Research Of Trademark Image Retrieval Based On Content

With the rapid development of Chinese market economy, the trademark requirements increased year by year. The traditional trademark retrieval method based on Classification Codes use lot of human resources can not solve the contradiction of the trademark registration. The content-based trademark retrieval which using computer vision technology and some computer-aided knowledge provides a good way to solve the current trademark registration problem.The current research status of the trademark image retrieval technology based on content is introduced and analyzed. The traditional feature extraction methods and the Scale Invariant Feature Transform method both be studied and analyzed. The experiment shows that the Scale Invariant Feature Transform method performs better than traditional methods. But this method to match two images by relative distance calculation, this explicit method causes time-consuming, being impacted easily by the complexity of image and some other problems. A new feature matching algorithm by using space pyramid encoding is proposed to solve these problems. Experimental results show that the algorithm not only can maintain a good matching accuracy, but also greatly reduce the matching time.In order to test the actual retrieval effect of the feature matching algorithm proposed in this paper, a online trademark retrieval system was designed and implemented. The system achieved good results by creating image features offline database and using space pyramid encoding method, which not only verify the feasibility of the method, but also make an in-depth attempt to take the trademark image retrieval technology based on content to practical from research.