CN106683263B

CN106683263B - The defect management method and system of valuable bills

Info

Publication number: CN106683263B
Application number: CN201710014778.1A
Authority: CN
Inventors: 眭俊华; 刘李泉; 王建鑫; 张健; 卢继兵; 宁焕成; 秦庆旺; 冯礼; 毛林; 王皓; 陈勇; 魏君; 孙晓刚; 张超
Original assignee: China Banknote Printing and Minting Corp; Institute of Printing Science and Technology Peoples Bank of China
Current assignee: China Banknote Printing Technology Research Institute Co ltd; China Banknote Printing and Minting Group Co Ltd
Priority date: 2017-01-10
Filing date: 2017-01-10
Publication date: 2019-07-19
Anticipated expiration: 2037-01-10
Also published as: CN109767430B; CN109767430A; CN109767544B; CN109767544A; CN109754395A; CN109754395B; CN109767545A; CN109767545B; CN109767546B; CN106683263A; CN109767546A

Abstract

The invention proposes a kind of defect management method of valuable bills and systems, wherein, the defect management method of valuable bills includes: the template image that the residual point of those suspected defects is carried by generating according to the sample image of valuable bills, according to the position of the position of the mass center of the geometric figure of the residual point composition of multiple those suspected defects in the template image being calculated and geometric center, judge whether the residual point of any those suspected defects in the residual point of multiple those suspected defects is the residual point of defect.According to the technical solution of the present invention, it is possible to reduce the verification efficiency of secondary checking system also can be improved in number of samples, calculation amount and the time of feature extraction, improve the accuracy rate verified, it reduces and verifies cost, improve nicety of grading, improve the robustness of quality detecting system.

Description

The defect management method and system of valuable bills

Technical field

The present invention relates to valuable bills technical field, a kind of defect management method in particular to valuable bills and A kind of fault management system of valuable bills.

Background technique

Currently, being all directly to be executed using cleaning-sorting machine for the quality verification of valuable bills.Although the verification of cleaning-sorting machine Speed quickly, but tends to cause erroneous judgement, and e.g., sample is very few to cause part that cannot lead to the discrepant normal picture of sample Detection is crossed, causes to report by mistake, sample excessively causes the range allowed wide, causes to fail to report, and originally qualified valuable bills are determined It is unqualified.In the related art, in order to avoid waste, underproof valuable bills is determined as cleaning-sorting machine, need to use Artificial mode carries out secondary verification, not only verifies low efficiency, while being also required to expend more human and material resources.

In addition, accurate defect management is conducive to raising detectability, while the precise classification of defect can be preceding process Important feedback information is provided, help promotes printing quality.Common classification method includes: decision tree, neural network, simple shellfish Ye Si, SVM (Support Vector Machine, support vector machines) etc..All there are respective advantage and disadvantage in practical applications, Such as:

One, Decision-Tree Method, advantage:

1) construction of decision tree does not need any domain knowledge or parameter setting, therefore is suitable for detection type Knowledge Discovery；

2) decision tree can handle high dimensional data, and processing speed is relatively fast；

3) learning procedure of Decision Tree Inductive is simple and quick；

Disadvantage:

1) classification robustness is not strong；

2) when classification is too many, mistake is possible will be increased than very fast；

Two, neural network, advantage:

1) algorithm robust, noise resistance data have the ability analyzed unbred data；

2) a variety of data modes such as discrete, continuous, vector be can handle；

3) the intrinsic concurrency of algorithm is suitable for parallel computation and accelerates calculating process；

Disadvantage:

1) net training time is long；

2) network model lacks interpretation, the information indigestion for including in hidden layer and weight；

3) since the excitation function of sigmoid type all has saturation region, network training is easy to produce paralysis phenomenon；

Three, SVM uses Nonlinear Mapping, and initial data is mapped to the space of more higher-dimension, is then looked in higher dimensional space again Separated to a hyperplane initial data it is best,

Its advantage:

1) classification strong robustness；

2) there is very strong extensive and study energy；

3) dimension space of traditional algorithm and overfitting can be overcome to ask very well.

Disadvantage: when data volume is excessive, the training time is long.

Classify in valuable bills printed matter defect management accurately and the calculating time is two key factors considered.Due to It is more to print product, classification processing those suspected defects is needed to be consequently increased, especially in the case where continuous useless, the classification processing time is just It is most important.

Therefore, number of samples, calculation amount and the time of feature extraction how are reduced, the secondary verification of multinode is being designed After system, how to ensure that secondary checking system can more reasonably distribute task, to improve the verification of secondary checking system Efficiency improves the accuracy rate of verification, reduces and verifies cost, in the design of defect management algorithm, how how to improve nicety of grading Calculating speed is dropped into acceptable range, so that improving the robustness of quality detecting system becomes skill urgently to be resolved at present Art problem.

Summary of the invention

The present invention is based on the above problems, proposes a kind of comprehensive judgement technology of new valuable bills, it is possible to reduce Number of samples, calculation amount and the time of feature extraction also can be improved the verification efficiency of secondary checking system, improves the standard of verification True rate reduces and verifies cost, improves nicety of grading, improves the robustness of quality detecting system.

In view of this, the first aspect of the present invention proposes a kind of defect management method of valuable bills, comprising: acquisition has The sample image of valence bill, and the template image for carrying the residual point of those suspected defects is generated according to the sample image；Described in calculating The position of the mass center of the geometric figure of the residual point composition of multiple those suspected defects in template image and the position of geometric center；According to institute The position of mass center and the position of the geometric center are stated, judges the residual point of any those suspected defects in the residual point of the multiple those suspected defects It whether is the residual point of defect.

In the technical scheme, according to the mass center of the geometric figure of the residual point composition of multiple those suspected defects in template image Position and the position of geometric center judge whether the residual point of any those suspected defects in the residual point of multiple those suspected defects is real defect Residual point, wherein can use LLE algorithm (Locally Linear Embedding, Local Liner Prediction) for valuable ticket According to sample image generate template image, therefore, through the above technical solutions, can to avoid in the related technology pass through height template Matching method or similitude detection method determine whether the residual point of those suspected defects is the veritably residual point of defect, will be doubtful to avoid the occurrence of The case where residual point of defect is judged by accident, effectively improves the detection accuracy of the residual point of defect, and then improve print quality inspection system The robustness of system.

In the above-mentioned technical solutions, it is preferable that calculate the several of the residual point composition of multiple those suspected defects in the template image The position of the mass center of what figure and the position of geometric center, with according to the position of the position of the mass center and the geometric center, The step of whether residual point of any those suspected defects for judging in the residual point of the multiple those suspected defects is defect residual includes: described in calculating The position of first mass center of the geometric figure of multiple residual point compositions of those suspected defects and the position of the first geometric center, and calculate and remove institute State the position and second of the second mass center of the geometric figure of the residual point composition of other those suspected defects except the residual point of any those suspected defects The position of geometric center；Judge whether the distance between the position of first mass center and the position of second mass center are greater than the One threshold value, and whether the distance between the position of first geometric center and the position of second geometric center are less than second Threshold value；If judging result be it is no, any residual point of those suspected defects be the residual point of the defect, otherwise, it is described it is any it is doubtful lack Falling into residual point is not the residual point of the defect.

In the technical scheme, by by multiple those suspected defects it is residual point composition geometric figure the first mass center position and Except the position of the second mass center of the geometric figure of the residual point composition of any those suspected defects is compared, and multiple those suspected defects are residual It puts the position of the first geometric center of the geometric figure of composition and puts the second of the geometric figure formed except any those suspected defects are residual The position of geometric center is compared, so as to determine whether the residual point of any those suspected defects is veritably scarce according to comparison result Residual point is fallen into, specifically, when the distance between the position of the first mass center and the position of the second mass center are greater than first threshold, and more than the first When the distance between the position and the position of the second geometric center at what center are less than second threshold, that is to say, the position of bright second mass center Changing greatly for the position generation compared to the first mass center is set, and the position of the second geometric center is compared to the first geometric center The variation that position occurs is smaller, it is determined that any residual point of those suspected defects has interference, i.e., any residual point of those suspected defects is not real The residual point of defect can exclude any residual point of defect, and otherwise, in addition to the conditions already mentioned, other situations are believed that those suspected defects are residual Point is the veritably residual point of defect, in this way, can relatively accurately judge whether the residual point of any those suspected defects is that real defect is residual Point.

In any of the above-described technical solution, it is preferable that described according to sample image generation, to carry those suspected defects residual It the step of template image of point, specifically includes: weight square is calculated according to the point of proximity of each sample point in the sample image Battle array；The template image for carrying the residual point of the those suspected defects is generated according to the weight matrix.

In the technical scheme, by calculating the weight matrix between each sample point and the point of proximity of the sample point, example Such as, the point of proximity of each sample point can be found, by the method for measuring Euclidean distance so as to raw according to the weight matrix At the template image for carrying the residual point of those suspected defects.

In any of the above-described technical solution, it is preferable that each sample point according in the sample image closes on Point calculates the step of weight matrix, specifically includes: being calculated by the following formula the weight matrix:

Wherein, ε (w) indicates error amount, X_iIndicate any sample point, X_j(j=1,2 ..., k) indicates any sample point K point of proximity, w_ijIndicate the weight matrix between any sample point and the point of proximity.

In the technical scheme, weight matrix is calculated by above-mentioned formula, wherein when the value minimum of ε (w) The value of weight matrix is calculated, in addition, k is previously given value.

In any of the above-described technical solution, it is preferable that every a line of the weight matrix and be 1.

In the technical scheme, every a line of weight matrix and be 1, i.e. ∑_jw_ij=1, in this way, by making weight matrix Meet above-mentioned constraint condition, it can be ensured that the validity of weight matrix.

The second aspect of the present invention proposes a kind of defect management method of valuable bills, comprising: obtains valuable bills Sample image；Fourier transform is carried out to the sample image, with the magnitude image and phase image of the determination sample image； The corresponding amplitude of the magnitude image also corresponding phase X of original image and the phase image is obtained to local derviation also original image With phase Y-direction local derviation also original image；According to the amplitude also original image, the phase X to local derviation also original image and the phase Y Original image is gone back to local derviation also original image generation；The sample image and the original image of going back are subjected to difference, according to difference result Determine the residual image of the valuable bills.

In the technical scheme, pass through the corresponding amplitude of the magnitude image also original image of the sample image of acquisition valuable bills Phase X corresponding with phase image is to local derviation also original image and phase Y-direction local derviation also original image, wherein the X of X indicates coordinate axis To, the Y-direction of Y indicates coordinate axis, and by amplitude also original image, phase X to local derviation also original image and phase Y-direction local derviation also original image What is generated goes back original image and sample image progress difference, and the residual image of valuable bills, Ke Yiyou are determined according to difference result Improve the robustness of printing quality checking system in effect ground.

