Image recognition system for currency counting and detecting machine

Abstract

The invention discloses an image recognition system for a currency counting and detecting machine, and relates to the technical field of currency counting and detecting machines, a positioning chip is arranged in the currency counting and detecting machine to realize effective tracking of the position of a counterfeit currency, a high-resolution camera is used for shooting the currency from multiple angles, and based on an optical character recognition technology, the image recognition efficiency is improved. The method comprises the following steps: identifying information of banknotes, carrying out deep analysis on an image by using a deep learning algorithm to accurately judge the authenticity of the banknotes, integrating counterfeit banknote information according to a specific format to obtain a sequence Q, combining the sequence Q with a henon mapping to obtain an initial encryption sequence Z, and converting the Z into image data, thereby effectively reducing the risk of tampering in a data transmission process. An administrator can obtain original information only by inputting a correct secret key, sorts out counterfeit money information collected in a fixed area, finds out an urgent processing area and obtains a counterfeit money occurrence frequency period R. According to the method, the circulation track of counterfeit money is effectively tracked, and appropriate measures are taken.

Description

Technical Field

[0001] The present invention relates to the technical field of currency verification machines, and particularly to an image recognition system for a currency verification machine. Background Art

[0002] With the increase in the currency circulation volume, the problem of counterfeit money has become increasingly serious, bringing great harm to the financial order and social economy. Traditional currency verification machines mainly focus on the identification of the authenticity of banknotes, lacking effective collection, analysis, and processing of counterfeit money circulation information, unable to track the circulation trajectory of counterfeit money in real time, and difficult to accurately prevent and control in high-incidence areas and time periods of counterfeit money. At the same time, in terms of data security, when traditional currency verification machines transmit counterfeit money information, the data is not effectively encrypted, resulting in the risk of being stolen and tampered with during the transmission process of counterfeit money information. This not only cannot guarantee information security but may also lead to wrong information misleading the anti-counterfeiting work. Therefore, it is of great practical significance to develop an image recognition system for a currency verification machine that can solve the above problems. Summary of the Invention

[0003] Aiming at the deficiencies of the prior art, the present invention provides an image recognition system for a currency verification machine, which solves the problems of insecurity in the transmission process of counterfeit money information and the inability to track the circulation trajectory of counterfeit money in real time.

[0004] To achieve the above objectives, the present invention is realized through the following technical solutions: An image recognition system for a currency verification machine, comprising: A positioning module, which effectively tracks the position of counterfeit money by setting a positioning chip inside the currency verification machine and transmits the detailed position information of the counterfeit money to the data analysis module; A data analysis module, which sorts out all counterfeit money information collected in a fixed area, divides the area into several equal-area regions, calculates the total face value of all counterfeit money in these regions, screens out key attention regions from them. For each key attention region, calculates the share of counterfeit money of different denominations and sets the weights corresponding to counterfeit money of different denominations to obtain the comprehensive value of each key attention region, selects key regions according to the comprehensive value, puts them into the urgent processing area, divides the time period of the urgent processing area, analyzes the number of counterfeit money appearing in different periods to obtain the counterfeit money appearance frequency period R, and transmits R to the protection execution module; A protection execution module, which, for the received counterfeit money appearance frequency period R, takes a series of protection measures in the week before every R weeks hereafter. The protection measures specifically include comprehensively maintaining and upgrading the currency verification equipment and training relevant staff on counterfeit money identification.

[0005] As a further solution of the present invention, before the data analysis module, there further includes an image data collection module, a data encryption module, and a data decryption module. The image data collection module uses a high-resolution camera to photograph banknotes from several set angles, recognizes information such as the denomination numbers and words on the banknotes based on optical character recognition technology, and performs in-depth analysis on the images using deep learning algorithms to accurately determine the authenticity of the banknotes. Integrate the denomination, authenticity label, geographical location, specific time, and unique identification ID of the counterfeit banknote in a specific format to obtain sequence Q, and transmit Q to the data encryption module; the data encryption module is used to combine Q with the Henon mapping to obtain the initial encryption sequence Z, convert Z into image data, and transmit the image data to the data decryption module; the data decryption module is used to, after obtaining the image data, convert the image information into the original data information by inputting the correct key and the inverse mapping formula, and transmit the obtained original data information to the data analysis module.

