频高图Es层回波的自动识别

提出了一种基于决策树算法的自动识别Es层回波的方法以提高Es识别效率.通过自适应二值化以及中值滤波算法对频高图进行预处理，并提取有效的Es回波区域.利用Es层在虚高轴分布的规律，通过图像投影的方法选择出特征，将这些特征以及时间信息作为决策树算法的输入；通过人工识别方式为每幅频高图标明是否存在Es回波的标签，与决策树算法的输出进行比较.对训练集进行训练、剪枝，获取最接近人工识别结果的决策树，使用测试集对获取的决策树进行测试验证.本文使用云南普洱站（22.7°N，101.5°E）记录的频高图数据作为训练集，分别使用云南普洱站（22.7°N，101.5°E）和四川乐山站（29.5°N，103.7°E）的频高图数据作为测试集对该方法进行测试验证，结果表明该方法对普洱站和乐山站两地Es二跳回波的识别均具有较高的准确率，分别达到84.2%和82.8%. 

Automatical Scaling of Es Layer Echoes on Ionograms ormalsize

Sporadic E (Es) and its echoes have a big impact on the measurement and inversion of the F-layer echoes, and it also affect the short-wave communication. This paper proposed a method based on decision tree algorithm to automatically identify Es layer and its echoes on ionograms. First, the ionogram was preprocessed by adaptive binarization and median filtering algorithms, the effective Es echo region was extracted from ionograms. Moreover, with characteristics of Es layer echoes on ionograms, the features were selected by image projection method. The features of Es layer on the projection values of the virtual height and the occurrence time of the Es layer were used as input of the decision tree algorithm to construct the decision tree. And manually labeled tag on ionograms and compare it with the output of the decision tree algorithm. Then trained to get closer to the results of manual labeled. In the present study, ionograms recorded at Pu'er Station (22.7°N, 101.5°E) in Yunnan province were used to construct the decision tree.the ionograms of Yunnan Pu'er Station (22.7°N, 101.5°E) and Sichuan Leshan Station (29.5°N, 103.7°E) were used as test sets respectively. The method was tested and verified. That method has high accuracy for the identification of Es two-hop echoes in both Pu'er and Leshan stations, reaching 84.2% and 82.8% respectively.