摘要: Six artificial neural network models are explored as potential methodologies forthe automated interpretation of satellite cloud images. The Multi-layer Perceptron Network is chosen to be the most applicable to the image interpreta... 展开 Six artificial neural network models are explored as potential methodologies forthe automated interpretation of satellite cloud images. The Multi-layer Perceptron Network is chosen to be the most applicable to the image interpretation problem. A complete, mathematical description of the methodology is presented. The neural network's classification capability is demonstrated using simple geometric patterns and alphabetic characters. A more complex test using GOES infrared imagery shows that the neural network can distinguish 53 of 54 large-scale cloud patterns. An architecture for a complete, automated cloud feature recognition system is proposed. 收起