[0006] As a further solution of the present invention, the specific steps for combining Q with the Henon mapping to obtain the initial encryption sequence Z are as follows: According to the obtained pure digital sequence Q, calculate the sequence length of Q to be 14; Use the Henon mapping to generate an encryption key with a length of 14. The specific formula of this mapping is: x(n + 1)=1 - a*(x(n)^2)+y(n), y(n + 1)=b*x(n); Select the x(n) sequence as the encryption key; Introduce the interleaved sequence P={-1,1,-1,1,...,(-1)^t}, where t∈[1,14]; Calculate the final encryption sequence through the formula Z = Q + P*x(n).

[0007] As a further solution of the present invention, the specific steps for converting Z into image data are as follows: For each number Ei in the encryption sequence Z, i∈[1,14], convert it into an RGB color value according to a specific formula. The specific formulas for calculating the r, g, and b values are as follows: (Ei)mod256 = ri...gi, bi=(31*ri + gi)mod256, where mod() represents the remainder function; It is collected that the encryption sequence Z has 14 elements. To ensure that the area of each color region is the same, select a rectangular image with a width of 1400 pixels and a height of 100 pixels, that is, divide the image horizontally into 14 regions with a width of 100 pixels and a height of 100 pixels evenly; Prepare a blank image. Fill each area with the color corresponding to the number in the encryption sequence Z in order. Starting from the left side of the image, fill the i-th area with the color (ri, gi, bi) corresponding to the number Ei, and so on until all 14 areas are filled. After filling, save the image in PNG format to ensure lossless image information.

[0008] As a further solution of the present invention, the specific steps to convert the image into the original information are as follows: Open the saved PNG image. Starting from the left side of the image, identify the color of each 100x100 pixel area in order. Select the pixel at the center position of each area and obtain its RGB color value. If the color value of the central pixel of the area is (ri, gi, bi), calculate it through the inverse mapping formula from color to number. The inverse mapping formula is: numberi = ri × 256 + gi. Arrange the obtained numbers in order to get the original encryption sequence Num = {number1, number2,..., number14}. According to the obtained key x(n), obtain the original data sequence Q according to the inverse transformation formula Q = Num - P * x(n).

[0009] As a further solution of the present invention, the specific method to find the urgent processing area is as follows: Sort out all the counterfeit money information collected in a fixed area and divide the area into m regions. Calculate the total value Ni of all counterfeit money in each region, i ∈ [1, m], that is, add up the denominations of each counterfeit money in the region. Set a total value threshold Nth. If Ni > Nth in a certain region, divide this region into the key attention regions Gj, j ∈ [1, p], where p is the number of key attention regions. In the key attention regions, calculate the share Qjt of each denomination of counterfeit money in each region and assign corresponding weights Wjt to each denomination of counterfeit money, where t is the denomination of the counterfeit money. Calculate the proportion of each denomination in each region according to the formula Hj = sum(QjtWjt). Normalize each region according to the total value Nj of each region in the key region Gj, that is, Nj / sum(Nj). Calculate the comprehensive value of each region in the key attention region Gj according to the formula Vj = Hj * (Nj / sum(Nj)). If Vj > Vth, then determine this region as the urgent processing area.

[0010] As a further solution of the present invention, the specific steps of obtaining the counterfeit banknote occurrence frequency period R are as follows: For the expedited processing area, the statistical time period is divided into K equal time periods; For each region, count the number of counterfeit banknotes of different denominations in each time period, where i represents the denomination, j represents the region, and k represents the time period. Calculate the total frequency of counterfeit banknotes in the area in each time period Fjk=sum(Aijk); According to the formula Calculate the autocorrelation function of the frequency of counterfeit banknotes in each region, where is the average frequency of counterfeit banknotes in the area, l is the delay order, and K is the total number of time periods; The frequency period R of counterfeit banknotes is obtained by calculating the autocorrelation function.

