基于特征融合的随机森林飞机尾流识别

Random Forest Aircraft Wake Recognition Based on Feature Fusion

为提高民航运输的高效性和安全性,利用随机森林模型,结合多普勒激光雷达技术,提出一种基于径向速度极差特征和边缘轮廓特征融合的方法,实现对飞机尾流的准确识别。将在双流机场采集的数据样本进行速度极差特征提取,同时将样本数据映射成灰度图,通过形态学梯度提取图像轮廓特征,再将二者融合,并以此构建随机森林尾流识别模型,最后进行对比实验。实验结果表明,特征融合后随机森林模型的分类准确率、精确率、召回率、F1-score分别为95.8%、87.3%、89.4%、88.4%,高于单一特征方式和决策树模型识别结果。本文提出的方法能够对具有复杂背景风场中的尾涡进行检测。

To improve the efficiency and safety of civil aviation transportation, the accurate recognition of aircraft wake vortex should be realized. We use a random forest model, combining with Doppler lidar technology, and propose a feature fusion method based on radial velocity range and edge contour. Experiment extracts the speed range characteristics of the data samples collected at Shuangliu Airport, and at the same time maps the sample data into a grayscale image, and extracts the image contour features through morphological gradients to construct a random forest wake recognition model, which is the same as a single. The characteristic method and the decision tree are compared and tested. The experimental results show that the classification accuracy, precision, recall, and F1-score of the random forest model after feature fusion are 95.8%, 87.3%, 89.4%, and 88.4%, respectively, which are higher than the recognition results of single feature method and decision tree model. The established method can detect aircraft wake vortex in wind fields with complex backgrounds.