In the above-mentioned technical solutions, it is preferable that described to obtain the corresponding amplitude of the magnitude image also original image, Yi Jisuo State the corresponding phase X of phase image to local derviation also original image and phase Y-direction local derviation also original image the step of, specifically include: by institute The phase of any image in phase image is stated as basic phase, and according to the basic phase to removing in the phase image The phase of other images except any image is handled, to obtain phase diagram aberration；According to the phase diagram aberration Determine the amplitude also original image, the phase X to local derviation also original image and the phase Y-direction local derviation also original image.

In the technical scheme, by using the phase of any image in phase image as basic phase, for example, by phase The phase of piece image in bit image is used as basic phase, and according to basic phase to other images in phase image Phase is handled, to obtain phase diagram aberration, so as to be calculated by carrying out local derviation to phase diagram aberration to compare accurately Ground determines phase X to local derviation also original image and phase Y-direction local derviation also original image.

In any of the above-described technical solution, it is preferable that it is described according to the amplitude also original image, the phase X to local derviation Also original image and the phase Y-direction local derviation also original image generation the step of going back original image, specifically include: according to the phase X to Local derviation also original image or the phase Y-direction local derviation also original image and initial point, determine unwrapped phase also original image；According to described Unwrapped phase also original image and the basic phase determine phase also original image；To the phase also original image and the amplitude Also original image carries out inverse fourier transform, to go back original image described in determination.

In the technical scheme, by by phase X to local derviation also original image (or phase Y-direction local derviation also original image) and width Degree goes back original image and is inversely generated as going back original image, be utilized " when the geometric displacement of image changes, width in the frequency spectrum of image Degree spectrum remain unchanged, only there is linear deflection in phase spectrum " principle, can effectively illustrate the defect of valuable bills of the invention Management method has stronger deformation tolerance.

In any of the above-described technical solution, it is preferable that described to determine the amplitude also original image according to the phase diagram aberration As, the phase X to local derviation also original image and the phase Y-direction local derviation also original image the step of, specifically include: to the phase Image difference carries out local derviation calculating to Y-direction in X respectively, with obtain the X of the phase diagram aberration to partial derivative and the phase The partial derivative of the Y-direction of image difference；To the magnitude image, the X to partial derivative, the Y-direction partial derivative carry out analysis and Reduction determines the amplitude also original image, the phase X to local derviation also original image and the phase according to analysis and reduction result Y-direction local derviation also original image.

In the technical scheme, by phase diagram aberration X to and Y-direction carry out local derviation calculating respectively, can be than calibrated Really according to X to partial derivative and the partial derivative of Y-direction obtain phase X to local derviation also original image and phase Y-direction local derviation also original image, So as to relatively accurately be obtained according to phase X to local derviation also original image, phase Y-direction local derviation also original image and amplitude also original image It gets and goes back original image.

In any of the above-described technical solution, it is preferable that it is described according to the basic phase to removing institute in the phase image The phase for stating other images except any image is handled, and the step of to obtain phase diagram aberration, is specifically included: will be described The phase of other images and the basic phase carry out subtraction, and carry out phase unwrapping around processing to calculated result, according to Phase unwrapping determines the phase diagram aberration around processing result.

In the technical scheme, by subtracting each other the phase of other images and basic phase and carrying out phase unwrapping around place Reason, available phase diagram aberration, so as to be calculated by carrying out local derviation to phase diagram aberration relatively accurately to determine phase Position X is to local derviation also original image and phase Y-direction local derviation also original image.

Third aspect present invention proposes a kind of defect management method of valuable bills, comprising: according to valuable bills training The defect characteristic of each valuable bills is concentrated, first order child node is generated；The partition value of each first order child node is calculated, And classified according to the partition value to the defect characteristic；Judge whether all first order child nodes can not divide again Class, when the judgment result is yes, classification are completed.

Wherein it is preferred to which the step of being classified according to the partition value to the defect characteristic, specifically includes: if institute Partition value is stated less than given threshold, then is classified using support vector machine method to the defect characteristic；If the partition value More than or equal to the given threshold, then classified using traditional decision-tree to the defect characteristic.

In the technical scheme, it is compared, is determined with the threshold values of setting by the partition value of each first order child node The classification carried out using vector machine method and traditional decision-tree, further judges whether all nodes can not classify again, determines and divides Class is completed, and is classified by using vector machine method and traditional decision-tree to defect characteristic, can effectively be improved classification Precision, calculating speed also fall below acceptable range, improve efficiency of algorithm.

In any of the above-described technical solution, it is preferable that according to lacking for valuable bills each in the training set of valuable bills The step of falling into feature, generating first order child node, specifically includes: calculating each valuable bills in the valuable bills training set Information gain value between defect characteristic and the defect characteristic of other valuable bills；According to the corresponding letter of each valuable bills It ceases yield value and constructs the first order child node.

In the technical scheme, have by the defect characteristic for calculating each valuable bills in valuable bills training set with other Information gain value between the defect characteristic of valence bill can effectively avoid the extraction of false defect, and then to the residual point of defect It is detected again, calculates defect attenuation degree, be normal defects or false defect according to attenuation degree differentiation, to attenuation degree Judgement, can effectively improve nicety of grading, and improve efficiency of algorithm.

In any of the above-described technical solution, it is preferable that further include: the residual points of each valuable bills is calculated to defect mass center Euclidean distance；When the Euclidean distance of residual point to the defect mass center of any valuable bills is greater than or equal to pre-determined distance, Delete the first order child node generated according to the defect characteristic of any valuable bills.

In the technical scheme, it is compared by calculating Euclidean distance with pre-determined distance, can effectively avoid exception Influence of the point to defect characteristic.

In any of the above-described technical solution, it is preferable that the defect characteristic includes: energy, density, residual dot density, residual Point saturation degree, residual divergence and/or residual black and white characteristic.

The fourth aspect of the present invention proposes a kind of quality detecting system of valuable bills, for carrying out matter to valuable bills Measure the secondary checking system verified, comprising: acquiring unit is detected as not for obtaining the cleaning-sorting machine in the secondary checking system The image information of qualified valuable bills；First allocation unit, for according to the secondary core recorded in the secondary checking system The image information of the underproof valuable bills is distributed to the secondary verification by the status data and allocation model for looking into node Node；Configuration unit, for obtaining the system configuration information of the secondary checking system, according to the system configuration information to institute Secondary verification node is stated to be configured；Trigger unit, for obtaining the batch information of the underproof valuable bills, according to institute It states the batch information triggering secondary verification node to start image analysis work or terminate image analysis work, not conform to described in acquisition The kind information of the valuable bills of lattice triggers the secondary verification node and is switched to image detection corresponding with the kind information Template.

In the technical scheme, according to the status data of the secondary verification node recorded in secondary checking system and distribution mould The image information that cleaning-sorting machine is detected as underproof valuable bills is distributed to secondary verification node, can made unqualified by formula Valuable bills image information distributively more rationally, effectively improve secondary checking system to underproof valuable bills Verification efficiency and improve the automation performance of secondary checking system.

In the above-mentioned technical solutions, it is preferable that further include: storage unit, for by it is described it is secondary verify node it is secondary It verifies result and the batch information is associated storage, to obtain summarizing data；Extraction unit, for summarizing data from described In extract it is described it is secondary verify the image information and crown word number information that result is underproof valuable bills, and according to described batch The secondary verification result is that database is written in the image information of underproof valuable bills and crown word number information by secondary information, with It is handled for comprehensive judgement system.

In the technical scheme, pass through the figure by the secondary secondary verification result for verifying node for underproof valuable bills As information and crown word number information write-in database, so that comprehensive judgement system can be to the image of underproof valuable bills Whether information is handled again with the underproof valuable bills of determination really for underproof valuable bills, is further promoted The accuracy rate verified.

In any of the above-described technical solution, it is preferable that the extraction unit is also used to, from it is described summarize in data extract Detection process data, and by the detection process data be supplied to the complete image real-time memory system of association carry out content association and Complete storage.

In the technical scheme, by the way that the complete image of association will be supplied to from summarizing the detection process data extracted in data Real-time memory system is got off with carrying out content association and complete storage so as to will test process data recording.

In any of the above-described technical solution, it is preferable that further include: the second allocation unit, if being used for all secondary cores Node failure is looked into, the image information of the underproof valuable bills is distributed into the comprehensive judgement system, for institute It states comprehensive judgement system and finally determines whether the underproof valuable bills are qualified.

It in the technical scheme, can be by the image of underproof valuable bills when secondary verification node breaks down Information distributes to comprehensive judgement system, so that comprehensive judgement system finally determines whether underproof valuable bills are qualified, in this way, The reliability that secondary checking system can be improved, avoiding cannot be to underproof valuable ticket when secondary verification node breaks down According to the situation for carrying out secondary verification and causing more underproof valuable bills misjudged.

In any of the above-described technical solution, it is preferable that the allocation unit is specifically used for, if the allocation model is first When allocation model, according to the secondary Connection Time for verifying node and quality detecting system in the status data, by institute The image information for stating underproof valuable bills distributes to the secondary verification node；If the allocation model is the second distribution mould When formula, according to the processing speed in the status data, the image information of the underproof valuable bills is distributed to described Secondary verification node；If the allocation model is third allocation model, according to the processed amount in the status data by institute The image information for stating underproof valuable bills distributes to the secondary verification node.

In the technical scheme, under different allocation models, underproof valuable bills is distributed into secondary verification and are saved The foundation of point is different, in this way, the distribution of underproof valuable bills can be made more reasonable, so that improving underproof has The verification efficiency of valence bill.

Fifth aspect present invention proposes a kind of quality determining method of valuable bills, comprising: by all valuable bills Sample set is divided into multiple detection zones；By the corresponding sample set of each detection zone in the multiple detection zone according to spy Sign is clustered, and the corresponding sample set of each detection zone is divided into multiple classifications；With in each detection zone The sample set of each classification in multiple classifications learns the parameter space of corresponding detection zone, to obtain each detection zone pair The parameter space answered；Quality testing is carried out to the sample set in corresponding detection zone using each parameter space.

In the technical scheme, multiple detection zones that sample set divides are clustered according to feature, is divided into multiple classes Not, in this way, the space that each subclass is constituted can be more uniform and flat, while sample size is reduced, with the sample of each classification This collection learns corresponding parameter space, carries out quality inspection to the sample set in corresponding detection zone according to each parameter space It surveys, the stability of Assured Mode parser, and effectively reduces number of samples, calculation amount and the time of feature extraction, Improve efficiency of algorithm.

In the above-mentioned technical solutions, it is preferable that by the corresponding sample of each detection zone in the multiple detection zone Collection is clustered according to feature, is specifically included: using any sample in each detection zone as cluster centre Initial value；Calculate the first Euclidean distance in the detection zone between other samples and any sample；Described first It, otherwise, will other described samples using other described samples cluster centre new as one when Euclidean distance is greater than pre-determined distance As the cluster centered on any sample.

