CN111738079A - Banknote denomination recognition method and device - Google Patents
Banknote denomination recognition method and device Download PDFInfo
- Publication number
- CN111738079A CN111738079A CN202010425302.9A CN202010425302A CN111738079A CN 111738079 A CN111738079 A CN 111738079A CN 202010425302 A CN202010425302 A CN 202010425302A CN 111738079 A CN111738079 A CN 111738079A
- Authority
- CN
- China
- Prior art keywords
- banknote
- image
- similarity
- full
- difference
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Inspection Of Paper Currency And Valuable Securities (AREA)
Abstract
The invention discloses a banknote denomination recognition method and a banknote denomination recognition device. The training part is characterized in that the training part can automatically extract features after collecting enough samples, so that human intervention is not needed, and the recognition efficiency is greatly improved. The identification portion is characterized by comprising: acquiring graph edge information; finding a graph coordinate through edge information fitting, and performing plane transformation according to coordinate information to obtain a transformed graph coordinate; acquiring and calculating one-dimensional characteristic information of the image according to a preset method; and matching with the one-dimensional characteristic information of the standard image to judge whether the matched characteristic is abnormal. The invention converts the characteristic two-dimensional characteristics into a plurality of one-dimensional characteristics, calculates the denomination one-dimensional characteristics, and matches the calculated one-dimensional characteristics with the standard denomination one-dimensional characteristics. The invention is independent of image type, is not influenced by CIS type and new and old difference of bank notes, and has strong adaptability to old bank notes, defective bank notes and stained bank notes.
Description
Technical Field
The invention relates to the field of image recognition, in particular to a method and a device for recognizing the denomination of a banknote.
Background
At present, more than 170 kinds of bank notes are arranged in the world, and various kinds of cash vouchers are added, the identification types are extremely large, meanwhile, each kind of bank note has a life cycle of only 3-10 years, the bank notes can be updated regularly, and a large amount of bank notes need to be added with identification every year. Meanwhile, denomination recognition is a basic function of the currency counting machine, the denomination of the currency needs to be recognized quickly and accurately in the currency counting process, the recognition rate is extremely high, and the denomination problem is basically not allowed to occur.
At present, the general identification method is used for distinguishing the denominations of various characteristics of various banknotes one by one, and under various environments of multiple countries, the workload is large, the identification effect is not ideal, and various limitations exist under the conditions of incomplete banknotes, partial banknotes and breakage, which can cause that the denominations cannot be accurately identified.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a method and a device for identifying the denomination of banknotes, and aims to solve the problem of a universal global feature identification mode of various banknotes, greatly simplify the identification process and simultaneously have a better identification effect on old banknotes and damaged banknotes.
The invention is realized by the following steps: the invention discloses a banknote denomination recognition method, which comprises the following steps:
acquiring an image of a bank note to be identified;
extracting full-width characteristic data of the banknote image;
and carrying out similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data, and selecting the sample data with the highest comprehensive similarity as a final banknote information result.
The invention discloses a banknote denomination recognition method, which comprises the following steps:
acquiring an image of a bank note to be identified;
extracting full-width characteristic data of the banknote image;
similarity calculation is carried out on the extracted full-width characteristic data of the banknote image and preset sample data of partial denomination versions extracted from each currency, and the currency of the banknote to be identified is preliminarily screened out;
and carrying out similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data corresponding to each denomination version of the currency screened preliminarily, and selecting the banknote with the highest comprehensive similarity as a final banknote information result.
In multinational currency identification, the denomination is acquired for the first time but the accuracy of the denomination at the time is to be evaluated, so that the denomination is matched for the second time, and local comprehensive matching is performed for the second time according to the first result. The invention carries out matching in 2 steps, can improve the speed, and directly and completely matches the state of multinational currency with more time consumption.
For example, when the method is used for identifying the American English 3 national currency, the special data of partial denomination versions of each currency is extracted for matching during the first similarity calculation, the obtained result is followed, for example, the obtained result is a 5-yuan version of British pounds, and then all the British pounds are selected for matching again during the second similarity calculation.
And further, after a final banknote information result is obtained, judging whether the similarity score and the difference score meet the requirements, if the similarity score and the difference score meet the requirements, judging that the identification is normal, and if at least one of the similarity score and the difference score does not meet the requirements, judging that the banknote is suspected. The suspected representative denomination identification may not be trustworthy, or may be of low confidence.
The method for calculating the difference score comprises the following steps: and performing similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data corresponding to each denomination version of the currency screened preliminarily to obtain a plurality of similarity scores, wherein the difference score refers to the difference between the similarity score with the highest similarity and other similarity scores, and when the similarity score with the highest similarity does not meet the requirement or the minimum difference score does not meet the requirement, the banknote is considered to be a suspected banknote. And when the difference score is small, namely the difference score is less than or equal to the set value, the difference score is considered to be not satisfied. And when the similarity score of the highest similarity is larger than a set value, the similarity score of the highest similarity is considered to be not satisfied with the requirement.
