【摘要】 针对现有冰雹检测算法过分依赖专家经验的问题。本文提出一种基于卷积神经网络的双极化气象雷达冰雹检测方法。首先对单个雷达分辨单元分割并利用相邻分辨单元的极化信息进行填充，将极化信息转化为数据矩阵。接着搭建卷积神经网络模型进行特征提取和分类，然后用样本数据对网络进行训练调整模型参数。最后使用训练好的模型进行冰雹检测。通过仿真数据和实测数据的实验结果表明，该方法能够在提取数据特征和自适应调节网络参数基础上有效进行冰雹检测。更多还原

【Abstract】 Current hail detection algorithm relies too much on expert experience. In this paper, a hail detection method based on convolutional neural network is proposed for dual-polarization weather radar. Firstly, a single radar resolution cell is segmented and the polarization information of adjacent resolution cells is used for data filling, transforming the polarization information into a data matrix. Secondly, a convolutional neural network model is built for feature extraction and classification, and then the sample data is used to train the network and adjust the model parameters. Finally, the trained model is used for hail detection. The trained model is tested using both simulation data and measured data. The test results show that the proposed method can effectively detect hail due to data feature extraction and adaptive network parameter adjustment.更多还原