In the technical scheme, it through Euclidean distance compared with pre-determined distance, can be clustered to avoid abnormal point to determining The influence at center.

In any of the above-described technical solution, it is preferable that when using other described samples cluster centre new as one, Calculate separately the second Euclidean distance and the residue between the remaining sample and any sample in the detection zone Third Euclidean distance between sample and other described samples, Euclidean is with cluster belonging to the determination remaining sample.

Wherein it is preferred to which the step of determining cluster belonging to the remaining sample, specifically includes: in second Euclidean When distance is less than or equal to the pre-determined distance, using the remaining sample as the cluster centered on any sample, When the third Euclidean distance is less than or equal to the pre-determined distance, using the remaining sample as being with other described samples in The cluster of the heart, when second Euclidean distance and the third Euclidean distance are all larger than the pre-determined distance, described in comparison The size of second Euclidean distance and the third Euclidean distance is greater than the third Euclidean distance in second Euclidean distance When, the remaining sample is less than institute as the cluster centered on other described samples, and in second Euclidean distance When stating third Euclidean distance, using the remaining sample as the cluster centered on any sample.

In the technical scheme, the cluster centered on sample is determined compared with pre-determined distance by Euclidean distance, this Sample can make the space constituted more uniform and flat, while reduce sample size, and the stability of Assured Mode algorithm mentions High efficiency of algorithm.

In any of the above-described technical solution, it is preferable that calculate first Euclidean distance, institute according to following calculation formula State the second Euclidean distance and the third Euclidean distance:

Wherein, D indicates the Euclidean distance,Indicate that the mean vector of the sample as cluster centre, C are overall association Variance matrix, x indicate sample.

Sixth aspect present invention proposes a kind of image analysis method of valuable bills, comprising: according to preset order to having Multiple Informations of the image of valence bill are detected, wherein the Information include positive information, back side information, thoroughly Visual information and/or infrared information；When detecting that any Information is unqualified, it is determined as that described image is unqualified；It is detecting When each Information is qualified into the multiple Information, it is determined as described image qualification；Table images are not conformed to described Type of error and wrong process analyzed and recorded, the table images that do not conform to are managed and be counted.

In the technical scheme, repeated detection is carried out to the much information of image respectively, a set of image contains front, back More Informations such as face, perspective, infrared, if wherein a certain towards unqualified, then it is assumed that the set image is unqualified, only works as institute Have Information it is all qualified when, just think that the set image is qualified.It is equivalent in this way in the unqualified judgement of progress information to image Repeated detection has been carried out, so that determining that result is relatively reliable, has effectively improved the accuracy rate of judgement, also instead of artificial inspection It surveys, reduces verification cost.

In the above-mentioned technical solutions, it is preferable that include: when the number of image for needing to detect is multiple, will be described more A image distributes to multiple and different detection nodes, simultaneously using the plurality of different detection node to described multiple images It is detected.

In the technical scheme, when the amount of images that certain a batch needs determines is excessive, multinode can be used and ship Capable mode carries out the judgement of image, to save the judgement time of described multiple images.Multinode operation is related to internal system Dispatching method, can respectively operating condition distributes image to be sentenced to different nodes according to each node.

In any of the above-described technical solution, it is preferable that further include: do not conform to table images and the qualification described in obtaining respectively The crown word number information of image；The crown word number information for not conforming to table images is added in blacklist, by the qualified images Crown word number information is added in white list, and by the crown word number information for not conforming to table images the or described qualified images The crown word number information for meeting specified requirements in crown word number information is added in gray list.

In the technical scheme, the access of result is determined all using database as main carriers, and system is sentenced in decision process Determine procedural information and final result is synchronously written in database, when software and hardware is abnormal or other collapse cases, system can be with intelligence Current judgement state is reverted in the slave database of energy, with the consistency for keeping system to run.Therefore, the crown word number of image is believed Breath, which is added in list, to be saved, and allows system in exception or other judgement knots for supporting to have access to image in special circumstances Fruit really restores the judgement information of previous image.Wherein, system supports the crown word number inventory of three types to export: white list is Finger system is determined as the crown word number collection of qualified images；Blacklist refers to that system is judged to not conforming to the crown word number collection of table images；Grey name Singly refer to the crown word number collection for meeting specified requirements.

In any of the above-described technical solution, it is preferable that the type of error for not conforming to table images include: set-off, ink blok, Dog-ear lacks print；The wrong process for not conforming to table images includes: blank sheet of paper, offset printing, gravure.

In any of the above-described technical solution, it is preferable that according to the type of error and/or mistake of the underproof image The preset order is arranged in process.

In the technical scheme, can be customized towards judgement sequence, it is in general, in need to institute to sentence when judging Fixed image first determines its positive information, if the front of certain set image be determined as it is unqualified, the set image be it is unqualified, should Cover image remaining towards no longer needing to determine；After the front of all images determines, the qualified image of front judgement, then into The row back side determines.Above-mentioned rule is equally applicable to other Informations such as judgement perspective, infrared.For example, known to this batch of image thoroughly Determine perspective towards determining that result determines its front, back again according to perspective first depending on towards unqualified more, can choose default Face etc. is towards to save the operational efficiency for determining the time and improving system.Furthermore it is also possible to be set as unified towards setting for multiple Priority, i.e., by multiple standards towards the judgement that is set as being locked in a tie for first.Because of certain information in some cases, on image It needs multi-panel to combine just can determine that whether it is qualified, can show any amount towards (for example only opening positive/infrared simultaneously at this time Two towards) carry out integrated information judgement.In this mode, the result that system determines is the final result of the set image, nothing Need to carry out again it is other towards judgement.

The seventh aspect of the present invention proposes a kind of fault management system of valuable bills, comprising: generation unit, for obtaining The sample image of valuable bills is taken, and generates the template image for carrying the residual point of those suspected defects according to the sample image；First Computing unit, for calculate multiple those suspected defects in the template image it is residual point composition geometric figure mass center position and The position of geometric center；Judging unit, for judging described more according to the position of the mass center and the position of the geometric center Whether the residual point of any those suspected defects in a residual point of those suspected defects is the residual point of defect.

In the above-mentioned technical solutions, it is preferable that first computing unit is specifically used for, and calculates the multiple those suspected defects The position of first mass center of the geometric figure of residual point composition and the position of the first geometric center, and calculate except described any doubtful scarce Fall into the position of the second mass center of the geometric figure of the residual point composition of other those suspected defects except residual point and the position of the second geometric center It sets；The judging unit is specifically used for, and judges the position of first mass center and the distance between the position of second mass center Whether first threshold is greater than, and the position of first geometric center and the distance between the position of second geometric center are It is no be less than second threshold, if judging result be it is no, any residual point of those suspected defects be the residual point of the defect, it is otherwise, described Any residual point of those suspected defects is not the residual point of the defect.

In any of the above-described technical solution, it is preferable that the generation unit includes: the second computing unit, for according to institute The point of proximity for stating each sample point in sample image calculates weight matrix；The generation unit is specifically used for, according to the power Value matrix generates the template image for carrying the residual point of the those suspected defects.

In any of the above-described technical solution, it is preferable that second computing unit is specifically used for, and is calculated by the following formula The weight matrix:

The eighth aspect of the present invention proposes a kind of fault management system of valuable bills, comprising: first acquisition unit is used In the sample image for obtaining valuable bills；Converter unit, for carrying out Fourier transform to the sample image, described in determination The magnitude image and phase image of sample image；Second acquisition unit, for obtaining the corresponding amplitude reduction of the magnitude image Image and the corresponding phase X of the phase image are to local derviation also original image and phase Y-direction local derviation also original image；Generation unit, For being generated according to the amplitude also original image, the phase X to local derviation also original image and the phase Y-direction local derviation also original image Also original image；Difference unit determines institute according to difference result for the sample image and the original image of going back to be carried out difference State the residual image of valuable bills.

In the above-mentioned technical solutions, it is preferable that the second acquisition unit includes: processing unit, is used for the phase The phase of any image in image is appointed in the phase image except described as basic phase, and according to the basic phase The phase of other images except one image is handled, to obtain phase diagram aberration；The second acquisition unit is specifically used for, Determine that the amplitude also original image, the phase X are inclined to local derviation also original image and the phase Y-direction according to the phase diagram aberration It leads and goes back original image.

In any of the above-described technical solution, it is preferable that the generation unit includes: the first determination unit, for according to institute Phase X is stated to local derviation also original image or the phase Y-direction local derviation also original image and initial point, determines unwrapped phase also original image Picture；Second determination unit, for determining phase also original image according to the unwrapped phase also original image and the basic phase； Inverse transformation unit, for carrying out inverse fourier transform to the phase also original image and the amplitude also original image, to determine It states and goes back original image.

In any of the above-described technical solution, it is preferable that the second acquisition unit includes: computing unit, for described Phase diagram aberration carries out local derviation calculating to Y-direction in X respectively, with obtain the X of the phase diagram aberration to partial derivative and described The partial derivative of the Y-direction of phase diagram aberration；Reduction unit, for the magnitude image, the X to partial derivative, the Y-direction Partial derivative is analyzed and is restored, and determines the amplitude also original image, the phase X to local derviation also according to analysis and reduction result Original image and the phase Y-direction local derviation also original image.

In any of the above-described technical solution, it is preferable that the processing unit is specifically used for, by the phase of other images Subtraction is carried out with the basic phase, and phase unwrapping is carried out around processing, according to phase unwrapping around processing to calculated result As a result the phase diagram aberration is determined.

Ninth aspect present invention proposes a kind of fault management system of valuable bills, comprising: generation unit is used for basis The defect characteristic of each valuable bills in valuable bills training set generates first order child node；Taxon, it is each for calculating The partition value of the first order child node, and classified according to the partition value to the defect characteristic；Processing unit is used for Judge whether all first order child nodes can not classify again, when the judgment result is yes, classification is completed.

Wherein it is preferred to which the taxon is specifically used for: if the partition value is less than given threshold, using support Vector machine method classifies to the defect characteristic, and if the partition value be more than or equal to the given threshold, use Traditional decision-tree classifies to the defect characteristic.

In any of the above-described technical solution, it is preferable that the generation unit is specifically used for: calculating the valuable bills instruction Information gain value between the defect characteristic of each valuable bills of white silk concentration and the defect characteristic of other valuable bills, and according to Each valuable bills corresponding information gain value construction first order child node.

In any of the above-described technical solution, it is preferable that further include: computing unit, for calculating each valuable bills Euclidean distance of the residual point to defect mass center；Unit is deleted, for the residual point in any valuable bills to the Europe of defect mass center When family name's distance is greater than or equal to pre-determined distance, the first order generated according to the defect characteristic of any valuable bills is deleted Child node.