[0011] As a further solution of the present invention, a threshold Tj is set for each focus area. , it indicates that there is a significant fluctuation in the frequency of counterfeit banknotes in this area during this time period.

[0012] The present invention provides an image recognition system for a banknote counting and checking machine. Compared with the prior art, the system has the following beneficial effects: (1) The present invention effectively tracks the geographical location and time of the appearance of counterfeit banknotes by embedding a positioning chip in the banknote counting and checking machine, and encrypts the detailed information about the counterfeit banknotes, thereby effectively reducing the risk of the counterfeit banknote information being stolen or tampered with during the transmission process, and ensuring the security of information transmission; (2) The present invention divides a region into multiple areas to find out the expedited processing area, and calculates the autocorrelation function in the expedited processing area to find out the frequency cycle of counterfeit banknotes in the area, thereby preparing for effective protection based on the cycle later. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a block diagram of the system principle of the present invention. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0015] like Figure 1 The present invention provides an image recognition system for a currency counting and verifying machine, comprising: Positioning module. This module is mainly used to accurately locate the currency counter. By installing a positioning chip inside the currency counter, the effective tracking of the position of counterfeit banknotes can be achieved; The built-in chip can specifically consider GPS chips, Beidou chips, cellular network positioning chips, and Wi-Fi positioning chips; GPS chips are suitable for open space scenarios and areas without obstructions. For example, in outdoor self-service banking halls and outdoor financial publicity activity venues of banks, at least four satellite signals can be received through GPS chips, and the position can be determined using the trilateration method. For example, in a bank branch in the suburbs of a city, if a GPS chip is installed inside the currency counter, the longitude and latitude can be accurately obtained, with an error within a few meters, providing a basis for positioning counterfeit banknotes; Beidou chips are suitable for currency counters with additional communication requirements. For example, in some financial service points in remote areas, in addition to the function of positioning counterfeit banknotes, in case of emergencies such as the concentrated appearance of counterfeit banknotes or equipment failures, communication with the superior agency can be achieved through the Beidou short message function. The Beidou chip locates by receiving satellite signals and has excellent positioning accuracy, especially in the Asia-Pacific region; Cellular network positioning chips are suitable for indoor places or places with weak satellite signals. For example, bank branches in underground shopping malls. In this case, satellite signals usually cannot penetrate buildings, and the purpose of positioning cannot be achieved by receiving satellite signals. However, cellular network positioning chips estimate the position by measuring signal parameters with base stations, avoiding the influence of large buildings on satellite signals. For example, for self-service currency counting equipment in a bank within a large shopping mall, although the positioning accuracy is within dozens of meters using base station positioning, the approximate position where counterfeit banknotes appear can be determined, and the search range can be narrowed down in combination with other information; Wi-Fi positioning chips are suitable for places with stable Wi-Fi coverage indoors, such as the business hall of a bank. The Wi-Fi network is stable. The Wi-Fi positioning chip locates by scanning Wi-Fi hotspots and comparing with the fingerprint database. For example, in the office area of the bank headquarters, the currency counter can be accurately located within a few meters using Wi-Fi, and the position where counterfeit banknotes appear in the office area can be accurately recorded.

