| With the development of multimedia technology, its utilization value for every profession gradually accepted, image as the media that what is utilization of the most widely used is unprecedented development. A wealth of information that carried by the image can make the person be clear. Image has the irreplaceable draw on text. With the massive image data increasing, how to effectively organize, manage and retrieve large data is urgent needed to address the problem.Base on this, we studied the algorithms and techniques of content-based aerial remote sensing image retrieval in the thesis。There are two critical issues for the technology of CBIR. This paper do research surround with these two problems. Feature representation and extraction is the basis of CBIR. This paper provides a comprehensive survey for these low-level features from two aspects——color and shape. After analyses and studies every extraction algorithm that the predecessors introduced, in color space feature extracted of image, we put forward HSV partition-based histogram feature extraction. HSV partition-based histogram full performance the image space distribution information.Through the experiment comparison we know that, HSV partition-based histogram achieve better outcome in the retrieval, it can improve performance. In shape space feature extracted of image, in this paper, we apply Hu invariant moments, Experiment results verify this method is application independent and effective to solve the problem brought out by image translation, scaling, rotation.One kind feature can not show the image totally. Integrate experiment conclusions we talked above, combined HSV partition-based histogram and Hu invariant moments, we use mufti-feature retrieval based exterior unitary manage. Experiment conclusions show that using mufti-feature achieves much better images retrieval than using a single image, at least as well as the best of a single image, improve the precision and recall.At last, based on theoretics analyzing, we design a test image retrieval framework system. In Experiment, we compare response time in each arithmetic.the results show that, as the system doesn't optimize database index structure, time will be done in 5 minutes; the time is satisfaction to them. As a whole, the system obtains the aim which we design expection before.This dissertation holds certain referential value and practical significance in promoting the development of retrieval technique of image database. |