The tenth aspect of the present invention proposes a kind of quality determining method of valuable bills, for carrying out matter to valuable bills Measure the secondary checking system verified, comprising: the cleaning-sorting machine obtained in the secondary checking system is detected as underproof valuable ticket According to image information；It, will according to the status data and allocation model of the secondary verification node recorded in the secondary checking system The image information of the underproof valuable bills distributes to the secondary verification node；Obtain the secondary checking system is System configuration information, configures the secondary verification node according to the system configuration information；Obtaining described underproof has The batch information of valence bill triggers the secondary verification node according to the batch information and starts image analysis work or end figure As analysis work；The kind information of the underproof valuable bills is obtained, the triggering secondary verification node is switched to and institute State the corresponding image detection template of kind information.

In the above-mentioned technical solutions, it is preferable that further include: by the secondary secondary verification result of node and described verified Batch information is associated storage, to obtain summarizing data；From it is described summarize extract the secondary verification result in data and be The image information and crown word number information of underproof valuable bills, and be by the secondary verification result according to the batch information Database is written in the image information and crown word number information of underproof valuable bills, so that comprehensive judgement system is handled.

In any of the above-described technical solution, it is preferable that described by the secondary secondary verification result for verifying node and institute State batch information and be associated storage, after obtaining the step of summarizing data, further includes: from it is described summarize in data extract Detection process data, and by the detection process data be supplied to the complete image real-time memory system of association carry out content association and Complete storage.

In any of the above-described technical solution, it is preferable that further include: it, will if all secondary verification nodes break down The image information of the underproof valuable bills distributes to the comprehensive judgement system, so that the comprehensive judgement system is final Determine whether the underproof valuable bills are qualified.

In any of the above-described technical solution, it is preferable that described according to the secondary verification recorded in the secondary checking system The image information of the underproof valuable bills is distributed to the secondary verification and saved by the status data and allocation model of node It the step of point, specifically includes: if the allocation model is the first allocation model, according to described secondary in the status data The Connection Time for verifying node and quality detecting system, the image information of the underproof valuable bills is distributed to described two Secondary verification node；It, will be described according to the processing speed in the status data if the allocation model is the second allocation model The image information of underproof valuable bills distributes to the secondary verification node；If the allocation model is third allocation model When, the image information of the underproof valuable bills distributed to according to the processed amount in the status data described secondary Verify node.

Tenth one side of the invention proposes a kind of quality detecting system of valuable bills, comprising: division unit, being used for will The sample set of all valuable bills is divided into multiple detection zones；Cluster cell, for will be every in the multiple detection zone The corresponding sample set of a detection zone is clustered according to feature, the corresponding sample set of each detection zone is divided into more A classification；Unit, for learning to correspond to the sample set of each classification in each detection zone in multiple classifications Detection zone parameter space, to obtain the corresponding parameter space of each detection zone；Detection unit, for using each ginseng Number space carries out quality testing to the sample set in corresponding detection zone.

In the above-mentioned technical solutions, it is preferable that the cluster cell includes: setting unit, is used for each detection Initial value of any sample as a cluster centre in region；Computing unit, for calculating other in the detection zone The first Euclidean distance between sample and any sample；Confirmation unit, it is default for being greater than in first Euclidean distance Apart from when, otherwise other described samples cluster centre new as one, other described samples is used as with any sample Cluster centered on this.

In any of the above-described technical solution, it is preferable that the computing unit is also used in other sample conducts by described in When one new cluster centre, the second Europe between the remaining sample and any sample in the detection zone is calculated separately Third Euclidean distance between family name's distance and the remaining sample and other described samples, with the determination remaining sample institute The cluster of category.

Wherein it is preferred to the determination unit, be also used to second Euclidean distance be less than or equal to it is described it is default away from From when, using the remaining sample as the cluster centered on any sample, be less than or wait in the third Euclidean distance When the pre-determined distance, using the remaining sample as the cluster centered on other described samples, in second Euclidean When distance and the third Euclidean distance are all larger than the pre-determined distance, second Euclidean distance and the third Europe The size of family name's distance, when second Euclidean distance is greater than the third Euclidean distance, using the remaining sample as with institute The cluster centered on other samples is stated, and when second Euclidean distance is less than the third Euclidean distance, it will be described surplus Remaining sample is as the cluster centered on any sample.

In any of the above-described technical solution, it is preferable that the computing unit is specifically used for, according to following calculation formula meter Calculate first Euclidean distance, second Euclidean distance and the third Euclidean distance:

Through above technical scheme, under conditions of not influencing detection accuracy, it is possible to reduce number of samples, calculation amount and The time of feature extraction, improve efficiency of algorithm.

The twelfth aspect of the present invention also proposed a kind of image analysis system of valuable bills, comprising: detection unit is used for It is detected according to multiple Informations of the preset order to the image of valuable bills, wherein the Information includes front Information, back side information, perspective information and/or infrared information；Judging unit, for detecting that any Information is unqualified When, it is determined as that described image is unqualified, and when each Information is qualified in detecting the multiple Information, sentence It is set to described image qualification；Processing unit, for being analyzed simultaneously the type of error for not conforming to table images and wrong process Record, the table images that do not conform to are managed and be counted.

In the above-mentioned technical solutions, it is preferable that include: allocation unit, for being more when the number for needing the image detected When a, described multiple images are distributed to multiple and different detection nodes, to use the plurality of different detection node simultaneously Described multiple images are detected.

In the above-mentioned technical solutions, it is preferable that further include: acquiring unit described does not conform to table images for obtaining respectively With the crown word number information of the qualified images；Adding unit is black for the crown word number information for not conforming to table images to be added to In list, the crown word number information of the qualified images is added in white list, and by the crown word number for not conforming to table images The crown word number information for meeting specified requirements in the crown word number information of in information the or described qualified images is added in gray list.

In any of the above-described technical solution, it is preferable that the type of error for not conforming to table images include: set-off, ink blok, Dog-ear and/or scarce print；The wrong process for not conforming to table images includes: blank sheet of paper, offset printing and/or gravure.

Pass through above technical scheme, it is possible to reduce number of samples, calculation amount and the time of feature extraction, also can be improved two The verification efficiency of secondary checking system improves the accuracy rate of verification, reduces and verifies cost, improves nicety of grading, improves quality testing The robustness of system.

Detailed description of the invention

Fig. 1 shows the defect management side of the valuable bills of one embodiment of embodiment according to the first aspect of the invention The flow diagram of method；

Fig. 2 shows the defect management sides of the valuable bills of one embodiment of embodiment according to the second aspect of the invention The flow diagram of method；

Fig. 3 shows the defect management side of the valuable bills of one embodiment of embodiment according to the third aspect of the invention we The schematic flow diagram of method；

Fig. 4 shows the defect management of the valuable bills of another embodiment of embodiment according to the third aspect of the invention we The schematic flow diagram of method；

Fig. 5 shows the quality testing system of the valuable bills of one embodiment of embodiment according to the fourth aspect of the invention The structural schematic diagram of system；

Fig. 6 shows the quality testing side of the valuable bills of one embodiment of embodiment according to the fifth aspect of the invention The schematic flow diagram of method；

Fig. 7 shows the quality testing of the valuable bills of another embodiment of embodiment according to the fifth aspect of the invention The schematic flow diagram of method；

Fig. 8 shows the image analysis side of the valuable bills of one embodiment of embodiment according to the sixth aspect of the invention The schematic flow diagram of method；

Fig. 9 shows the image analysis of the valuable bills of another embodiment of embodiment according to a sixth aspect of the present invention The system processing schematic of method；

Figure 10 shows the defect management of the valuable bills of one embodiment of embodiment according to the seventh aspect of the invention The structural schematic diagram of system；

Figure 11 shows the defect management of the valuable bills of one embodiment of embodiment according to the eighth aspect of the invention The structural schematic diagram of system；

Figure 12 shows the defect management of the valuable bills of one embodiment of embodiment according to the ninth aspect of the invention The schematic block diagram of system；

Figure 13 shows the quality testing of the valuable bills of one embodiment of embodiment according to the tenth aspect of the invention The flow diagram of method；

Figure 14 shows the quality inspection of the valuable bills of another embodiment of embodiment according to the tenth aspect of the invention The flow diagram of survey method；

Figure 15 shows the structure of the secondary checking system of one embodiment of embodiment according to the tenth aspect of the invention Schematic diagram；

Figure 16 shows the quality inspection of the valuable bills of one embodiment of embodiment according to the eleventh aspect of the invention The schematic block diagram of examining system；

Figure 17 shows the images of the valuable bills of one embodiment of embodiment according to the twelfth aspect of the invention point The schematic block diagram of analysis system.

Specific embodiment

To better understand the objects, features and advantages of the present invention, with reference to the accompanying drawing and specific real Applying mode, the present invention is further described in detail.It should be noted that in the absence of conflict, the implementation of the application Feature in example and embodiment can be combined with each other.

In the following description, numerous specific details are set forth in order to facilitate a full understanding of the present invention, still, the present invention may be used also To be implemented using other than the one described here other modes, therefore, protection scope of the present invention is not by described below Specific embodiment limitation.

Fig. 1 shows the defect management side of the valuable bills of one embodiment of embodiment according to the first aspect of the invention The flow diagram of method.

As shown in Figure 1, the defect management method of valuable bills according to an embodiment of the invention, comprising:

Step 102, the sample image of valuable bills is obtained, and it is residual according to sample image generation to carry those suspected defects The template image of point；

Step 104, the position of the mass center of the geometric figure of the residual point composition of multiple those suspected defects in the template image is calculated Set the position with geometric center；

Step 106, according to the position of the position of the mass center and the geometric center, judge that the multiple those suspected defects are residual Whether the residual point of any those suspected defects in point is the residual point of defect.

In the above-mentioned technical solutions, it is preferable that step 104 and step 106 include: to calculate the multiple residual point of those suspected defects The position of first mass center of the geometric figure of composition and the position of the first geometric center, and calculate except any those suspected defects are residual The position of second mass center of the geometric figure of the residual point composition of other those suspected defects except point and the position of the second geometric center；Sentence Whether break the distance between the position of first mass center and the position of second mass center is greater than first threshold, and described first Whether the distance between the position of geometric center and the position of second geometric center are less than second threshold；If judging result is No, then the residual point of any those suspected defects is the residual point of the defect, and otherwise, any residual point of those suspected defects is not the defect Residual point.

In any of the above-described technical solution, it is preferable that step 102 specifically includes: according to each of described sample image The point of proximity of sample point calculates weight matrix；The mould for carrying the residual point of the those suspected defects is generated according to the weight matrix Domain picture.

Fig. 2 shows the defect management sides of the valuable bills of one embodiment of embodiment according to the second aspect of the invention The flow diagram of method.