The higher the similarity score is, the lower the similarity is, and when the similarity score is 0, the two pictures are considered to be identical.
Further, extracting the full-width feature data of the banknote image specifically comprises: drawing the shape of the bank note, finding the edge characteristics of the graph to obtain the coordinates of four corner points of the image of the bank note, giving out preliminary information judgment of the bank note through shape detection, obtaining plane transformation parameters of the bank note according to the coordinates, carrying out linear-like plane transformation on the coordinate information of the image of the bank note, zooming the picture to a set size, then carrying out discrete cosine transformation on the picture zoomed to the set size, converting the picture to a frequency domain, and obtaining full-width characteristic data.
Further, similarity calculation is carried out on the extracted full-width feature data of the banknote image and preset sample data, and the similarity calculation specifically comprises the following steps: the method comprises the steps that the full-width characteristic data and sample data of a banknote image are frequency domain matrixes, the coefficients of a set number of middle and low frequency regions are extracted from the frequency domain matrixes, the arithmetic mean value D is calculated, each value in the frequency domain matrixes is compared with D, the value is larger than D and is 1, the value is smaller than D and is 0, the image information can be obtained through storage according to bits, the number capable of quantifying the difference degree of the images is obtained through calculating the difference between the image information of the banknote image and the image information of the sample data, when the difference is 0, the two images are considered to be identical, and the difference is larger, the more dissimilar the images are indicated.
Furthermore, the method also comprises the step of correspondingly loading preset sample data according to the set currency before the image of the banknote to be identified is acquired.
The invention discloses a banknote denomination recognition device, which comprises a banknote image acquisition module, a banknote image acquisition module and a banknote recognition module, wherein the banknote image acquisition module is used for acquiring an image of a banknote to be recognized;
the characteristic extraction module is used for extracting the full-width characteristic data of the banknote image;
and the similarity calculation module is used for performing similarity calculation on the full-width characteristic data of the extracted banknote image and preset sample data, and selecting the banknote image with the highest comprehensive similarity as a final banknote information result.
The invention discloses a banknote denomination recognition device, which comprises a banknote image acquisition module, a banknote image acquisition module and a banknote recognition module, wherein the banknote image acquisition module is used for acquiring an image of a banknote to be recognized;
the characteristic extraction module is used for extracting the full-width characteristic data of the banknote image;
the similarity calculation module is used for carrying out similarity calculation on the full-width feature data of the extracted banknote image and preset sample data of partial denomination versions extracted from each currency, and preliminarily screening out the currency of the banknote to be identified; and the similarity calculation module is used for carrying out similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data corresponding to each denomination version of the currency screened preliminarily, and selecting the banknote with the highest comprehensive similarity as a final banknote information result.
The banknote denomination recognition device further comprises a confirmation module, wherein the confirmation module is used for judging whether the similarity score and the difference score meet the requirements, if the similarity score and the difference score meet the requirements, the recognition is considered to be normal, and if at least one of the similarity score and the difference score does not meet the requirements, the banknote is considered to be a suspected banknote.
The invention also comprises a sample file loading module which is used for correspondingly loading preset sample data according to the set currency.
Further, the feature extraction module is used for drawing the shape of the bank note, finding the edge features of the graph to obtain the coordinates of four corner points of the image of the bank note, giving out preliminary information judgment of the bank note through shape detection, obtaining plane transformation parameters of the bank note according to the coordinates, carrying out line-like plane transformation on the coordinate information of the image of the bank note, zooming the picture to a set size, then carrying out discrete cosine transformation on the picture zoomed to the set size, and converting the picture into a frequency domain to obtain full-width feature data.
Further, the similarity calculation module is used for extracting the coefficients of a set number of middle and low frequency regions from the frequency domain matrix of the full-width characteristic data of the banknote image and the sample data, calculating an arithmetic mean value D, comparing each value in the frequency domain matrix with the value D, counting the value D to 1, counting the value D to 0, storing the values according to positions to obtain picture information, obtaining a number capable of quantifying the difference degree of the pictures by calculating the difference between the picture information of the banknote image and the picture information of the sample data, and when the difference is 0, considering that the two pictures are completely the same, and the difference is larger, so that the pictures are more dissimilar.
The invention has the beneficial effects that: the invention adopts the following steps: acquiring original data of an image; drawing the appearance of the bank note, finding the edge characteristics of the graph, carrying out coordinate transformation, and zooming the graph to an ideal state to obtain coordinate information; obtaining the image after coordinate transformation, extracting global feature data, and performing similarity calculation on the feature data and the loaded sample data; screening information of a plurality of sample files with higher similarity through similarity to serve as currency information alternatives of the actually acquired graphs; and calculating the similarity of the whole breadth by comparing the frequency domain information of the alternative information, and selecting the highest comprehensive similarity as a final currency information result, namely acquiring information such as the denomination, version direction, breadth and the like of the graph. The invention is independent of image type, is not influenced by CIS type and new and old difference of paper money, and has strong adaptability to old paper money, defective paper money and stained paper money.
