Research Of Table Characters Recognition Based On Hough Transform And Neural Network

In actual application , many processing data are table text . As a common text , table text is a kind of text that is often processed in daily work , and is applied in all kinds of occasion catholically . For example , bank bills , reports of tax or finance , book cards , personnel archives , time sheets and so on . Therefore , the technology of characters recognition has a great value in research and application .In this paper , we research and explore the technology of automatic characters recogrition and presents a new method for characters recognition according to the property of characters' structure by using the duality of point-line in Hough transform. In this method, we use the number of the points which are on the same line and its corresponding coordinate in a character's parametric space as the feature input vector of BP neural network recognition system. To do this can reduce the dimensions of the feature vector greatly and can reflect the general distribution of the character's dot matrix and the essential feature of its figuration in the round. This method is simple and easy to do, and it has well equabliity. Besides this , we expand and improve on the above-mentioned method, design two systems for characters recognition , one is a system for printed charaters recognition based on skeleton thinning processing , the other is a system for characters recognition based on the characteristic of character's edge . The improved method reduced the redundant information of the character and the dimensions of its feature vector, and eliminate the noise on keeping the important feature of the original image's geometrical topology , so that it can make the character's architectural feature more clear and it is more advantageous for abstracting feature based on the character's architecture in this paper. This can increase the rate of recognition ,and at the same time,can reduce the amount of calculation, and can improve the speed of recognition. The results of system test show that these methods are feasible and availably .The method for characters recognition based on the Hough transform and BP neural network which is presented in this paper is a active exploer on the technology of characters recognition .It has some expanding space , and needs to be integrity . If it is combined with other methods of characters recognition , it is predicted to get a better result .