[0016] Image data collection module. This module is mainly used to collect detailed information of counterfeit banknotes. When banknotes enter the currency counter, high-resolution cameras are used to take pictures of the banknotes from multiple angles, covering the front, back, and edge details of the banknotes to ensure that the obtained image information is complete and clear; For the banknote image information obtained in real time, based on optical character recognition technology, information such as the denomination numbers and texts on the banknotes is recognized. At the same time, deep learning algorithms are used to deeply analyze the images. These algorithms are trained with a large number of real and fake banknote images and can accurately identify the authenticity of banknotes according to the image features of the input banknotes. For example, by learning multi-dimensional features such as the watermark clarity of real and fake banknotes, the magnetic characteristics of security threads, and the printing quality of patterns, the trained deep learning model can accurately judge the authenticity of banknotes; The deep learning algorithms described above include convolutional neural networks. This algorithm has a powerful ability to extract local features and can automatically learn features such as banknote textures and patterns. Its hierarchical structure can not only efficiently process images but also has a high training efficiency; Recurrent neural networks and their variants, long short-term memory networks, and gated recurrent units. This algorithm has good effects in processing sequential data, can capture the order and context information of characters and patterns on banknotes, and can judge the authenticity of banknotes according to the specific arrangement rules of banknotes to improve the reliability of recognition; Generative adversarial networks. This algorithm has good effects in data augmentation, can generate diverse simulated banknote images, increases the diversity of training data, enables the model to adapt to banknotes of different qualities and styles, and enhances the generalization ability of the model; Residual networks. This algorithm can train deeper networks, learn more complex and subtle features in banknotes, and improve the accuracy of recognition by accurately identifying the imperceptible differences in counterfeit banknotes. Especially for highly simulated counterfeit banknotes, its identification effect is remarkable; If the output result obtained through the trained discriminant model is a counterfeit banknote, start collecting the detailed information of the counterfeit banknote and combining it in a specific format; The denomination information is directly taken from the optical character recognition result. For example, if "100" is recognized, the denomination is determined to be 100 yuan; the authenticity label is represented by "F", and the number 6 in A-Z can be used; the geographical location information is obtained by the positioning module. If the currency detector is equipped with a Beidou positioning chip, accurate longitude and latitude coordinates can be obtained, such as "30.67°N, 104.06°E"; the specific time is obtained through the high-precision clock built into the currency detector, accurate to the second, for example, "2024-11-10 14:25:30"; the unique identification ID is composed of the device number of the currency detector and the serial number of this banknote verification. Assuming the device number is "5" and the serial number is "123", the unique identification ID is "5-123"; Integrate the collected counterfeit banknote information in a specific format to form a complete data record, such as "100-F-30.67N-104.06E-2024-11-10-14-25-30-5-123", and convert it into a pure digital sequence Q = {100, 6, 30, 67, 104, 6, 2024, 11, 10, 14, 25, 30, 5, 123}. This set of pure digital sequences will be transmitted to the data encryption module for encryption to prevent the leakage of counterfeit banknote information.

[0017] The data encryption module calculates the sequence length of Q to be 14 based on the obtained pure digital sequence Q, and uses the Henon map to generate an encryption key with a length of 14. The Henon map is a chaotic map, and the generated sequence has strong chaos. Due to the characteristics of chaotic sequences such as sensitivity to initial conditions and long-term unpredictability, the generated encryption key has a high degree of randomness and security. The specific formula of this map is: x(n + 1) = 1 - a * (x(n)^2) + y(n), y(n + 1) = b * x(n). To obtain a more dispersed and disordered sequence, a is usually taken as 1.38 and b is usually taken as 0.2. Finally, two chaotic sequences x(n) and y(n) are obtained. Considering the three aspects of chaotic characteristics, key space, and computational complexity, the x(n) sequence is selected as the encryption key; During the encryption process, an interleaved sequence P = {-1, 1, -1, 1,..., (-1)^t} is introduced, where t ∈ [1, 14]. The final encrypted sequence is calculated by the formula Z = Q + P * x(n). Here, the numbers corresponding to P and x(n) are multiplied, and the numbers corresponding to Q and P * x(n) are added; The specific steps to convert the encrypted sequence Z into an image are as follows: For each number Ei in the sequence, i ∈ [1, 14], it is converted into an RGB color value according to a specific formula. The specific formulas for calculating the r, g, and b values are as follows: (Ei) mod 256 = ri... gi, bi = (31 * ri + gi) mod 256; Determine the size of the image. Considering that the sequence Z has 14 elements, to ensure that the area of each color region is the same, a rectangular image with a width of 1400 pixels and a height of 100 pixels can be selected, that is, the image is evenly divided horizontally into 14 regions with a width of 100 pixels and a height of 100 pixels; After preparing the blank image, in the order of the numbers in the sequence Z, fill the color corresponding to each number into the corresponding region in turn. Starting from the left side of the image, the i-th region is filled with the color (ri, gi, bi) corresponding to the number Ei, and so on until all 14 regions are filled. After filling, save the image in PNG format to ensure the lossless of image information.

