Research And Implementation Of Image Retrieval Based On Trademark Content

Research on trademark content image retrieval technology is of great significance in today's society. In the application for registered trademarks, the use of image retrieval can help to avoid duplication of similar trademarks appearing in the trademark rights, to prevent deliberately counterfeit similar trademarks and to safeguard trademark rights.Therefore, the trademark retrieval system to speed up the registration of trademarks, to prevent similar forged trademarks and the protection of trademark rights and interests is of great significance.Trademark images have unique distinctive features and characteristics, that is, simple,significant, easy to identify the image, graphics or text symbols for the intuitive language.According to the characteristics of trademark images and the technical methods of image processing, the research work and achievements of this thesis are mainly reflected in the following aspects:(1) The analysis of trademark image retrieval status and related work. Through the application of cases and literature to understand the status quo of research questions,analysis and comparison of image processing feature description of the main methods of the pros and cons, for the work of this proposed research issues, feature description and vector composition methods, and classifiers and Similarity measure of the main technology.(2) The Base of the morphological characteristics of trademark images and the annotation of image types commonly used in the application system, it is proposed to realize the classification of trademark image types that is a method of directional histogram description and K nearest neighbor (KNN) classification based on Tamura texture feature. The experimental results show that this method can better describe the trend of text graphics, and can get enough information to correctly classify and identify the low error rate.(3) The Aim of the deficiency of trademark image and global image description in trademark image retrieval, a method based on SIFT and SURF local feature description method and image similarity measure method is proposed to realize the retrieval method of trademark feature description and matching metric. The experimental results show that the SIFT and SURF local features have good anti - jamming and can guarantee higher recognition rate and retrieval when image changes such as illumination, scale, rotation and blur appear.(4) The realization of trademark image retrieval system is to develop a simple version of the trademark-based image retrieval system. The system is simple, but it involves the database management technology, image processing technology, program interactive interface development technology, and this thesis studies image classification technology and image retrieval technology integration in the system.