The invention aims to solve the problem that a universal global feature recognition mode of various banknotes can be used for carrying out multinational currency denomination recognition, when multinational currency recognition is carried out, the denomination is acquired for the first time, but the accuracy of the denomination at the moment is to be evaluated, so that the denomination is matched for the second time, and local comprehensive matching is carried out for the second time according to the first result. The time consumption is high in the multi-country currency state of direct and complete matching, the method is matched in two steps, the speed can be increased, the identification process is greatly simplified, and meanwhile, the method can achieve a good identification effect on old and damaged banknotes. The invention can automatically extract the characteristics after collecting enough samples without human intervention, greatly improves the recognition efficiency, and can quickly add various banknotes while ensuring the recognition rate through a training mode.
Drawings
FIG. 1 is a general flow chart of the banknote denomination identification method of the present invention;
FIG. 2 is a flow chart of the present invention for obtaining four corner coordinates of a banknote image;
FIG. 3 is a functional block diagram of the banknote denomination recognition arrangement of the present invention;
FIG. 4 is a diagram illustrating a discrete cosine transform according to the present invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
Example one
Referring to fig. 1, the invention discloses a multinational currency denomination identification method, which comprises the following steps:
setting currency types, and correspondingly loading preset sample data according to the set currency types;
acquiring an image of a bank note to be identified;
extracting full-width characteristic data of the banknote image;
matching for the first time: similarity calculation is carried out on the extracted full-width characteristic data of the banknote image and preset sample data of partial denomination versions extracted from each currency, and the currency of the banknote to be identified is preliminarily screened out;
and (3) second matching: and carrying out similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data corresponding to each denomination version of the currency screened preliminarily, and selecting the banknote with the highest comprehensive similarity as a final banknote information result.
If the similarity between one sample data (50 yuan if all) of the currencies of America, Europe, English, China and the like and the whole feature data of the extracted banknote image is selected for the first matching, the similarity calculation is carried out to primarily determine which currency belongs to, such as the currency of China, then, the similarity calculation is carried out to each sample data of the currency of China and the whole feature data of the extracted banknote image when the matching is carried out for the second time, and the sample data with the highest comprehensive similarity is selected as the final banknote information result.
Further, sample data is obtained through an upper computer training sample, and the training main process comprises the following steps:
1. a certain amount of sample data is collected.
2. And classifying the sample data and obtaining the denomination version direction of the sample data. And then training the global features through training software, and meanwhile, reversely calculating the similarity and the score by using the training result, if the similarity and the score meet the requirements, considering that the normal data is the sample feature file extracted through the training software.
3. The generated feature file is imported into the machine and automatically takes effect, and the target currency can be normally identified.
After a sufficient number of samples are collected according to requirements, the samples are stored according to the directory rule in different directions, and after a path is selected, the processing is started by clicking, and then bin files with sample characteristics can be generated in the designated directory. Each template corresponds to a direction of a version of a denomination in one currency, exemplified by renminbi.
There are six denominations of 1 yuan, 5 yuan, 10 yuan, 20 yuan, 50 yuan and 100 yuan, three versions of 5 yuan, 10 yuan, 50 yuan and 100 yuan, 1990 version, 1999 version and 2005 version, two versions of 20 yuan, 100 yuan also 2015 version and space coin, 1 yuan only 1999 version, and four directions in total, so there are 68 templates in total. And each template needs a minimum of 20 sample data, and the total sample data of 20 × 68 ═ 1360 is required by adding the banknotes of one country, so that the training of one kind of banknotes can be realized, and the full recognition of the banknotes can be realized. Meanwhile, the dependence on the banknote sample can be reduced by repeatedly collecting data, and the banknote can be added under the condition of only a small amount of samples.
And further, after a final banknote information result is obtained, judging whether the similarity score and the difference score meet the requirements, if the similarity score and the difference score meet the requirements, judging that the identification is normal, and if at least one of the similarity score and the difference score does not meet the requirements, judging that the banknote is suspected.
The method for calculating the difference score comprises the following steps: and performing similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data corresponding to each denomination version of the currency screened preliminarily to obtain a plurality of similarity scores, wherein the difference score refers to the difference between the similarity score with the highest similarity and other similarity scores, and when the similarity score with the highest similarity does not meet the requirement or the minimum difference score does not meet the requirement, the banknote is considered to be a suspected banknote. And when the difference score is small, namely the difference score is less than or equal to the set value, the difference score is considered to be not satisfied. And when the similarity score of the highest similarity is larger than a set value, the similarity score of the highest similarity is considered to be not satisfied with the requirement.
The higher the similarity score is, the lower the similarity is, and when the similarity score is 0, the two pictures are considered to be identical.