[0018] Data decryption module. After transmitting the encrypted image information to the administrator, the administrator needs to input the correct key to convert the image information into the original data information. Open the saved PNG image. Starting from the left side of the image, identify the colors of each 100x100 pixel area in sequence. For example, select the pixel at the center of each area and obtain its RGB color value. Assume the color value of the center pixel of the first area is (ri, gi, bi). Calculate through the inverse mapping formula from color to number. The inverse mapping formula is: numberi = ri×256 + gi. Arrange the obtained numbers in sequence to get the original encrypted sequence Num = {number1, number2,..., number14}. According to the obtained key x(n), obtain the original data sequence Q according to the inverse transformation formula Q = Num - P*x(n).

[0019] Data analysis module. Organize all the counterfeit currency information collected in a fixed area and divide the area into m regions. Calculate the total value Ni of all counterfeit currency in each region, i ∈ [1, m], that is, add up the denominations of each counterfeit currency in the region. Set a total value threshold Nth. If Ni > Nth in a certain region, divide this region into the key attention regions Gj, j ∈ [1, p], where p is the number of key attention regions. In the key attention regions, calculate the share Qjt of counterfeit currency of each denomination in each region, where t is the denomination of the counterfeit currency. For example, in a certain key attention region, there are 10 counterfeit currency with a denomination of 100 yuan, 5 counterfeit currency with a denomination of 50 yuan, and 3 counterfeit currency with a denomination of 20 yuan. The total number of counterfeit currency is 18. Then the share of counterfeit currency with a denomination of 100 yuan is Qj100 = 10 / 18, the share of counterfeit currency with a denomination of 50 yuan is Qj50 = 5 / 18, and the share of counterfeit currency with a denomination of 20 yuan is Qj20 = 3 / 18. Assign corresponding weights Wjt to each denomination of counterfeit currency. The weights can be determined through expert evaluation combined with historical data statistical analysis. For example, according to historical data, counterfeit currency with a denomination of 100 yuan has a greater impact on the financial order, so the weight Wj100 = 0.5; the weight of counterfeit currency with a denomination of 50 yuan is Wj50 = 0.3; the weight of counterfeit currency with a denomination of 20 yuan is Wj20 = 0.2. Thus, obtain the comprehensive situation of the proportion of each denomination in each region: sum(QjtWjt). Then, according to the total value Nj of each region in these key regions Gj, and perform normalization processing on each region, that is, Nj / sum(Nj). Calculate the comprehensive value Vj of each region in the key attention regions Gj = sum(QjtWjt)*(Nj / sum(Nj)). Sort these key attention regions according to the comprehensive value. The regions ranked higher not only have a larger total value but also have a larger proportion of counterfeit currency with larger denominations, such as counterfeit currency with a denomination of 100 yuan or 50 yuan. These regions are determined as the urgent processing regions. For the urgent processing area, the statistical time period is divided into K equal time cycles. The division of the time cycles can be selected according to the actual situation, such as by week, month, quarter, etc. If it is in weeks, and if the statistical time period is 3 months, about 13 weeks, then it can be divided into 13 time cycles; for each area, count the number of occurrences Aijk of counterfeit banknotes of different denominations within each time cycle, where i represents the denomination type, j represents the area, and k represents the time cycle, and calculate the total occurrence frequency Fjk = sum(Aijk) of counterfeit banknotes in this area within each time cycle; calculate the autocorrelation function of the occurrence frequency of counterfeit banknotes in each area. The specific formula is as follows: , where is the average value of the occurrence frequency of counterfeit banknotes in this area, l is the order of delay, and K is the total number of time cycles; obtain the period R of the occurrence frequency of counterfeit banknotes by calculating the autocorrelation function.

