基于形态学和K近邻算法的飞机尾流识别

Aircraft Wake Recognition Based on Morphology and KNN

航空器尾流的准确识别对于实现动态尾流间隔缩减技术具有重要意义。与传统利用尾流特征对称关系的方法相比,研究关注飞机尾流的形态特征,提出了一种基于数学形态学和K近邻算法(KNN)的飞机尾涡识别框架。在模型中,主要工作如下:1)根据数学形态学中开闭算子提取飞机尾涡样本的形态特征;2)引入卷积神经网络的池化层来降低特征维度;3)通过KNN机器学习模型识别所提取的尾涡特征。实验在双流机场(ZUUU)的02跑道附近布置Wind3D 6000多普勒激光雷达,通过采集进近风场数据来对模型进行验证。结果表明,所提模型分类准确率、精确率、召回率和F1-score分别为97.1%、92.6%、91.1%和91.9%。

Accurate recognition of aircraft wakes is of great significance for realizing dynamic wake interval reduction technology.Compared with the traditional method using the symmetry relationship of wake characteristics, this paper focuses on the morphological characteristics of the aircraft wake, and proposes an aircraft wake vortex recognition framework based on mathematical morphology and K-nearest neighbor algorithm(KNN).In the framework, the main work is as follows: 1)Extract the morphological characteristics of the airplane wake vortex sample according to the opening and closing operators in mathematical morphology;2)Introduce the pooling layer of the convolutional neural network to reduce the dimensionality of the morphological characteristics;3)Identify the extracted wake vortex features through the KNN machine learning model.In the experiment, Wind3 D 6000 Doppler Lidar was deployed near runway 02 of Shuangliu Airport(ZUUU),and the model was verified by collecting approach wind field data.The results show that the classification accuracy, precision, recall and F1-score of the proposed model are 97.1%,92.6%,91.1% and 91.9%,respectively.