Feature Extraction And Retrieval Based On Sun Image

In recent years, with the development of the observation techniques and equipment of solar image, the quantity and quality of the solar image is growing more and more faster. For example, the American Solar Dynamics Observatory mission, with its Atmospheric Imaging Assembly (AIA), is generating eight 4096 pixels x 4096 pixels images every 10 seconds, leading to a data transmission rate of approximately 700 Gigabytes per day from only the AIA component. These data are generally named according to the time that the data generated, how to retrieve from these massive solar image data to a particular image has become a serious problem. Therefore, content-based image retrieval system in the field of solar image processing also has broad application prospects. The main work of this paper consists of two aspects.First, from the start with content-based image retrieval, we research some feature describe methods such as color, texture, shape and spatial relationships. On this basis, the article conducted a preliminary research on the solar image retrieval. According to the particularity of solar image, we use the texture features as solar image features. Then, this paper proposes a method that fusion GLCM and Gabor wavelets to extract texture features. The weight of the two features is determined by the ratio of their precision, the feature which has a high precision has relatively large weight. Then we established a CBIR system which is based on MATLAB, the experimental results show that the fusion feature has a good results.Second, in order to improve the retrieval results, most of today’s CBIR systems have relevance feedback function, generally use artificial feedback. Given the special nature of the sun image, adopt the artificial feedback method has some difficult, maybe people will give some error feedback. Therefore, this paper proposes three weights in adaptive algorithm, the system can automatically adjust the weights of adjustment and reorientation, then by setting a certain termination condition is retrieved to end without human intervention. We then compared with the previous search results retrieved by the experimental results after adjustment for weight and determining the optimal algorithm, the results showed that this method has a good effect.