Content-based Image Retrieval Applications

How to achieve intelligently non-fault driving, automation detection and emergence kicking off? At subway, high speed rail and light rail, how to effectively discover fault ahead of time? And then evaluate the situation by software of speciality in order to make non-faulty driving happen, specially all these cases rely on content based image retrieval, or CBIR, as the system may make use of the association between already known image for defect and the reason, characteristical & rectification. Then retrieve the target image by newly discovered suspicious images, and kick off accordingly the back up scheme if so.Because there is a gap between the symantical and the physical features in image, we need unite all the features into one so as to conquer the gap between the symantical and the physical features. The paper introduces the foundation on the study in the field of CBIR such as index fabrication, similarity checking, all the feedback method, image classification and feature substruction. The core technology is the substraction handling of features in CBIR, for example, in the shape one including block and perimeter, and in color one composing full layout and partial layout.The key point is index fabrication based on features already substracted so as to improve completeness rate and precision one.In CBIR, there are some new structures as R tree coming from B tree and quad tree so we can enquery image quickly in system. The exceptional characteristical is that it is a dynamical one as it facilitate addition, deletion and modification in real time.Fullly discusses the ways for color based retrieval, firstly bring in the color spaces and transformation and selection between them, then analyze the effects on index and storage, and the selection and space of digitalization, finally point out the con and pron of color based histogram. Then propose the enhanced solution, or the simulation based on histogram, finally hand over the best one, or the way for norm distribution consideration with weighted and errer-nonprone. In order to consider geograph arrange of color in image, firstly bring in block based histograph, then calculate the distance between blocks to justify matching between images. At the end, propose multi features combination retrieval with object oriented.Implement the algorithms in the organization where I had a intern. In the paper, introduce the framework, the design principal and functional module, and then analyze the testing data. At the last, bring out the next move toward the open question.