As shown in Fig. 2, the defect management method of valuable bills according to an embodiment of the invention includes:

Step 202, the sample image of valuable bills is obtained；

Step 204, Fourier transform is carried out to the sample image, with the magnitude image and phase of the determination sample image Bit image；

Step 206, the corresponding amplitude of the magnitude image also original image and the corresponding phase of the phase image are obtained X is to local derviation also original image and phase Y-direction local derviation also original image；

Step 208, according to the amplitude also original image, the phase X to local derviation also original image and the phase Y-direction local derviation Original image is gone back in also original image generation；

Step 210, the sample image and the original image of going back are subjected to difference, are determined according to difference result described valuable The residual image of bill.

In the above-mentioned technical solutions, it is preferable that step 206 specifically includes: by any image in the phase image Phase is used as basic phase, and according to the basic phase to other figures in the phase image in addition to any image The phase of picture is handled, to obtain phase diagram aberration；The amplitude also original image, described is determined according to the phase diagram aberration Phase X is to local derviation also original image and the phase Y-direction local derviation also original image.

In any of the above-described technical solution, it is preferable that step 208 specifically includes: according to the phase X to local derviation also original image Picture or the phase Y-direction local derviation also original image and initial point, determine unwrapped phase also original image；According to the unwrapped phase Also original image and the basic phase determines phase also original image；To the phase also original image and the amplitude also original image into Row inverse fourier transform, to go back original image described in determination.

Fig. 3 shows the defect management side of the valuable bills of one embodiment of embodiment according to the third aspect of the invention we The flow chart of method.

As shown in figure 3, the defect management method of the valuable bills of embodiment according to the present invention, comprising:

Step 302, according to the defect characteristic of valuable bills each in valuable bills training set, first order child node is generated；

Step 304, the partition value of each first order child node is calculated, and special to the defect according to the partition value Sign is classified；

Step 306, judge whether all first order child nodes can not classify again, when the judgment result is yes, classification It completes.

In any of the above-described technical solution, it is preferable that according to each valuable bills in the training set of valuable bills It the step of defect characteristic, generation first order child node, specifically includes: calculating each valuable bills in the valuable bills training set Defect characteristic and other valuable bills defect characteristic between information gain value；It is corresponding according to each valuable bills Information gain value constructs the first order child node.

Specifically, technical solution of the present invention can be embodied by following multiple embodiments:

Embodiment one: by according to the energy of valuable bills each in valuable bills training set, density, residual dot density, residual The defects of putting saturation degree, residual divergence and/or residual black and white characteristic feature, generates first order child node, and calculate each first The partition value of grade child node classifies to defect characteristic using support vector machine method with being less than given threshold in partition value； It is more than or equal to given threshold in partition value, is classified using traditional decision-tree to the defect characteristic, until all first order Child node can not classify again, determine that classification is completed, can effectively improve nicety of grading, calculating speed, which is also fallen below, to be received Range, improve efficiency of algorithm.Wherein, training set is a certain number of samples being trained to the parameter of system.

Embodiment two: on the basis of example 1, the defect characteristic of each valuable bills of calculating can also specifically be passed through Information gain value between the defect characteristic of other valuable bills, and according to the corresponding information gain value structure of each valuable bills First order child node is made, can effectively avoid the extraction of false defect, and then detected again to the residual point of defect, calculates defect Attenuation degree, being distinguished according to attenuation degree is normal defects or false defect, and the judgement to attenuation degree can be mentioned effectively High-class precision, and improve efficiency of algorithm.

Embodiment three: on the basis of example 1, abnormal point specifically can also further be excluded to the shadow of defect characteristic It rings: calculating the residual point of each valuable bills to the Euclidean distance of defect mass center, and in the residual point of any valuable bills to defect matter When the Euclidean distance of the heart is greater than or equal to pre-determined distance, the first order generated according to the defect characteristic of any valuable bills is deleted Node.

Technical solution of the present invention is described further below in conjunction with Fig. 4.

As shown in figure 4, the defect management method of valuable bills according to the present invention, comprising:

Step 402, feature 1 is extracted, feature 1 and 1 is compared, if feature 1 is greater than 1, enters step 404；If feature 1 When less than or equal to 1,414 are entered step.

Step 404, it extracts feature 2 and enters step 406 if feature 2 is less than 0；If feature 2 is more than or equal to 0, enter Step 408.

Step 406, classification 1 is obtained.

Step 408, SVM carries out svm classifier using all training sets on the node.

Step 410, classification 2 is obtained.

Step 412, classification n is obtained.

Step 414, SVM carries out svm classifier using all training sets on the node.

Step 416, classification 1 is obtained.

Step 418, classification n is obtained.

Specific steps are as follows:

One, feature extraction optimizes

It 1) is the extraction for avoiding false defect, point residual to defect is detected again, gradually reinforces parameter, calculates defect decaying Degree.It is normal defects or false defect according to attenuation degree differentiation.False defect label is added in classification learning simultaneously, Whether can be distinguished by sorting algorithm is real defect.

2) influence in order to avoid abnormal point to defect characteristic increases anti-interference process in characteristic extraction procedure.Pass through It polymerize the residual point of defect, calculates each residual point to the Euclidean distance of defect mass center, delete apart from excessive noise spot.

3) effective defect characteristic, energy, area, residual dot density, residual saturation degree, residual divergence, residual black and white are designed Characteristic etc..

Two, sorting algorithm step

If defect characteristic is F={ f₁,f₂,...,f_n, classification marker C={ C₁,C₂,...,C_m}。

1) according to the defect characteristic of training set, root node, that is, first order child node of spanning tree.Classify in child node Partition value is maximum, so that gap width is from maximum.

2) each node allocation value in first order child node is calculated, if partition value is less than given threshold, illustrates the node In characteristic value, be difficult to reach preferable classifying quality with decision tree, then classified using SVM；For can further divide The node of class, classifies according to traditional decision-tree.

3) when carrying out svm classifier, should be classified using the characteristic value on the node, or risen using kernel function Dimension classification.

4) it calculates whether all nodes can not classify again, such as if so, classification is completed, if not having, then repeats second and third Step, it is through to divide again.

Three, feature extraction optimizes

It 1) is the extraction for avoiding false defect, point progress residual to defect is detecting, and gradually reinforces parameter, calculates defect decaying Degree.It is normal defects or false defect according to attenuation degree differentiation.False defect label is added in classification learning simultaneously, Whether can be distinguished by sorting algorithm is real defect.

Four, sorting algorithm step

Assuming that obtaining defect attribute is F={ f₁,f₂,...,f_k, defect type is denoted as C={ C₁,C₂,...,C_m}.Training Sample sample set S={ x₁,x₂,...,x_n, decision Tree algorithms use ID3.

1) information gain value Gain (S, the f between computation attribute F_i), wherein i=1,2 ..., k, indicate attribute f_iGathering Information gain on S.

2) maximum attribute Gain (S, f are selected_i) it is used as decision tree nodes.

3) according to attribute f_iDiscrete value d construct child node d_j, j=1,2 ..., l, and be S sample set S points_jIt is right respectively It should be in d_j, indicate f_iThere is l kind probable value.

4) all child node d are calculated_jCorresponding sample set S_jInformation gain value Gain (S, f_p), wherein p=1,2 ..., k, p ≠i。

If 5) Gain (S, f_p) >=T (T is gain threshold), it can continue to repeat step 1) to 3) progress decision tree point Class；If Gain (S, f_p) < T, then carry out step 6).

6) child node d_jCorresponding sample set S_jClassified using support vector machines (SVM) method, classification results Directly as d_jLeaf node.

Step 4) is repeated to 6) until completing all child node d_jClassification, calculate the leaf section of all classification results Point.

Fig. 5 shows the quality testing system of the valuable bills of one embodiment of embodiment according to a fourth aspect of the present invention The structural schematic diagram of system.

As shown in figure 5, the quality detecting system 500 of valuable bills according to an embodiment of the invention, comprising: obtain Unit 502, the first allocation unit 504, configuration unit 506 and trigger unit 508, wherein the acquiring unit 502, for obtaining The cleaning-sorting machine in the secondary checking system is taken to be detected as the image information of underproof valuable bills；First allocation unit 504, for the status data and allocation model according to the secondary verification node recorded in the secondary checking system, by described in not The image information of qualified valuable bills distributes to the secondary verification node；The configuration unit 506, for obtaining described two The system configuration information of secondary checking system configures the secondary verification node according to the system configuration information；It is described Trigger unit 508, for obtaining the batch information of the underproof valuable bills, according to batch information triggering described two Secondary verification node starts image analysis work or terminates image analysis work, obtains the kind letter of the underproof valuable bills Breath triggers the secondary verification node and is switched to image detection template corresponding with the kind information.

In the above-mentioned technical solutions, it is preferable that further include: storage unit 510, for by it is described it is secondary verify node two Secondary verification result and the batch information are associated storage, to obtain summarizing data；Extraction unit 512 is used for from the remittance The secondary result of verifying is extracted in total data as the image information and crown word number information of underproof valuable bills, and according to The batch information is by the secondary result of verifying for image information and crown word number information the write-in number of underproof valuable bills According to library, so that comprehensive judgement system is handled.

In any of the above-described technical solution, it is preferable that the extraction unit 512 is also used to, from it is described summarize in data mention Detection process data are taken out, and the detection process data are supplied to the complete image real-time memory system of association and carry out content pass Connection and complete storage.

In any of the above-described technical solution, it is preferable that further include: the second allocation unit 514, if for all described secondary It verifies node to break down, the image information of the underproof valuable bills is distributed into the comprehensive judgement system, for The comprehensive judgement system finally determines whether the underproof valuable bills are qualified.

In any of the above-described technical solution, it is preferable that first allocation unit 504 is specifically used for, if the distribution mould When formula is the first allocation model, when according to the connection of the secondary verification node and quality detecting system in the status data Between, the image information of the underproof valuable bills is distributed into the secondary verification node；If the allocation model is the When two allocation models, according to the processing speed in the status data, by the image information of the underproof valuable bills point Secondary verification node described in dispensing；If the allocation model is third allocation model, according to the place in the status data The image information of the underproof valuable bills is distributed to the secondary verification node by reason amount.

Fig. 6 shows the quality testing side of the valuable bills of one embodiment of embodiment according to a fifth aspect of the present invention The schematic flow diagram of method.

As shown in fig. 6, the quality determining method of valuable bills according to an embodiment of the invention, comprising:

Step 602, the sample set of all valuable bills is divided into multiple detection zones；

Step 604, the corresponding sample set of each detection zone in the multiple detection zone is gathered according to feature The corresponding sample set of each detection zone is divided into multiple classifications by class；

Step 606, learn corresponding inspection with the sample set of each classification in each detection zone in multiple classifications The parameter space in region is surveyed, to obtain the corresponding parameter space of each detection zone；

Step 608, quality testing is carried out to the sample set in corresponding detection zone using each parameter space.

