The Recognition Of Rotated Multi-font Character Based On Eigen Space

With the rapid development of computer, multimedia and communication technologies, multimedia information based on image has become an important medium of information transmission. Text in the image contains important semantic information, For example, text instructions on road signs image and the palace name on the map all can indicate the location information associated with them. The title and author on the book cover, the product name and other information on the product packaging image, they all can explain the main content of them. Accurate extraction and recognition of image information can be applied in the fields of image retrieval system, vehicle automatic navigation system, visual disturbances auxiliary equipment, mobile phone auxiliary systems, etc. Therefore, accurate extraction and recognition of the text in the image has become a key issue in intelligent information processing, and has a wide range of applications.For the problems in the current text recognition system such as, For example, due to the resolution of different images of different equipment acquired its character recognition rate is also very different, when the image resolution is low or tilted image recognition rate will be greatly reduced, and some character recognition systems are also limited by text font type. This article will conduct research to identify the texts which is low resolution, containing a rotation of several fonts on image, and a character recognition method based on feature space was proposed. This method can improve the text recognition effect and improve the robustness of character recognition, so that it can be better applied in the field of machine vision.The main differences between character recognition based on feature space method and the traditional character recognition method are the methods of feature extraction and classification. The main contents of the method based on feature space are as follows: firstly, according to the principal component analysis to extract text features, create text feature space and determine the trajectory of the text; Secondly, the characters are classified according to the improved nearest neighbor method and their corresponding rotation angles are also identified. Finally, through the compared experiments with GCC and Hanwang OCR show that the character recognition method based on feature space to identify variety fonts of rotated texts has better results.