Intelligent paper money serial number identification method based on DSP

Abstract

The invention relates to the technical field of serial number identification, in particular to an intelligent paper money serial number identification method based on a DSP. The method comprises the steps that paper money images are collected; information such as the currency, current value, front and back sides, reversal and the like of paper money is judged; one-quarter images containing serial numbers are captured from the obtained paper money images; the one-quarter images are cut, and images only containing the paper money parts are split out; the serial number areas of the images of the paper money parts are subjected to rough location and precise location; binary serial number area images are obtained; character images are separated out from the serial number area images, when comparing with a character template image, similarity is calculated, and then characters in the character images are identified; the identified characters which are easily-confused character images are subjected to skeletal feature extraction, second judgement is conducted through skeletal features, and then the characters corresponding to the character images are determined; identified serial number characters are output by combining the template comparison result and the skeletal feature judgement result.

Description

technical field [0001] The invention relates to the technical field of serial number identification, in particular to a DSP-based intelligent identification method for banknote serial numbers. Background technique [0002] Most of the current mainstream serial number recognition methods use a single algorithm such as artificial neural network, bone feature, and feature point matching to realize the recognition of the serial number. [0003] Chinese patent publication number is CN105261110A invention patent discloses a kind of efficient DSP banknote serial number recognition method, and its specific steps are: [0004] a) Collect banknote images and find edge points of banknotes; [0005] b) Discharge the abnormal edge points found at the edge of the banknote, and fit the edge straight line to obtain the slope and intercept of the upper, lower, left and right sides of the banknote; [0006] c) Identify the face and direction of the banknote; [0007] d) Locate the serial n...

AI Technical Summary

Problems solved by technology

[0011] 1. The edge calculation method is complicated. The banknotes used in various countries are usually rectangular, that is, the upper and lower sides are parallel, and the left and right sides are parallel. When using the above method, a large number of edge points need to be sampled to fit the straight line equations of all edges. There is redundancy in its calculation amount;

[0012] 2. The artificial neural network has a large amount of calculation. The neural network model commonly used for character discrimination is at least a 3-4 layer model, that is, at least including an input layer, an output layer and an intermediate layer.

If dimensionality reduction is not performed, the number of points in the input layer of the algorithm is w c × h c The number of nodes in the output layer is 34, and there are more than 2 layers of intermediate layers, so there are a large number of floating point operations, and the performance requirements and budget real-time requirements of DSP are very high, which cannot be achieved by general systems.

Images