Study On RMB Image Information Recognition Algorithm

With the development of economy, the application of RMB classification and circulation system is becoming more and more widely, it played a huge role in replacing the manual labor and service masses. At present, there is still a gulf between the domestic classification equipment and the foreign products, it is urgent to improve the performance and reduce the cost. Under this background, the research on the RMB image recognition algorithm is of great theoretical and practical significance. The main purpose of the RMB image recognition algorithm is based on the characteristics of RMB image, studying deeply on the problem of the RMB image preprocessing, the recognition of face value& face direction, worn currency detection, the classification of old and new and finally reaching the expected requirements successfully.Due to the existence of interference of the RMB images which the CIS sensor acquires, in order to ensure the accuracy of the results, first of all, the preprocessing of RMB image is needed. It includes the correction and the position of RMB image, the smoothing of feature area sub-image. For the calibration and positioning parts, the method used is based on the tilt angle of the horizontal and vertical direction and the position of the four corners, solving the mapping matrix that original RMB images to standard images. The feature area sub-image is extracted according to the mapping matrix. After preprocessing, then recognizes face value& face direction of RMB image. Through selecting the suitable feature area and extracting the characteristics of the face value and face direction, the information of face value and face direction can be obtained after once recognition. In the experiment, the accuracy rate can reach one hundred percent in the case of a small amount of rejection is allowed. Then the damage of images is detected according to the information of face value& face direction of RMB images. The damage of RMB images includes defects?wear?damage and defacing of images. In order to improve the accuracy and efficiency of the algorithm, the processing of binaryzation and morphological is firstly carried out then block count the damaged area. The classification of the old-new of RMB images is one of the key problems of the RMB images recognition algorithm. According to the fuzzy degree difference between old images and new images, this paper puts forward an operator based on local average gray difference to represent the fuzzy difference between old and new RMB images. Then it uses the SVM which has good classification results of two class classification problems as the classifier for classification learning.Experimental results show that the RMB image recognition algorithm has achieved good results and it fully meets the expectations of the project in the case of a small amount of rejection is allowed.