[0020] The protection execution module, after obtaining the period R of the occurrence frequency of counterfeit banknotes in the urgent processing area, takes a series of protection measures in the week before every R weeks in the future. For example, financial institutions conduct comprehensive maintenance and upgrading of the banknote verification equipment to ensure that the equipment can operate normally during the high-incidence period of counterfeit banknotes; conduct training on counterfeit banknote identification for relevant staff to improve the identification ability; to analyze whether there are obvious fluctuations in a specific time period, set a threshold Tj for each key area of concern. The setting of the threshold can refer to factors such as the standard deviation of the historical data of the occurrence frequency of counterfeit banknotes in this area. If , it indicates that there is an obvious fluctuation in the occurrence frequency of counterfeit banknotes in this area during this time cycle. At this time, it is necessary to strengthen the review of cash transactions, especially the review of counterfeit banknotes with larger denominations, such as 100 yuan and 50 yuan, increase the manual review link, and use more advanced banknote verification equipment for secondary inspection, etc., to effectively prevent the circulation of counterfeit banknotes.

[0021] Some of the data in the above formula are numerically calculated after removing their dimensions, and the content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0022] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An image recognition system for a banknote counting and verifying machine, characterized in that: include: The positioning module, which effectively tracks the location of counterfeit banknotes by installing a positioning chip inside the banknote counting and checking machine, and transmits the detailed location information of the counterfeit banknotes to the data analysis module; The data analysis module organizes all the counterfeit banknote information collected in a fixed area, divides the area into several areas of equal area, calculates the total face value of all counterfeit banknotes in these areas, and selects key focus areas from them. For each key focus area, calculates the share of counterfeit banknotes of different denominations and sets the corresponding weights of counterfeit banknotes of different denominations to obtain the comprehensive value of each key focus area. The key areas are selected according to the comprehensive value and placed in the expedited processing area. The expedited processing area is divided into time periods, and the number of counterfeit banknotes appearing in different periods is analyzed to obtain the counterfeit banknote frequency period R, and R is transmitted to the protection execution module; A protection execution module is used to perform a series of protection measures for the counterfeit banknotes received with a frequency period R in the week before every R weeks. The protection measures specifically include comprehensive maintenance and upgrading of banknote detection equipment and counterfeit banknote identification training for relevant staff.

2. The image recognition system for a banknote counting and checking machine according to claim 1, characterized in that: Before the data analysis module, it also includes an image data collection module, a data encryption module, and a data decryption module. The image data collection module uses a high-resolution camera to shoot banknotes from several set angles, and based on optical character recognition technology, identifies the denomination numbers, text and other information on the banknotes, and uses a deep learning algorithm to perform in-depth analysis of the image to accurately determine the authenticity of the banknotes. The denomination, authenticity label, geographic location, specific time, and unique ID of the counterfeit banknotes are integrated in a specific format to obtain a sequence Q, and Q is transmitted to the data encryption module; the data encryption module is used to combine Q and henon mapping to obtain an initial encryption sequence Z, convert Z into image data, and transmit the image data to the data decryption module; the data decryption module is used to convert the image information into original data information by inputting the correct key and inverse mapping formula after obtaining the image data, and transmit the obtained original data information to the data analysis module.

3. The image recognition system for a banknote counting and checking machine according to claim 2, characterized in that: The specific steps of combining Q and henon mapping to obtain the initial encryption sequence Z are: According to the obtained pure digital sequence Q, the sequence length of Q is calculated to be 14; Use Henon mapping to generate an encryption key of length 14. The specific formula of the mapping is: x(n+1)=1-a*(x(n)^2)+y(n), y(n+1)=b*x(n); Select the x(n) sequence as the encryption key; Introduce the interleaved sequence P = {-1, 1, -1, 1, ..., (-1)^t}, where t∈[1, 14]; The final encryption sequence is calculated using the formula Z=Q+P*x(n).

