基于卷积神经网络的地基云人工智能分类器研究

Research on ground-based cloud artificial intelligence classifier based on convolutional neural network

随着科学技术的发展,对地基云的研究也越来越深入,地基云的研究对天气预报、水资源管理、农业生产等领域具有重要的意义。传统地基云分类方法存在数据需求大、运行速率慢等问题。研究为了解决这些问题构建了一种融合双通道卷积神经网络(Convolutional Neural Networks,CNN)算法与压缩感知的地基云分类器。首先通过对CNN进行改进,得到双通道的CNN算法,然后将其与压缩感知进行融合,得到地基云人工智能分类器;最后通过不同方法进行对,以验证构建的地基云分类器对地基云的分类能力。结果表明在光线正常、光线较暗、水平角度、俯仰角度的观测前提下,地基云分类器的识别准确率平均值为73.95%、45.39%、92.61%和43.82%,均高于对照算法。这表明该地基云分类器具有较高的准确率和鲁棒性。

With the development of science and technology,the research on ground-based clouds has become more and more in-depth,and the research on ground-based clouds is important for weather forecasting,water resources management,agricultural production and other fields.Traditional ground-based cloud classification methods have problems such as large data requirements and slow operation rates.The study constructs a ground-based cloud classifier that combines a two-channel Convolutional Neural Networks(CNN)algorithm with compression-awareness in order to solve these problems.Firstly,a dual-channel CNN algorithm is obtained by improving the CNN,and then it is fused with compressive sensing to obtain a ground-based cloud artificial intelligence classifier;finally,a pair of different methods is conducted to verify the classification ability of the constructed ground-based cloud classifier for ground-based clouds.The results show that the average recognition accuracy of the ground-based cloud classifier is 73.95%,45.39%,92.61%and 43.82%under the observation premise of normal light,low light,horizontal angle and pitch angle,which are higher than the control algorithm.This indicates that this ground-based cloud classifier has high accuracy and robustness.