Research On Information Extraction And Seal Identification Technology Of Bank Cheques

With image processing and pattern recognition technology continuing to develop, check automatic identification has gradually become the important financial automation component. Meanwhile standardization and integration trends of check will bring broader application to domestic check automatic identification. But domestic check automation level and recognition rate are relatively low, compared to the application of foreign checks. The full text is discussed by five chapters, including the introduction, check image preprocessing, extraction and identification of check stamps, check information extraction, check identification core interface design, as follows:Firstly, the chapter analyzes current research states and development trends of check identification techology, and discusses the domestic check features and business development direction, then points out the existing issues of check identification, purpose of research, significance of topics.Secondly, the check preprocessing part is discussed by frequency domain smoothing, local binarization, and tilt correction with the direction projection.Thirdly, the part give a few analysis and comparison of stamps extraction methods based on the gray and color space; then methods of optimization rotation registration for stamps are posed; At last, respectively based on Zernike moment invariant features weighted continental distance and the overall poor image matching recognition methods are raised for seals identification.Fourthly, this section tells about information extraction from check. At the first the chapter analyses the domestic bank check features, advances checks layout classification with BP Neural Network and Support Vector Machine(SVM); Then based on the use of blank check template difference from real check, the textual information filled in real one is extracted, which achieves a good performance.Finally, with development direction of software programming, interface designs of checks identification core is put forward against general demands of checks identification.By analyzing the experiment results, series of novel methods are advanced. This paper has posed tilt correction method with the direction projection, imprint extraction technology based on HSV color space, check layout classification methods, information extraction ways based on template of blank check and checks identification core interface design, which provide more abundant ways for check automatic recogonition.