Aurora Image Classification Based On Adaptive Lifting Wavelet Variation And Local Binary Mode

The aurora phenomenon is colorful, and often appears in the Earth's polar regions. It is the concentrated reflection of meet the Sun and the Earth's magnetosphere role, and it is a better field for researching the sun on the earth.Studying aurora images morphological transform process and getting a lot of magnetosphere information from the Sun on the Earth's magnetosphere, it can be more in-depth understanding other functions. Moreover, the research also can help to know the weather process of transformation. In addition, according to observing satellites shows that the solar system other galaxies is also existing the aurora phenomenon, so studying aurora images, and extension it to other galaxies, providing a strong technical support to human exploration of the mysteries of the solar system other galaxies, the research meaning not only can't be measured, but also scientific value can't be matched for other phenomena.Aurora images classification is an important prerequisite to study the evolution of aurora image morphological, and it also has been highly described by scientists and other related research areas. Modern scientific innovation relying on the relevant field of classical research methods reference each other. On this point, the introduction of pattern recognition in aurora classified research methods, The great support aurora image classification researchin science.Pattern recognition is introduced not only can specific identify the aurora image,but can provide a good data to support the classification.The main purpose of this paper is to classify aurora images, so we introducing the two-scale algorithm adaptive lifting wavelet transform, meanwhile, and using the research method is that improved local binary pattern and fuzzy classification algorithm based on neighbor.Firstly, aurora images feature representation, adaptive wavelet transform algorithm to enhance the two-scale introduction of aurora image processing, so that the image characteristics of the aurora more obvious, and greatly reduce the interference factors, local binary pattern improved after the aurora used to extract feature information of the image, and generates a relative histogram. On this basis, the fuzzy neighbor classification applied to the classification of the aurora. Finally, Good aurora images in four kinds of classification of each selected 10 as the training sample,and then randomly selected 800 as the testing samples, and then the experimental samples and test samples were compared to verify the efficiency of its classification.In the experimental part, this paper verifys the efficiency of the proposed method to classify and also tests its robustness to noise for each test sample after adding noise. In both cases, the efficiency of this method is slightly higher than other methods of experimental comparison. In the end of the experiment, verifying time complexity of the various aurora classification algorithm, by experimenting, this paper has made relatively satisfactory test results...