Further, extracting the full-width feature data of the banknote image specifically comprises: drawing the shape of the bank note, finding the edge characteristics of the graph to obtain the coordinates of four corner points of the image of the bank note, giving out preliminary information judgment of the bank note through shape detection, obtaining plane transformation parameters of the bank note according to the coordinates, carrying out linear-like plane transformation on the coordinate information of the image of the bank note, zooming the picture to a set size, then carrying out discrete cosine transformation on the picture zoomed to the set size, converting the picture to a frequency domain, and obtaining full-width characteristic data.
And obtaining coordinates of four corner points of the banknote image, giving out preliminary information judgment of the banknote through shape detection, and obtaining plane transformation parameters of the banknote according to the coordinates for region matting.
Referring to fig. 2, depicting the shape of the banknote, finding the edge features of the graph, and obtaining the coordinates of four corner points of the banknote image specifically include: and respectively searching the IR channel images for edge point sequences of four edges of the banknote image from four directions, wherein the edge point sequences comprise an upper edge point sequence, a lower edge point sequence, a left edge point sequence and a right edge point sequence.
And (5) fitting a straight line by using a least square method for the edge point sequence to obtain the k, b, deg, confidence values of four edges.
The straight line describes the equation Y — KX + B, k is slope, B is offset, Deg is angle, and confidence is confidence.
Carrying out linear-simulated plane transformation on the image coordinate information;
the linear planar transformation is a linear transformation from two-dimensional coordinates to two-dimensional coordinates, which maintains the "straightness" (i.e., straight lines remain straight lines after transformation) and the "parallelism" (i.e., the relative positional relationship between two-dimensional patterns remains unchanged, parallel lines remain parallel lines, and the positional order of points on the straight lines does not change) of two-dimensional patterns. Through this transformation, rotation, translation, and scaling operations may be performed on the image.
Specifically, in this embodiment, in order to obtain global information of an image, the following one-dimensional feature extraction is facilitated, and a plane transformation formula is:
from the matrix coordinates, we derive:
x′0＝a00x0+a01y0+a02
y′0＝a10x0+a11y0+a12
x′1＝a00x1+a01y1+a02
y′1＝a10x1+a11y1+a12
x′2＝a00x2+a01y2+a02
y′2＝a10x2+a11y2+a12
obtaining:
the plane transformation parameters are obtained by the gaussian elimination method to complete image correction, and it can be understood that the matrix parameters are different for different transformation requirements, which is not limited in this embodiment.
And carrying out discrete cosine transform on the acquired image to acquire full-width feature data.
Discrete cosine transform is an important method for simplifying fourier transform, and it can be known from the property of fourier transform that when a continuous real function f (x) or f (xy) is an even function, the calculation of the transform has only a cosine term, so that cosine transform has definite physical significance as fourier transform, and cosine transform is a special case of fourier transform. Complex operations in fourier transforms, which are real-based orthogonal transforms, are avoided. The basis vectors of the transformation matrix closely approximate the Toeplitz matrix (a feature vector where the coefficient matrix is symmetric and the elements are equal along any diagonal parallel to the main diagonal, which in turn embodies the relevant characteristics of human speech and image signals, this transformation is often considered a quasi-optimal transformation of the image signals.
Referring to fig. 4, from the spatial domain to the frequency domain, the discrete cosine transform is defined by the following formula, where N and M are the number of rows and columns of the matrix.
where
and
p, q ═ discrete frequency variables;
amn is the pixel value of the image in m rows and n columns;
bpq is the result of the discrete cosine transform.
The discrete cosine transform is performed in the transformed matrix after some special processing, so that the operation is relatively simple, the real-time performance is met, meanwhile, the expandability is strong, the newly added template only needs to collect enough samples and retrain the template, almost all currency types are suitable, when the image has no obvious quality problem or deformation, the performance is relatively reliable, even if the problem occurs, the problem can be rapidly solved through retrain, the interpretability is strong when the problem occurs, and the understanding is convenient.
Further, similarity calculation is carried out on the full-width feature data of the extracted banknote image and preset sample data, and a full-width similarity score is obtained, and the method specifically comprises the following steps: the full-width characteristic data and the sample data of the banknote image are frequency domain matrixes, and the method is realized by comparing frequency domain information when the image difference is compared. In the implementation of this embodiment, the picture is first scaled to a specified size, then discrete cosine transform is applied to convert the picture to the frequency domain, a certain number of coefficients of the medium-low frequency region are extracted from the frequency domain matrix, and an arithmetic mean value D is calculated. Such as the formula:
and comparing each value in the matrix with D, counting more than D and 1 and less than D and counting 0, and storing according to bits to obtain picture information. By calculating the difference between the two picture information, a number that can quantify the picture difference can be obtained. When the difference is 0, it can be considered that the two pictures are identical, and the larger the difference is, the more dissimilar the pictures are.