In any of the above-described technical solution, it is preferable that when using other described samples cluster centre new as one, Calculate separately the second Euclidean distance and the residue between the remaining sample and any sample in the detection zone Third Euclidean distance between sample and other described samples, with cluster belonging to the determination remaining sample.

Fig. 7 shows the quality testing of the valuable bills of another embodiment of embodiment according to a fifth aspect of the present invention The schematic flow diagram of method.

As shown in fig. 7, the quality determining method of valuable bills according to another embodiment of the invention, comprising:

Step 702, sample set is obtained, product area is divided into k detection zone, respectively step 704, step 706, Step 708.

Step 704, region 1 is divided.

Step 706, region 2 is divided, all products of sample set according to the feature clustering of detection zone, total m class, difference Enter step 710, step 714, step 718.

Step 708, region k is divided.

Step 710, sample set 1 learns the 1st detection zone with the sample set of the 1st class.

Step 712, parameter space 1 is obtained.

Step 714, sample set 2 learn the 2nd detection zone with the sample set of the 2nd class.

Step 716, parameter space 2 is obtained.

Step 718, sample set m learns m-th of detection zone with the sample set of m class.

Step 720, parameter space m is obtained.

Specific steps are as follows:

One, learn part

Printing product is divided into k detection zone；

Each detection zone is carried out the following processing；

All products of sample set according to the feature clustering of i-th of detection zone, total m class, wherein

A), appoint and take a sample as the initial value centered on a cluster sample, such as enable z₁=x₁, z₁Indicate cluster sample Centered on this.

B), distance is calculatedWherein D indicates distance,It is the sample of cluster centre Mean vector, C is overall covariance matrix.

If c), D₂₁> T, wherein T indicates threshold value, it is determined that a new cluster sample is center z₂=x₂, otherwise, x₂Belong to With z₁For the cluster centered on sample.

D), assume that having cluster sample is center z₁、z₂, calculate distance D₃₁, D₃₂。

If e), D₃₁> T and D₃₂> T, then obtaining a new cluster sample is center z₃=x3, otherwise, x3 belongs to from z₁、z₂In Nearest person cluster.

F), so repeat down, until all sample classifications are finished, total m class.

G), with jth class, the parameter space ψ of sample set i-th of detection zone of study of j=1,2 ..., m_ij。

H), until all detection zones all handle completion, parameter space ψ is obtained_ij, wherein i=1,2 ..., k, j=1, 2,...,m

Two, detection part:

1) detection printing product x', is divided into k detection zone.

2) the i-th region, is calculated separately to center z_jDistance D_ij.Wherein i=1,2 ..., k, j=1,2 ..., m.

3) the i-th region of x', is classified as D_ijThe smallest one kind, such as x'=z₁。

4), using ψ_i1Parameter detects the i-th region of x'.

Step 2) is repeated to 4), until completion is all detected in all k regions.

The technical scheme of the present invention has been explained in detail above with reference to the attached drawings, according to the technical solution of the present invention, is not influencing Under conditions of detection accuracy, it is possible to reduce number of samples, calculation amount and the time of feature extraction, improve efficiency of algorithm.

Fig. 8 shows the image analysis side of the valuable bills of one embodiment of embodiment according to the sixth aspect of the invention The flow chart of method.

As shown in figure 8, the image analysis of the valuable bills of one embodiment of embodiment according to the sixth aspect of the invention Method, comprising:

Step 802, it is detected according to multiple Informations of the preset order to the image of valuable bills, wherein towards Information includes positive information, back side information, perspective information and/or infrared information.

Step 804, when detecting that any Information is unqualified, it is determined as that image is unqualified, is detecting multiple faces When each Information is qualified into information, it is determined as image qualification.

Step 806, the type of error and wrong process that do not conform to table images are analyzed and is recorded, with to not conforming to table images It is managed and counts.

In the technical scheme, repeated detection is carried out to the much information of image respectively, a set of image contains front, back More Informations such as face, perspective, infrared, if wherein a certain towards unqualified, then it is assumed that the set image is unqualified.Only work as institute Have Information it is all qualified when, just think that the set image is qualified.It is equivalent in this way in the unqualified judgement of progress information to image Repeated detection has been carried out, so that determining that result is relatively reliable, has effectively improved the accuracy rate of judgement, also instead of artificial inspection It surveys, reduces verification cost.

In the above-mentioned technical solutions, it is preferable that include: when the number for the image for needing to detect is multiple, by multiple figures As distributing to multiple and different detection nodes, to be detected simultaneously using multiple and different detection nodes to multiple images.

In the technical scheme, when the amount of images that certain a batch needs determines is excessive, multinode can be used and ship Capable mode carries out the judgement of image, to save the judgement time of multiple images.Multinode operation is related to the tune of internal system Degree method, can respectively operating condition distributes image to be sentenced to different nodes according to each node.

In any of the above-described technical solution, it is preferable that further include: acquisition does not conform to table images and qualified images respectively Crown word number information；The crown word number information for not conforming to table images is added in blacklist, the crown word number information of qualified images is added It is specified into white list, and by meeting in the crown word number information for not conforming to table images or qualified images crown word number information The crown word number information of condition is added in gray list.

In any of the above-described technical solution, it is preferable that the type of error for not conforming to table images includes: set-off, ink blok, folding Angle lacks print；The wrong process for not conforming to table images includes: blank sheet of paper, offset printing, gravure.

In any of the above-described technical solution, it is preferable that according to the type of error of underproof image and/or wrong work Preset order is arranged in sequence.

It embodiment one: can preset suitable according to the type of error and/or wrong process, setting of underproof image first Sequence, wherein the type of error for not conforming to table images includes: set-off, ink blok, dog-ear and/or scarce print etc.；The wrong work of table images is not conformed to Sequence includes: blank sheet of paper, offset printing and/or gravure etc., is then believed according to preset order the positive information of the image of valuable bills, the back side Multiple Informations such as breath, perspective information and/or infrared information are detected, when detecting that any Information is unqualified, It is determined as that image is unqualified；When each Information is qualified only in detecting multiple Informations, just it is determined as image Qualification, and the type of error and wrong process that do not conform to table images are analyzed and recorded, to carry out pipe to not conforming to table images Reason and statistics effectively improve the accuracy rate of judgement so that determining that result is relatively reliable, also instead of artificial detection, reduce Verification cost.

Embodiment two: on the basis of can also be embodiment one, increasing the detection of the number of the image detected to needs, when When the number for the image for needing to detect is multiple, multiple images are distributed to multiple and different detection nodes, simultaneously using more A different detection node detects multiple images, can be with so that when certain amount of images for determining of a batch needs is excessive The judgement of image is carried out using the mode that multinode is concurrently run, to save the judgement time of multiple images, multinode operation is related to And the dispatching method to internal system, it can respectively operating condition distributes image to be sentenced to different nodes according to each node.

Embodiment three: can also increase on the basis of example 1 to not conforming to table images and qualified images prefixs The detection of number information, does not conform to table images and qualified images crown word number information by obtaining respectively, will not conform to the hat of table images Font size information is added in blacklist, the crown word number information of qualified images is added in white list, and will not conform to table images Crown word number information in or qualified images crown word number information in the crown word number information for meeting specified requirements be added to grey name Dan Zhong allows system in exception or other judgements for supporting to have access to image in special circumstances as a result, the previous figure of true reduction The judgement information of picture.

Technical solution of the present invention is described further below in conjunction with Fig. 9.

As shown in figure 9, cleaning-sorting machine 902 is examined according to multiple Informations of the preset order to the image of valuable bills It surveys, wherein Information includes positive information, back side information, perspective information and/or infrared information.Cleaning-sorting machine 902 and secondary core It looks into automatic checkout system 904 to be connected, the testing result of the image to certain amount (such as 7000) valuable bills is sent to two Secondary verification automatic checkout system 904, secondary verification automatic checkout system 904 distinguish qualified products, one by its secondary verification As waste product and serious waste product, and by distinguish result be sent to secondary verification image synthesis Analysis And Evaluation system 906.Secondary verification Image synthesis Analysis And Evaluation system 906 analyzes the type of error for not conforming to table images and wrong process, and records, with right Do not conform to table images to be managed and count, obtains the crown word number letter of the image of qualified products, general waste product and serious waste product respectively Breath, its crown word number information is sent in corresponding output inventory.Wherein, system supports the crown word number inventory of three types defeated Out: white list refers to that system is determined as the crown word number collection of qualified images；Blacklist refers to that system is judged to not conforming to the hat of table images Font size collection；Gray list refers to the crown word number collection for meeting specified requirements.

In the technical scheme, repeated detection is carried out to the much information of image respectively, a set of image contains front, back More Informations such as face, perspective, infrared, if wherein a certain towards unqualified, then it is assumed that the set image is unqualified.Only work as institute Have Information it is all qualified when, just think that the set image is qualified.It is equivalent in this way in the unqualified judgement of progress information to image Repeated detection has been carried out, so that determining that result is relatively reliable, has effectively improved the accuracy rate of judgement.

In addition, the access of judgement result is all using database as main carriers, system is in decision process, decision process information Be synchronously written in database with final result, when software and hardware is abnormal or other collapse cases, system can intelligence slave data Current judgement state is reverted in library, with the consistency for keeping system to run.Therefore, the crown word number information of image is added to name It is saved in list, allows system in exception or other judgements for supporting to have access to image in special circumstances as a result, true go back The judgement information of former previous image.

Figure 10 shows the defect management system of the valuable bills of one embodiment of embodiment according to a seventh aspect of the present invention The structural schematic diagram of system.

As shown in Figure 10, the fault management system of valuable bills according to an embodiment of the invention, comprising: generate single Member 1002, the first computing unit 1004 and judging unit 1006, wherein the generation unit 1002, for obtaining valuable bills Sample image, and generated according to the sample image and carry the template image of the residual point of those suspected defects；Described first calculates list Member 1004, for calculating the position of mass center of the geometric figure of the residual point composition of multiple those suspected defects in the template image and several The position at what center；The judging unit 1006, for according to the position of the mass center and the position of the geometric center, judgement Whether the residual point of any those suspected defects in the multiple residual point of those suspected defects is the residual point of defect.

In the above-mentioned technical solutions, it is preferable that first computing unit 1004 is specifically used for, and calculates the multiple doubtful The position of first mass center of the geometric figure of the residual point composition of defect and the position of the first geometric center, and calculate except described any doubtful Position and the second geometric center like the second mass center of the geometric figure of the residual point composition of other those suspected defects except the residual point of defect Position；The judging unit 1006 is specifically used for, judge first mass center position and second mass center position it Between distance whether be greater than first threshold, and between the position of first geometric center and the position of second geometric center Distance whether be less than second threshold, if judging result be it is no, any residual point of those suspected defects be the residual point of the defect, it is no Then, the residual point of any those suspected defects is not the residual point of the defect.

