基于图像的终端区相似气象运行场景识别分析

Analysis of similar weather operating scenarios in terminal areas based on image patterns

天气影响是空中交通运行不确定性的主要原因之一,基于历史相似气象场景识别是一种有效辅助管制运行决策的依据.相较于传统的基于距离度量空间分布的相似性识别方法,从视觉角度提出了一种基于CNN(Convolutional Neural Network)图像模式的终端区相似气象运行场景识别方法.将处理后的WAF(Weather Avoidance Field)天气避让区图像作为神经网络训练集,构建CNN+K means++模型实现对气象场景分类.进一步依据评估场景分类的实际效果和管制决策状态,构建了基于气象、交通、策略多领域的终端区特征集.从典型场景-典型日视角切入,分析不同场景对终端区运行能力与策略发布的影响,并以广州终端区为例进行验证,结果表明,基于CNN图像模式的对流天气识别具有较显著的效果,可为交通流量管理措施提供指导辅导决策.

Weather impact is one of the main reasons for the uncertainty of air traffic operation,and the recognition of similar meteorological scenes based on history is an effective basis to assist the decision-making of control operation.Compared with the traditional similarity recognition method based on distance measurement and spatial distribution,this paper proposed a method for recognizing similar meteorological operation scenes in terminal area based on CNN(Convolutional Neural Network)image pattern from the perspective of vision.The processed WAF(Weather Avoidance Field)weather avoidance area image was used as the neural network training set,and the CNN+K means++model was constructed to classify the weather scene.Further,according to the actual effect of the evaluation scenario classification and the control decision status,the terminal area feature set based on the weather,traffic and strategy fields was constructed.From the perspective of typical scenario-typical day,this paper analyzed the impact of different scenarios on the operation capacity and strategy release of the terminal area,and took the Guangzhou terminal area as an example to verify the results.The results showed that the recognition of convective weather based on CNN image mode has a significant effect,which can provide guidance and guidance for traffic flow management measures.