Example two
The invention discloses a banknote denomination recognition method, which comprises the following steps:
setting currency types, and correspondingly loading preset sample data according to the set currency types;
acquiring an image of a bank note to be identified;
extracting full-width characteristic data of the banknote image;
and carrying out similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data, and selecting the sample data with the highest comprehensive similarity as a final banknote information result.
And further, after a final banknote information result is obtained, judging whether the similarity score and the difference score meet the requirements, if the similarity score and the difference score meet the requirements, judging that the identification is normal, and if at least one of the similarity score and the difference score does not meet the requirements, judging that the banknote is suspected.
Further, extracting the full-width feature data of the banknote image specifically comprises: drawing the shape of the bank note, finding the edge characteristics of the graph to obtain the coordinates of four corner points of the image of the bank note, giving out preliminary information judgment of the bank note through shape detection, obtaining plane transformation parameters of the bank note according to the coordinates, carrying out linear-like plane transformation on the coordinate information of the image of the bank note, zooming the picture to a set size, then carrying out discrete cosine transformation on the picture zoomed to the set size, converting the picture to a frequency domain, and obtaining full-width characteristic data.
Further, similarity calculation is carried out on the extracted full-width feature data of the banknote image and preset sample data, and the similarity calculation specifically comprises the following steps: the method comprises the steps that the full-width characteristic data and sample data of a banknote image are frequency domain matrixes, the coefficients of a set number of middle and low frequency regions are extracted from the frequency domain matrixes, the arithmetic mean value D is calculated, each value in the frequency domain matrixes is compared with D, the value is larger than D and is 1, the value is smaller than D and is 0, the image information can be obtained through storage according to bits, the number capable of quantifying the difference degree of the images is obtained through calculating the difference between the image information of the banknote image and the image information of the sample data, when the difference is 0, the two images are considered to be identical, and the difference is larger, the more dissimilar the images are indicated.
EXAMPLE III
The invention discloses a banknote denomination recognition device, which comprises a banknote image acquisition module, a banknote image acquisition module and a banknote recognition module, wherein the banknote image acquisition module is used for acquiring an image of a banknote to be recognized;
the characteristic extraction module is used for extracting the full-width characteristic data of the banknote image;
and the similarity calculation module is used for performing similarity calculation on the full-width characteristic data of the extracted banknote image and preset sample data, and selecting the banknote image with the highest comprehensive similarity as a final banknote information result.
The banknote denomination recognition device further comprises a confirmation module, wherein the confirmation module is used for judging whether the similarity score and the difference score meet the requirements, if the similarity score and the difference score meet the requirements, the recognition is considered to be normal, and if at least one of the similarity score and the difference score does not meet the requirements, the banknote is considered to be a suspected banknote.
The invention also comprises a sample file loading module which is used for correspondingly loading preset sample data according to the set currency.
Further, the feature extraction module is used for drawing the shape of the bank note, finding the edge features of the graph to obtain the coordinates of four corner points of the image of the bank note, giving out preliminary information judgment of the bank note through shape detection, obtaining plane transformation parameters of the bank note according to the coordinates, carrying out line-like plane transformation on the coordinate information of the image of the bank note, zooming the picture to a set size, then carrying out discrete cosine transformation on the picture zoomed to the set size, and converting the picture into a frequency domain to obtain full-width feature data.
Further, the similarity calculation module is used for extracting the coefficients of a set number of middle and low frequency regions from the frequency domain matrix of the full-width characteristic data of the banknote image and the sample data, calculating an arithmetic mean value D, comparing each value in the frequency domain matrix with the value D, counting the value D to 1, counting the value D to 0, storing the values according to positions to obtain picture information, obtaining a number capable of quantifying the difference degree of the pictures by calculating the difference between the picture information of the banknote image and the picture information of the sample data, and when the difference is 0, considering that the two pictures are completely the same, and the difference is larger, so that the pictures are more dissimilar.
Example four
Referring to fig. 3, the invention discloses a banknote denomination recognition apparatus, comprising a banknote image acquisition module, a banknote recognition module and a banknote recognition module, wherein the banknote image acquisition module is used for acquiring an image of a banknote to be recognized;
the characteristic extraction module is used for extracting the full-width characteristic data of the banknote image;
the similarity calculation module is used for carrying out similarity calculation on the full-width feature data of the extracted banknote image and preset sample data of partial denomination versions extracted from each currency, and preliminarily screening out the currency of the banknote to be identified; and carrying out similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data corresponding to each denomination version of the currency screened preliminarily, and selecting the banknote with the highest comprehensive similarity as a final banknote information result.
The banknote denomination recognition device further comprises a confirmation module, wherein the confirmation module is used for judging whether the similarity score and the difference score meet the requirements, if the similarity score and the difference score meet the requirements, the recognition is considered to be normal, and if at least one of the similarity score and the difference score does not meet the requirements, the banknote is considered to be a suspected banknote.
