Study Of Official Seal Recognition Technology Under Specific Background

The rapid development of the Internet and the popularity not only builds a free interactive platform for people but also trigger crimes. For example, criminals counterfeit authoritative department official seals and fabricated documents which published on official website in government departments. These fabricated government documents will guide the resulting in a very bad public opinion, and affect social stability and unity. Because most of these documents are image files, so in government websites, image censorship should be a necessary step before file uploading. This is the research background of this paper.The study object in this paper is color picture with particular background on Internet, and the research purpose is to identify the special seal in the background, such as pentagram, national emblem, and Party emblem. The so-called particular background is mainly common documents, government documents and licenses, and in this condition, the seal has authoritative and force of law.Identification process of the official seal included two major parts in this paper. Image preprocessing is aim to extract the seal outline in order to make preparations for seal-recognition. Pretreatment work consists of four steps, that's image segmentation, image enhancement red noise elimination and outline tracking. There're great difficulties in seal extracting because of the different picture quality between image files. So a new splitting way, which is based on RGB model and HSI model is proposed in this paper, using this way, the seal image can be extracted clearly. The experiments proved that this way has wide applicability among the three special backgrounds.In official seal recognition process, a new solution is presented in this paper. First, the basic character of seals is analyzed to determine the object region. Then, different recognition way is designed according to the object region. Recognition of national emblem is based on statistical nature, and the recognition of pentagram pattern and party emblem is based on its morphological character. If there're seals in the image, computing central coordinate, radius and seal category are followed.