In any of the above-described technical solution, it is preferable that the generation unit 1002 includes: the second computing unit 10022, is used According to the point of proximity of each sample point in sample image calculating weight matrix；The generation unit 1002 is specifically used In being generated according to the weight matrix and carry the template image of the residual point of the those suspected defects.

It is described in detail below by one embodiment and generates the template figure for carrying the residual point of those suspected defects according to sample image The method of picture.

Carrying the template image of the residual point of those suspected defects according to sample image generation is exactly by the data in high-dimensional space Point (i.e. sample point) is mapped in low dimensional space.Specific steps are divided into three steps: the first step finds the k of each sample point A point of proximity；Second step is calculated the weight matrix of the sample point by the point of proximity of each sample point；Third step, according to each The point of proximity and weight matrix of sample point calculate the output valve of the sample point, finally, generating template image according to output valve.

Wherein, it is calculated by the following formula the weight matrix:

Wherein, ε (w) indicates error amount, X_iIndicate any sample point, X_j(j=1,2 ..., k) indicates the k of any sample point A point of proximity, w_ijIndicate the weight matrix between any sample point and point of proximity.

Then, in the case where keeping weight matrix constant, if output valve (data point i.e. in low dimensional space) is Y_i, Can then output valve be calculated by the following formula:

Wherein, φ (w) indicates loss function value, Y_j(j=1,2 ..., k) indicates output valve Y_iK point of proximity, w_ijTable Show Y_iAnd Y_jBetween weight matrix (i.e. X_iAnd X_jBetween weight matrix)

Above-mentioned formula can convert are as follows:

Wherein, M_ij=(I-w_ij)^T(I-w_ij), I indicates that a unit covariance matrix, T are indicated to (I-w_ij) seek transposition square Battle array.

In addition, Y_iNeed to meet two constraint conditions, i.e. ∑_iY_i=0 He(the number of N expression sample point Amount, T are indicated to Y_iSeek transposed matrix).

Calculating output valve Y_iAfterwards, so that it may according to Y_iGenerate template image.

Figure 11 shows the defect management of the valuable bills of one embodiment of embodiment according to the eighth aspect of the invention The structural schematic diagram of system.

As shown in figure 11, the fault management system 1100 of valuable bills according to an embodiment of the invention, comprising: the One acquiring unit 1102, converter unit 1104, second acquisition unit 1106, generation unit 1108 and difference unit 1110, wherein The first acquisition unit 1102, for obtaining the sample image of valuable bills；The converter unit 1104, for the sample This image carries out Fourier transform, with the magnitude image and phase image of the determination sample image；The second acquisition unit 1106, for obtaining the corresponding amplitude of the magnitude image also corresponding phase X of original image and the phase image to local derviation Also original image and phase Y-direction local derviation also original image；The generation unit 1108, for according to the amplitude also original image, described Phase X goes back original image to local derviation also original image and the phase Y-direction local derviation also original image generation；The difference unit 1110 is used In the sample image and the original image of going back are carried out difference, the residual plot of the valuable bills is determined according to difference result Picture.

In the above-mentioned technical solutions, it is preferable that the second acquisition unit 1106 includes: processing unit 11062, and being used for will The phase of any image in the phase image is used as basic phase, and according to the basic phase in the phase image The phase of other images in addition to any image is handled, to obtain phase diagram aberration；The second acquisition unit 1106 are specifically used for, according to the phase diagram aberration determine the amplitude also original image, the phase X to local derviation also original image and The phase Y-direction local derviation also original image.

In any of the above-described technical solution, it is preferable that the generation unit 1108 includes: the first determination unit 11082, is used In, to local derviation also original image or the phase Y-direction local derviation also original image and initial point, determining unwrapped phase according to the phase X Also original image；Second determination unit 11084, for determining phase according to the unwrapped phase also original image and the basic phase Original image is gone back in position；Inverse transformation unit 11086, for carrying out Fourier to the phase also original image and the amplitude also original image Inverse transformation, to go back original image described in determination.

In any of the above-described technical solution, it is preferable that the second acquisition unit 1106 includes: computing unit 11064, is used In to the phase diagram aberration X to and Y-direction carry out local derviation calculating respectively, with obtain the X of the phase diagram aberration to local derviation The partial derivative of several and the phase diagram aberration Y-direction；Reduction unit 11066, for the magnitude image, the X to it is inclined Derivative, the Y-direction partial derivative analyzed and restored, according to analysis and reduction result determine the amplitude also original image, institute Phase X is stated to local derviation also original image and the phase Y-direction local derviation also original image.

In any of the above-described technical solution, it is preferable that the processing unit 11062 is specifically used for, will other described images Phase and the basic phase carry out subtraction, and phase unwrapping is carried out around processing, according to phase unwrapping to calculated result The phase diagram aberration is determined around processing result.

Figure 12 shows the defect management of the valuable bills of one embodiment of embodiment according to the ninth aspect of the invention The schematic block diagram of system.

As shown in figure 12, the fault management system 1200 of the valuable bills of embodiment according to the present invention, comprising: generate single Member 1202, taxon 1204 and processing unit 1206.

Wherein, generation unit 1202 are generated for the defect characteristic according to valuable bills each in valuable bills training set First order child node；Taxon 1204, for calculating the partition value of each first order child node, and according to the segmentation Value classifies to the defect characteristic；Processing unit 1206, for judging whether all first order child nodes can not be again Classification, when the judgment result is yes, classification are completed.

Wherein it is preferred to which the taxon 1204 is specifically used for: if the partition value is less than given threshold, using Support vector machine method classifies to the defect characteristic, and if the partition value be more than or equal to the given threshold, Classified using traditional decision-tree to the defect characteristic.

In any of the above-described technical solution, it is preferable that the generation unit 1202 is specifically used for: calculating the valuable ticket According to the information gain value between the defect characteristic of valuable bills each in training set and the defect characteristic of other valuable bills, and According to each valuable bills corresponding information gain value construction first order child node.

In any of the above-described technical solution, it is preferable that further include: computing unit 1208, for calculating each valuable ticket According to residual point to defect mass center Euclidean distance；Unit 1210 is deleted, for the residual point in any valuable bills to defect mass center Euclidean distance when being greater than or equal to pre-determined distance, delete the institute that generates according to the defect characteristic of any valuable bills State first order child node.

Figure 13 shows the quality testing of the valuable bills of one embodiment of embodiment according to the tenth aspect of the invention The flow diagram of method.

As shown in figure 13, the quality determining method of valuable bills according to an embodiment of the invention, comprising:

Step 1302, the image letter that the cleaning-sorting machine in the secondary checking system is detected as underproof valuable bills is obtained Breath；

Step 1304, according to the status data of the secondary verification node recorded in the secondary checking system and distribution mould The image information of the underproof valuable bills is distributed to the secondary verification node by formula；

Step 1306, the system configuration information for obtaining the secondary checking system, according to the system configuration information to institute Secondary verification node is stated to be configured；

Step 1308, the batch information for obtaining the underproof valuable bills, according to batch information triggering Secondary verification node starts image analysis work or terminates image analysis work；

Step 1310, the kind information of the underproof valuable bills is obtained, the secondary verification node switching is triggered To image detection template corresponding with the kind information.

As shown in figure 14, the quality determining method of valuable bills according to another embodiment of the invention, comprising:

Step 1402, it acquires the useless image data of a machine and (obtains the figure that cleaning-sorting machine is detected as underproof valuable bills As information), and store the useless image data of a machine；

Step 1404, according to the secondary status data and allocation model for verifying node, by the useless image data distribution of a machine To secondary verification node；

Step 1406, the secondary secondary verification result for verifying node and batch information are associated storage, to be converged Total data；

Step 1408, secondary qualified crown word number inventory, i.e., qualification after secondary verifications node is verified are exported in data from summarizing Valuable bills image information；

Step 1410, secondary verification result and the useless image data of a machine are merged, to obtain among secondary treatment Result data is prepared for the secondary number of choosing of cleaning-sorting machine.

Figure 15 shows the structure of the secondary checking system of one embodiment of embodiment according to the tenth aspect of the invention Schematic diagram.

As shown in figure 15, secondary checking system 1500 according to an embodiment of the invention includes: quality detecting system 1502 (quality detecting systems for being equivalent to the valuable bills in the embodiment shown in Figure 15), secondary verification node 1504 are comprehensive Decision-making system 1506, cleaning-sorting machine 1508, product quality data center (data report center) 1510 and invalidated ticket crown word number amended record System 1512.Quality detecting system 1502 obtains the image information that cleaning-sorting machine 1508 is detected as underproof valuable bills, and will The image information of underproof valuable bills distributes to secondary verification node 1504, and secondary verification node 1504 is examined to from quality The image informations of the underproof valuable bills of examining system 1502 carries out secondary verification, and by the image of underproof valuable bills The crown word number of information and corresponding secondary verification result are sent to product quality data center 1510 and are stored, meanwhile, by two It is secondary to verify the image information hat corresponding with the image information of the underproof valuable bills that result is underproof valuable bills Font size is sent to comprehensive judgement system 1506.

Comprehensive judgement system 1506 carries out real time parsing to the image information of the underproof valuable bills received, and mentions The security characteristic information in the image information of underproof valuable bills is taken, security characteristic information and default characteristic information are carried out Relatively and show comparison result, and according to the decision instruction received judge underproof valuable bills image information whether Qualification, and the crown word number of the image information of underproof valuable bills is sent to product quality data with corresponding judgement result Center 1510.

Product quality data center 1510 is used to integrate secondary verification node 404 and comprehensive judgement system 1506 uploads respectively Underproof valuable bills image information verification as a result, i.e. that comprehensive judgement subsystem 1506 is corresponding to any crown word number Underproof valuable bills the judgement result made of image information replace verification that secondary verification node 1504 is made as a result, Invalidated ticket crown word number amended record subsystem 1512 carries out amended record to underproof valuable bills are determined as.

In addition, quality detecting system 1502 can be in several ways (for example, dynamic network passively receives mode and static state File active reading manner) obtain the useless image data of a machine (the i.e. image letter of the underproof valuable bills of cleaning-sorting machine output Breath), meanwhile, quality detecting system 1502 can also use network dynamic connection type, to obtain preceding process related information in real time, For example, determining whether secondary verification node 1504 starts image according to the batch information of the underproof valuable bills got It analyzes work and terminates the continuous switching between image analysis work, and continuous different batches of product；Or according to getting Template replacement information, for notifying whether secondary verification node 1504 needs more new template.Quality detecting system 1502 also props up Automatically accessing and exiting for floating node (for example, secondary verifying node 1504, comprehensive judgement system 1506) is held, and is supported not The floating access of same type node (for example, secondary verify node 1504 and comprehensive judgement system 1506).Quality detecting system 1502 also support the multi-mode of system configuration to import, for example, multi-mode importing, which can be Local or Remote static file, imports mould Formula is also possible to database introduction model.