The invention also comprises a sample file loading module which is used for correspondingly loading preset sample data according to the set currency.
Further, the feature extraction module is used for drawing the shape of the bank note, finding the edge features of the graph to obtain the coordinates of four corner points of the image of the bank note, giving out preliminary information judgment of the bank note through shape detection, obtaining plane transformation parameters of the bank note according to the coordinates, carrying out line-like plane transformation on the coordinate information of the image of the bank note, zooming the picture to a set size, then carrying out discrete cosine transformation on the picture zoomed to the set size, and converting the picture into a frequency domain to obtain full-width feature data.
Further, the similarity calculation module is used for extracting the coefficients of a set number of middle and low frequency regions from the frequency domain matrix of the full-width characteristic data of the banknote image and the sample data, calculating an arithmetic mean value D, comparing each value in the frequency domain matrix with the value D, counting the value D to 1, counting the value D to 0, storing the values according to positions to obtain picture information, obtaining a number capable of quantifying the difference degree of the pictures by calculating the difference between the picture information of the banknote image and the picture information of the sample data, and when the difference is 0, considering that the two pictures are completely the same, and the difference is larger, so that the pictures are more dissimilar.
The invention can automatically extract the characteristics after collecting enough samples without human intervention, thereby greatly improving the identification efficiency. The invention comprises the following steps: acquiring graph edge information; finding a graph coordinate through edge information fitting, and performing plane transformation according to coordinate information to obtain a transformed graph coordinate; acquiring and calculating one-dimensional characteristic information of the image according to a preset method; and matching with the one-dimensional characteristic information of the standard image to judge whether the matched characteristic is abnormal. The invention converts the characteristic two-dimensional characteristics into a plurality of one-dimensional characteristics, calculates the denomination one-dimensional characteristics, and matches the calculated one-dimensional characteristics with the standard denomination one-dimensional characteristics. The invention is independent of image type, is not influenced by CIS type and new and old difference of paper money, and has strong adaptability to old paper money, defective paper money and stained paper money.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A banknote denomination identification method, comprising the steps of:
acquiring an image of a bank note to be identified;
extracting full-width characteristic data of the banknote image;
and carrying out similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data, and selecting the sample data with the highest comprehensive similarity as a final banknote information result.
2. A banknote denomination identification method, comprising the steps of:
acquiring an image of a bank note to be identified;
extracting full-width characteristic data of the banknote image;
similarity calculation is carried out on the extracted full-width characteristic data of the banknote image and preset sample data of partial denomination versions extracted from each currency, and the currency of the banknote to be identified is preliminarily screened out;
and carrying out similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data corresponding to each denomination version of the currency screened preliminarily, and selecting the banknote with the highest comprehensive similarity as a final banknote information result.
3. The banknote denomination recognition method according to claim 1 or 2, wherein: after the final banknote information result is obtained, judging whether the similarity score and the difference score meet the requirements or not, if the similarity score and the difference score meet the requirements, judging that the identification is normal, and if at least one of the similarity score and the difference score does not meet the requirements, judging that the banknote is suspected;
the method for calculating the difference score comprises the following steps: and performing similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data corresponding to each denomination version of the currency screened preliminarily to obtain a plurality of similarity scores, wherein the difference score refers to the difference between the similarity score with the highest similarity and other similarity scores, and when the similarity score with the highest similarity does not meet the requirement or the minimum difference score does not meet the requirement, the banknote is considered to be a suspected banknote.
4. The banknote denomination recognition method according to claim 1 or 2, wherein: the method for extracting the full-width characteristic data of the banknote image specifically comprises the following steps: drawing the shape of the bank note, finding the edge characteristics of the graph to obtain the coordinates of four corner points of the image of the bank note, giving out preliminary information judgment of the bank note through shape detection, obtaining plane transformation parameters of the bank note according to the coordinates, carrying out linear-like plane transformation on the coordinate information of the image of the bank note, zooming the picture to a set size, then carrying out discrete cosine transformation on the picture zoomed to the set size, converting the picture to a frequency domain, and obtaining full-width characteristic data.
5. The banknote denomination recognition method according to claim 1 or 2, wherein: similarity calculation is carried out on the extracted full-width characteristic data of the banknote image and preset sample data, and the similarity calculation specifically comprises the following steps: the method comprises the steps that the full-width characteristic data and sample data of a banknote image are frequency domain matrixes, the coefficients of a set number of middle and low frequency regions are extracted from the frequency domain matrixes, the arithmetic mean value D is calculated, each value in the frequency domain matrixes is compared with D, the value is larger than D and is 1, the value is smaller than D and is 0, the image information can be obtained through storage according to bits, the number capable of quantifying the difference degree of the images is obtained through calculating the difference between the image information of the banknote image and the image information of the sample data, when the difference is 0, the two images are considered to be identical, and the difference is larger, the more dissimilar the images are indicated.
