| The Cloud plays an important role in the regulation of the climate on Earth, and it also has an impact influence on the change of the climate. With the development of the photography technology and the digital image processing technology, several kinds of ground-based observation equipments have been developed, which are applied to take the all sky picture automatically and then generate digital images. As a result, the number of the captured ground-based digital cloud picture is increasingly large. However, at present, the only way of cloud pictures retrieval is skimming over cloud pictures by manual work. Undoubtedly, it is difficult to find the right picture considering the large size of the database. Therefore, the application demands in the retrieval and identification of content-based digital cloud pictures.With the help of the traditional CBIR(Content-Based Image Retrieval) technology, the research of local features of content-based digital cloud pictures and the feature extraction methods related to local features has been conducted. Further, the methods above have been applied to cloud retrieval system. The main research work falls into the following parts:1) The paper combines the local feature descriptor extraction technology, BoW Concept and the Inverted Index technology of the text search engine. Then analyzes the impact of different parameters on the local-features-based cloud image retrieval methods by comparing the test data.2) The paper combines the local feature descriptor extraction technology, BoW Concept and the Spatial Pyramid Matching technology. Then analyzes the impact of different parameters on the global-features-based cloud image retrieval methods by comparing the test data.3) The paper designs and implements experimental platform for cloud image retrieval, including the extraction of feature descriptor, descriptor clustering, quantification of the descriptor, the SPM eigenvector creating, inverted index, KD-Tree index and results evaluation system. |
PDF Full Text Request