When the image information of underproof valuable bills is after 1502 inter-process of quality detecting system, it is available with Lower output data:

1, secondary useless image data file (deposit) complete (containing real useless and critical useless)；

2, two subcritical useless description information, in a manner of data-base recording, storage is in the database；

3, secondary qualified crown word number inventory, qualified information (the i.e. qualified valuable ticket generated including secondary verification node 404 According to image information) and comprehensive judgement system 406 generate qualified information, in a manner of data-base recording, storage in the database；

4, the tracking information of secondary checking system 1500, is saved with file mode.

Quality detecting system 1502 can be according to the system configuration information got, to the inside of quality detecting system 1502 It is configured and is arranged；Secondary verification node 1504 is triggered according to the batch information of the underproof valuable bills acquired to start Image analysis work terminates image analysis work, according to the kind information of the underproof valuable bills got, triggering two Secondary verification node 1504 is switched to image detection template corresponding with the kind information.

Quality detecting system 1502 also supports the floating node of multiple and different types, and system is supported in the form of floating node The system access of (such as comprehensive judgement system 1506) and exiting for system, quality detecting system 1502 (are not conformed to by total task number The sum of the valuable bills of lattice) and system hardware performance, entrance connection and flow are controlled.

Quality detecting system 1502 realizes that task is distributed automatically to floating node, and dynamic records the place of different floating nodes Reason ability and characteristic, the disposed of in its entirety speed of dynamic analysis difference floating node, dynamic analysis different task are saved by different floatings Point disposition (as processing the time, number of repetition), to individual (numbers of i.e. underproof valuable bills) processing time-out into Row task is redistributed, super to individual processing secondary to carry out (more than the number redistributed) lagging staying disk.

Quality detecting system 1502 supports 3 kinds of allocation models, i.e. allocation model (i.e. the first allocation model) at first, average to appoint Business allocation model (i.e. the second allocation model) and the preferential allocation model of ability (i.e. third allocation model), wherein distribute mould at first Formula: being having time sequence between multiple connections in more floating node dynamic floating connections, does not have in each floating node task In the case where fully loaded, the Connection Time is preceding preferentially to be distributed, until this floating node task is full, then when selection connection Between behind nearest connection carry out task distribution, and so on；Average task allocation model: quality detecting system 1502 records The entirety ability of each floating node carries out task in the case where ensuring that each floating node processing total amount is average Distribution；The preferential allocation model of ability: quality detecting system 1502 obtains the processing capacity of each floating node by data statistics (such as processing speed) generates a newest floating node processing speed ranking list, and quality detecting system 1502 is according to the seniority among brothers and sisters List, it is preferential that speed is selected to handle fast floating node progress task distribution.

Quality detecting system 1502 can receive automatically verifies transmit after node 1504 is handled two through different secondary It is secondary verify as a result, by it is secondary verification result extract and cleaning-sorting machine 1508 export a machine give up image data merge, So that including complete raw image data, single treatment intermediate result data, in secondary treatment in the image file finally saved Between result data and secondary treatment final result data.Wherein, secondary treatment intermediate result data can come from different secondary Node 1504 is verified, can be from comprehensive judgement system 1506.The secondary treatment intermediate result data is with the side of data record Formula is association with batch information, database is written, prepares for the secondary number of choosing of cleaning-sorting machine.In addition, the secondary useless processing data of analysis (image informations of underproof valuable bills in the i.e. secondary secondary verification result for verifying node 1504), to secondary useless processing number Secondary useless relevant information in extracts, and in a manner of data record, is association with batch information, database is written, is The verification of comprehensive judgement system 1506 is prepared.

As shown in figure 16, the quality detecting system 1600 of valuable bills according to an embodiment of the invention, comprising: draw Sub-unit 1602, cluster cell 1604, unit 1606 and detection unit 1608.

Wherein, division unit 1602, for the sample set of all valuable bills to be divided into multiple detection zones；Cluster is single Member 1604, for the corresponding sample set of each detection zone in the multiple detection zone to be clustered according to feature, with The corresponding sample set of each detection zone is divided into multiple classifications；Unit 1606 is used for each detection zone The sample set of each classification in domain in multiple classifications learns the parameter space of corresponding detection zone, to obtain each detection zone The corresponding parameter space in domain；Detection unit 1608, for using each parameter space to the sample set in corresponding detection zone Carry out quality testing.

In the above-mentioned technical solutions, it is preferable that the cluster cell 1604 includes: setting unit 16042, and being used for will be described Initial value of any sample as a cluster centre in each detection zone；Computing unit 16044, for calculating the inspection Survey the first Euclidean distance in region between other samples and any sample；Confirmation unit 16046, for described first It, otherwise, will other described samples using other described samples cluster centre new as one when Euclidean distance is greater than pre-determined distance As the cluster centered on any sample.

As shown in figure 17, the image analysis system 1700 of the valuable bills of embodiment according to the present invention, comprising: detection is single Member 1702, judging unit 1704 and processing unit 1706.

Wherein, detection unit 1702, for being carried out according to multiple Informations of the preset order to the image of valuable bills Detection, wherein the Information includes positive information, back side information, perspective information and/or infrared information；Judging unit 1704, it is described more for being determined as that described image is unqualified when detecting that any Information is unqualified, and detecting When each Information is qualified in a Information, it is determined as described image qualification；Processing unit 1706, for it is described not The type of error of qualified images and wrong process are analyzed and are recorded, the table images that do not conform to are managed and be counted.

In the above-mentioned technical solutions, it is preferable that include: allocation unit 1708, for the number when the image for needing to detect When being multiple, described multiple images are distributed to multiple and different detection nodes, to use the plurality of different detection simultaneously Node detects described multiple images.

In any of the above-described technical solution, it is preferable that further include: acquiring unit 1710, it is described not for acquisition respectively The crown word number information of qualified images the and described qualified images；Adding unit 1712, for by the prefix for not conforming to table images Number information is added in blacklist, and the crown word number information of the qualified images is added in white list, and does not conform to described The crown word number information for meeting specified requirements in the crown word number information of in the crown word number information of table images the or described qualified images It is added in gray list.

In any of the above-described technical solution, it is preferable that further include: setting unit 1714, for according to described unqualified Image type of error and/or wrong process, the preset order is set.

The technical scheme of the present invention has been explained in detail above with reference to the attached drawings, and the present invention proposes the comprehensive of the new valuable bills of one kind Close decision technology, it is possible to reduce number of samples, calculation amount and the time of feature extraction, the core of secondary checking system also can be improved Efficiency is looked into, the accuracy rate of verification is improved, reduces and verifies cost, nicety of grading is improved, improves the robustness of quality detecting system.

The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field For art personnel, the invention may be variously modified and varied.All within the spirits and principles of the present invention, made any to repair Change, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.

Claims

1. a kind of defect management method of valuable bills characterized by comprising

The sample image of valuable bills is obtained, and generates the template figure for carrying the residual point of those suspected defects according to the sample image Picture；

Calculate position and the geometric center of the mass center of the geometric figure of the residual point composition of multiple those suspected defects in the template image Position；

According to the position of the position of the mass center and the geometric center, judge any doubtful in the residual point of the multiple those suspected defects It whether is the residual point of defect like the residual point of defect.

2. the defect management method of valuable bills according to claim 1, which is characterized in that calculate in the template image Multiple those suspected defects it is residual point composition geometric figure mass center position and geometric center position, according to the mass center The position of position and the geometric center judges whether the residual point of any those suspected defects in the residual point of the multiple those suspected defects is scarce The step of falling at residual include:

Calculate the position of the first mass center of the geometric figure of the residual point composition of the multiple those suspected defects and the position of the first geometric center It sets, and calculates the second mass center of the geometric figure of the residual point composition of other those suspected defects in addition to any residual point of those suspected defects Position and the second geometric center position；

Judge whether the distance between the position of first mass center and the position of second mass center are greater than first threshold, and institute State whether the distance between the position of the first geometric center and the position of second geometric center are less than second threshold；

When the position of first mass center and the distance between the position of second mass center are greater than the first threshold, and it is described When the distance between the position of first geometric center and the position of second geometric center are less than the second threshold, it is determined that The residual point of any those suspected defects is not the residual point of the defect；

Otherwise, the residual point of any those suspected defects is the residual point of the defect.

3. the defect management method of valuable bills according to claim 1 or 2, which is characterized in that described according to the sample The step of generation of this image carries the template image of the residual point of those suspected defects, specifically includes:

Weight matrix is calculated according to the point of proximity of each sample point in the sample image；

The template image for carrying the residual point of the those suspected defects is generated according to the weight matrix.

4. the defect management method of valuable bills according to claim 3, described according to each of described sample image The point of proximity of sample point calculates the step of weight matrix, specifically includes:

It is calculated by the following formula the weight matrix:

Wherein, ε (w) indicates error amount, X_iIndicate any sample point, X_j(j=1,2 ..., k) indicates the k of any sample point A point of proximity, w_ijIndicate the weight matrix between any sample point and the point of proximity.

5. the defect management method of valuable bills according to claim 4, which is characterized in that the weight matrix it is each It is capable and be 1.

6. a kind of fault management system of valuable bills characterized by comprising

Generation unit carries those suspected defects for obtaining the sample image of valuable bills, and according to sample image generation The template image of residual point；

First computing unit, the mass center of the geometric figure for calculating the residual point composition of multiple those suspected defects in the template image Position and geometric center position；

Judging unit, for judging the multiple those suspected defects according to the position of the mass center and the position of the geometric center Whether the residual point of any those suspected defects in residual point is the residual point of defect.

7. the fault management system of valuable bills according to claim 6, which is characterized in that the first computing unit tool Body is used for,

The judging unit is specifically used for, and judges the position of first mass center and the distance between the position of second mass center Whether first threshold is greater than, and the position of first geometric center and the distance between the position of second geometric center are It is no to be less than second threshold, when the distance between the position of first mass center and the position of second mass center are greater than described first Threshold value, and the distance between the position of first geometric center and the position of second geometric center are less than second threshold When value, it is determined that any residual point of those suspected defects is not the residual point of the defect；Otherwise, the residual point of any those suspected defects is institute State the residual point of defect.

8. the fault management system of valuable bills according to claim 6 or 7, which is characterized in that the generation unit packet It includes:

Second computing unit, for calculating weight matrix according to the point of proximity of each sample point in the sample image；

The generation unit is specifically used for, and the template for carrying the residual point of the those suspected defects is generated according to the weight matrix Image.

9. the fault management system of valuable bills according to claim 8, second computing unit is specifically used for,

It is calculated by the following formula the weight matrix:

10. the fault management system of valuable bills according to claim 9, which is characterized in that the weight matrix it is every A line and be 1.