6. The banknote denomination recognition method according to claim 1 or 2, wherein: and loading preset sample data correspondingly according to the set currency before acquiring the image of the bank note to be identified.
7. A banknote denomination recognition device, characterized by: the identification system comprises a banknote image acquisition module, a banknote image acquisition module and a banknote identification module, wherein the banknote image acquisition module is used for acquiring an image of a banknote to be identified;
the characteristic extraction module is used for extracting the full-width characteristic data of the banknote image;
and the similarity calculation module is used for performing similarity calculation on the full-width characteristic data of the extracted banknote image and preset sample data, and selecting the banknote image with the highest comprehensive similarity as a final banknote information result.
8. A banknote denomination recognition device, characterized by: the identification system comprises a banknote image acquisition module, a banknote image acquisition module and a banknote identification module, wherein the banknote image acquisition module is used for acquiring an image of a banknote to be identified;
the characteristic extraction module is used for extracting the full-width characteristic data of the banknote image;
the similarity calculation module is used for carrying out similarity calculation on the full-width feature data of the extracted banknote image and preset sample data of partial denomination versions extracted from each currency, and preliminarily screening out the currency of the banknote to be identified; and the similarity calculation module is used for carrying out similarity calculation on the extracted full-width characteristic data of the banknote image and preset sample data corresponding to each denomination version of the currency screened preliminarily, and selecting the banknote with the highest comprehensive similarity as a final banknote information result.
9. The banknote denomination recognition device of claim 6 or 7, wherein: the banknote identification system further comprises a confirmation module, wherein the confirmation module is used for judging whether the similarity score and the difference score meet the requirements or not, if the similarity score and the difference score meet the requirements, the identification is considered to be normal, and if at least one of the similarity score and the difference score does not meet the requirements, the banknote identification system is considered to be a suspected banknote;
the system further comprises a sample file loading module, wherein the sample file loading module is used for correspondingly loading preset sample data according to the set currency.
10. The banknote denomination recognition device of claim 6 or 7, wherein: the method for extracting the full-width characteristic data of the banknote image specifically comprises the following steps: drawing the shape of the bank note, finding the edge characteristics of the graph to obtain four corner point coordinates of the bank note image, giving out preliminary information judgment of the bank note through shape detection, obtaining plane transformation parameters of the bank note according to the coordinates, carrying out linear-like plane transformation on the coordinate information of the bank note image, zooming the picture to a set size, then carrying out discrete cosine transformation on the picture zoomed to the set size, converting the picture to a frequency domain, and obtaining full-width characteristic data;
similarity calculation is carried out on the extracted full-width characteristic data of the banknote image and preset sample data, and the similarity calculation specifically comprises the following steps: the method comprises the steps that the full-width characteristic data and sample data of a banknote image are frequency domain matrixes, the coefficients of a set number of middle and low frequency regions are extracted from the frequency domain matrixes, the arithmetic mean value D is calculated, each value in the frequency domain matrixes is compared with D, the value is larger than D and is 1, the value is smaller than D and is 0, the image information can be obtained through storage according to bits, the number capable of quantifying the difference degree of the images is obtained through calculating the difference between the image information of the banknote image and the image information of the sample data, when the difference is 0, the two images are considered to be identical, and the difference is larger, the more dissimilar the images are indicated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010425302.9A CN111738079A (en) | 2020-05-19 | 2020-05-19 | Banknote denomination recognition method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010425302.9A CN111738079A (en) | 2020-05-19 | 2020-05-19 | Banknote denomination recognition method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111738079A true CN111738079A (en) | 2020-10-02 |
Family
ID=72647379
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010425302.9A Pending CN111738079A (en) | 2020-05-19 | 2020-05-19 | Banknote denomination recognition method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111738079A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112329845A (en) * | 2020-11-03 | 2021-02-05 | 深圳云天励飞技术股份有限公司 | Method and device for replacing paper money, terminal equipment and computer readable storage medium |
| CN112906696A (en) * | 2021-05-06 | 2021-06-04 | 北京惠朗时代科技有限公司 | English image region identification method and device |