4. The image recognition system for a banknote counting and checking machine according to claim 3, characterized in that: The specific steps to convert Z into image data are: For each number Ei, i∈[1,14] in the encrypted sequence Z, it is converted into an RGB color value according to a specific formula. The specific formula for calculating the r, g, b values ​​is as follows: (Ei)mod256=ri...gi, bi=(31*ri+gi)mod256, where mod() represents the remainder function; The collected encrypted sequence Z has 14 elements. To ensure that the area of ​​each color region is the same, a rectangular image with a width of 1400 pixels and a height of 100 pixels is selected, that is, the image is evenly divided into 14 regions with a width of 100 pixels and a height of 100 pixels; Prepare a blank image and fill the corresponding area with the color corresponding to each number in the order of the numbers in the encryption sequence Z. Starting from the left side of the image, the i-th area is filled with the color corresponding to the number Ei (ri, gi, bi), and so on, until all 14 areas are filled. After filling, save the image in PNG format to ensure that the image information is lossless.

5. The image recognition system for a banknote counting and checking machine according to claim 2, characterized in that: The specific steps to convert the image into original information are: Open the saved PNG image and identify the color of each 100x100 pixel area in sequence, starting from the left side of the image; Select the pixel point at the center of each region and obtain its RGB color value. If the color value of the central pixel of the region is (ri, gi, bi), calculate it using the inverse mapping formula from color to number: numberi=ri×256+gi; Arrange the obtained numbers in order to obtain the original encrypted sequence Num={number1,number2,...,number14}; According to the obtained key x(n), the original data sequence Q is obtained according to the inverse transformation formula Q=Num-P*x(n).

6. The image recognition system for a banknote counting and checking machine according to claim 1, characterized in that: The specific method to find the expedited processing area is: All the collected counterfeit money information in a fixed area is sorted out and the area is divided into m areas; Calculate the total value of all counterfeit banknotes in each region Ni, i∈[1,m], that is, add up the denominations of each counterfeit banknote in the region; Set a total value threshold Nth. If Ni>Nth in a certain area, divide the area into the key focus area Gj, j∈[1,p], p is the number of key focus areas; In the key focus areas, calculate the share of counterfeit banknotes of each denomination Qjt in each area and assign corresponding weights Wjt to each denomination of counterfeit banknotes, where t is the denomination of the counterfeit banknotes; Calculate the proportion of each denomination in each region according to the formula Hj=sum(QjtWjt); According to the total value Nj of each area in the key area Gj, each area is normalized, that is, Nj / sum(Nj); The comprehensive value of each area in the focus area Gj is calculated according to the formula Vj=Hj*(Nj / sum(Nj)); If Vj>Vth, the area is determined as an expedited processing area.

7. The image recognition system for a banknote counting and checking machine according to claim 6, characterized in that: The specific steps to obtain the counterfeit banknote frequency period R are: For the expedited processing area, the statistical time period is divided into K equal time periods; For each region, count the number of counterfeit banknotes of different denominations in each time period, where i represents the denomination, j represents the region, and k represents the time period. Calculate the total frequency of counterfeit banknotes in the area in each time period Fjk=sum(Aijk); According to the formula Calculate the autocorrelation function of the frequency of counterfeit banknotes in each region, where is the average frequency of counterfeit banknotes in the area, l is the delay order, and K is the total number of time periods; The frequency period R of counterfeit banknotes is obtained by calculating the autocorrelation function.

8. The image recognition system for a banknote counting and checking machine according to claim 7, characterized in that: A threshold Tj is set for each focus area. , it indicates that there is a significant fluctuation in the frequency of counterfeit banknotes in this area during this time period.

Images