Aurora Images Retrieval Based On Features Inspired By Visual Cortex

Aurora is the only physical phenomenon which describes the properties of poles and can be captured by human eyes. The observation of its morphology and evolution can be used to obtain the information about of the sun-earth space electromagnetic activities. With the digital all-sky imager(ASI) emergence, millions of images are captured annually, providing important data source for aurora research. But if no efficient and accurate image retrieval tools is accessible, it is difficult for people to obtain aurora images they need in the massive image dataset, which leads to the insufficiency use of the dataset. However, existent image retrieval tools are text-based, and cannot meet the demand of retrieving the content of images. Therefore content-based aurora image retrieval is of importance for aurora research for geoscientists.First, an meta-data aurora image retrieval system is introduced. The system could retrieve aurora images according to the meta-data inputted by user. Then feature extraction of aurora images is studied. Features inspired by visual cortex are adopted and applied for image retrieval. Compared with the retrieval results based on other features, the retrieval precision of our method is higher than that of other features, and the index of callback is almost as same as that of others. Considering for the drawbacks of precision and callback in retrieval performance evaluation, the paper defined a new method, named User Satisfaction Degree(USD),for evaluation from human view to evaluation the performance for subjective measure. Experiments proves the features inspired by visual cortex get a higher user satisfaction degree than that of other features.