| CN114419644A (en) * | 2021-12-30 | 2022-04-29 | 武汉卓目科技有限公司 | Banknote denomination recognition method and system |
| CN117671849A (en) * | 2023-12-14 | 2024-03-08 | 浙江南星科技有限公司 | A vertical image scanning banknote counting machine using a banknote sliding structure |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102034108A (en) * | 2010-12-06 | 2011-04-27 | 哈尔滨工业大学 | Multi-resolution network characteristic registration-based method for sorting face values and face directions of notes in sorter |
| CN102971746A (en) * | 2010-04-08 | 2013-03-13 | 多尔斯研发有限公司 | Method for the classification of banknotes |
| CN106846609A (en) * | 2016-12-16 | 2017-06-13 | 恒银金融科技股份有限公司 | Banknote face value face identification method based on perceptual hash |
| CN107705418A (en) * | 2017-10-10 | 2018-02-16 | 深圳怡化电脑股份有限公司 | A kind of bank note towards recognition methods, device, equipment and readable storage medium storing program for executing |
| CN108198324A (en) * | 2018-02-08 | 2018-06-22 | 中南大学 | A kind of multinational bank note currency type recognition methods based on finger image |
| CN110956737A (en) * | 2020-01-07 | 2020-04-03 | 武汉卓目科技有限公司 | Safety line identification method and device |
-
2020
- 2020-05-19 CN CN202010425302.9A patent/CN111738079A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102971746A (en) * | 2010-04-08 | 2013-03-13 | 多尔斯研发有限公司 | Method for the classification of banknotes |
| CN102034108A (en) * | 2010-12-06 | 2011-04-27 | 哈尔滨工业大学 | Multi-resolution network characteristic registration-based method for sorting face values and face directions of notes in sorter |
| CN106846609A (en) * | 2016-12-16 | 2017-06-13 | 恒银金融科技股份有限公司 | Banknote face value face identification method based on perceptual hash |
| CN107705418A (en) * | 2017-10-10 | 2018-02-16 | 深圳怡化电脑股份有限公司 | A kind of bank note towards recognition methods, device, equipment and readable storage medium storing program for executing |
| CN108198324A (en) * | 2018-02-08 | 2018-06-22 | 中南大学 | A kind of multinational bank note currency type recognition methods based on finger image |
| CN110956737A (en) * | 2020-01-07 | 2020-04-03 | 武汉卓目科技有限公司 | Safety line identification method and device |
Non-Patent Citations (1)
| Title |
|---|
| 任俊丽 等；: "《自适应尺度突变目标跟踪》", 《中国图象图形学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112329845A (en) * | 2020-11-03 | 2021-02-05 | 深圳云天励飞技术股份有限公司 | Method and device for replacing paper money, terminal equipment and computer readable storage medium |
| CN112329845B (en) * | 2020-11-03 | 2024-05-07 | 深圳云天励飞技术股份有限公司 | Method and device for changing paper money, terminal equipment and computer readable storage medium |
| CN112906696A (en) * | 2021-05-06 | 2021-06-04 | 北京惠朗时代科技有限公司 | English image region identification method and device |
| CN114419644A (en) * | 2021-12-30 | 2022-04-29 | 武汉卓目科技有限公司 | Banknote denomination recognition method and system |
| CN117671849A (en) * | 2023-12-14 | 2024-03-08 | 浙江南星科技有限公司 | A vertical image scanning banknote counting machine using a banknote sliding structure |
| CN117671849B (en) * | 2023-12-14 | 2024-05-14 | 浙江南星科技有限公司 | Vertical image scanning banknote counter adopting banknote sliding structure |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108647681B (en) | An English text detection method with text orientation correction | |
| CN111738079A (en) | Banknote denomination recognition method and device | |
| US9135518B2 (en) | Robust and efficient image identification | |
| CN109658584B (en) | Bill information identification method and device | |
| EP2187359B1 (en) | Paper sheet identification device and paper sheet identification method | |
| Guo et al. | A reliable method for paper currency recognition based on LBP | |
| Wang et al. | Shrinking the semantic gap: spatial pooling of local moment invariants for copy-move forgery detection | |
| CN105184225B (en) | A kind of multinational banknote image recognition methods and device | |
| CN109947273B (en) | A kind of point reading positioning method and device | |
| CN113011426A (en) | Method and device for identifying certificate | |
| CN107195069A (en) | A kind of RMB crown word number automatic identifying method | |
| CN112614167A (en) | Rock slice image alignment method combining single-polarization and orthogonal-polarization images | |
| Tahir et al. | MD-LBP: an efficient computational model for protein subcellular localization from HeLa cell lines using SVM | |
| van der Maaten et al. | Computer vision and machine learning for archaeology | |
| CN116778503A (en) | A seal authenticity identification method and system based on OCR recognition | |
| CN114863163B (en) | A method and system for cell classification based on cell images | |
| Chakraborty et al. | Review of various image processing techniques for currency note authentication | |
| CN120953347A (en) | AI-driven method and system for detecting the external dimensions of precision cutting tools | |
| Liu et al. | An improved image retrieval method based on sift algorithm and saliency map | |
| CN117474916B (en) | Image detection method, electronic equipment and storage medium | |
| KR100888674B1 (en) | Method of measuring similarity of bills using frequency domain | |
| CN117314861B (en) | Method for detecting, identifying and aligning silicon wafer overlay pattern | |
| CN119723111A (en) | An image processing method and system based on artificial intelligence | |
| Jiang et al. | Image copy-move forgery detection and localization scheme: How to avoid missed detection and false alarm | |
| CN112233313B (en) | Paper money identification method, device and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201002 |
|
| RJ01 | Rejection of invention patent application after publication |