CN101373516B - Method and system for analyzing and evaluating processability of finger print acquisition image library - Google Patents

Method and system for analyzing and evaluating processability of finger print acquisition image library Download PDF

Info

Publication number
CN101373516B
CN101373516B CN2008102221677A CN200810222167A CN101373516B CN 101373516 B CN101373516 B CN 101373516B CN 2008102221677 A CN2008102221677 A CN 2008102221677A CN 200810222167 A CN200810222167 A CN 200810222167A CN 101373516 B CN101373516 B CN 101373516B
Authority
CN
China
Prior art keywords
handlability
fingerprint
fingerprint image
absolute
algorithm
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CN2008102221677A
Other languages
Chinese (zh)
Other versions
CN101373516A (en
Inventor
李扬渊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
CHENGDU FINCHOS ELECTRON Co Ltd
Original Assignee
CHENGDU FINCHOS ELECTRON Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by CHENGDU FINCHOS ELECTRON Co Ltd filed Critical CHENGDU FINCHOS ELECTRON Co Ltd
Priority to CN2008102221677A priority Critical patent/CN101373516B/en
Publication of CN101373516A publication Critical patent/CN101373516A/en
Application granted granted Critical
Publication of CN101373516B publication Critical patent/CN101373516B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Collating Specific Patterns (AREA)
  • Image Input (AREA)

Abstract

The invention relates to a treatable analysis and assessment method and a system of a fingerprint collection image base. The system comprises a fingerprint sensor which is connected with the fingerprint image base; the fingerprint image base which is respectively connected with an absolute treatable analysis module, a relative treatable analysis module and a fingerprint image recognition algorithm module; the fingerprint image recognition algorithm module which is further connected with the relative treatable analysis module; the absolute treatable analysis module which is used for analyzing the fingerprint image base and outputting the absolute treatability of the fingerprint image base; the relative treatable analysis module which is used for analyzing the fingerprint image base and the recognition performance evaluation of the fingerprint image base obtained by the fingerprint image recognition algorithm and outputting the relative treatability of the fingerprint image base. The method and the system integrate the algorithm performance evaluation of the different fingerprint bases which are taken as a test set, thereby obtaining the in-depth and thorough evaluation of the algorithm performance; and the fingerprint base is assessed by the relative treatability, thereby having more direct significance for the construction of a fingerprint recognition system.

Description

The handlability analyzing evaluation method and the system of fingerprint collecting image library
Technical field
The present invention relates to the analysis and evaluation field in fingerprint image storehouse, relate in particular to the handlability analyzing evaluation method and the system of fingerprint collecting image library.
Background technology
Bio-identification is as Secure Application, because its ease for use and unforgeable to a certain extent have vast market prospect.Fingerprint collecting is convenient, and what can be used for discerning contains much information, and has great using value.Fingerprint recognition system mainly is made of fingerprint collecting sensor and Fingerprint Image Recognition Algorithms, and the former obtains digital picture from fingerprint, and the latter carries out fingerprint characteristic according to image and extracts and discern, to realize the function of identification finger.Huge industrial prospect has promoted the development of fingerprint collecting sensor on the one hand, promotes the research and development at the finger print image recognizer of fingerprint collecting image on the other hand.As the building block of fingerprint recognition system, both are conditional and interdependent, and the image that sensor should be suitable for algorithm process with collection is a target, and algorithm must be optimized at the problem that sensor characteristic and gatherer process cause simultaneously.But both research and developments separately are also more isolated at present, are difficult to mutual fusion, and wherein most important reason is exactly the appraisal procedure that lacks the fingerprint collecting image of generally acknowledging, to point out both design responsibilities separately.
As a kind of typical pattern clustering problem, the Fingerprint Image Recognition Algorithms of different principle, system, realization constraint, the check of its performance is all based on test set, i.e. the fingerprint image storehouse.Is the image tagged that belongs to identical finger in the fingerprint base same group, compares with recognition result, just can be sincere, know false rate and add up to refusing, thus algorithm performance is obtained roughly understanding.But different algorithms respectively has quality corresponding to different image problems and image to problem, does not separate the influence that these factors cause separately, just can't estimate the performance of Fingerprint Image Recognition Algorithms accurately.The reason that this just is being to use the instability of the algorithm evaluation result of different fingerprint bases generations to produce.Fingerprint base is carried out analysis and evaluation, pointed out the mode that will merge based on the algorithm evaluation result that a large amount of fingerprint bases obtain: the algorithm performance evaluation that will once obtain is defined as the algorithm performance on the problem coordinate, the performance difference of the same algorithm performance of comprehensive different problem coordinate just can have comprehensively algorithm performance and estimate accurately.
On the other hand, the Performance evaluation criterion of effective fingerprint collecting sensor is not arranged as yet at present, often adopt simple image property statistics, but this is not equivalent to and estimates this sensor and how be worth when forming fingerprint recognition system.Based on understanding to the relation of fingerprint base assessment and Fingerprint Image Recognition Algorithms performance evaluation, the great amount of images of being gathered with certain sensor is with respect to the handlability of Fingerprint Image Recognition Algorithms, the evaluation of the performance that shows when forming fingerprint recognition system as this sensor.Be shielding gatherer process and the finger characteristic different uncertain factors that cause that distribute, the inconsistent factor that only causes owing to gatherer process difference that need have nothing to do by Flame Image Process and screening normalization and sensor is with the correct evaluation of realization to sensor performance.
Above function was based on necessary module of the present invention provided, the assessment that the fingerprint base handlability is carried out, and finish by expansion of the present invention.
Be the difference on avoiding understanding, to important term definition be described below:
Fingerprint recognition system: by gathering finger print, discern this fingerprint image, to reach this finger of identification and electronic system that should the biology people.Mainly form by fingerprint collecting, fingerprint image identification two parts function.
Fingerprint collecting sensor: if no special declaration abbreviates " sensor " as.According to differences of physical properties such as finger skin electricity, magnetic, sound, light, heat, pressures, from pointing direct collection to obtain to have the digital picture of fingerprint texture information.
Finger is to the suitability of sensor: if no special declaration abbreviates " suitability " as.Because the difference of the physical property difference degree of different finger skins, different fingers, even the different times of same finger, collection effect has nothing in common with each other when using the sensor acquisition of different principle.
Face sensor: the sensor of gathering whole finger print simultaneously.Except that acoustic sensor, all gathering with the panel contact is feature, so be called face sensor.
Line sensor: sensor array is classified the rectangle of width much smaller than length as, finger needs contact and broad ways to move during collection, and the image mosaic that will gather one by one with image split-joint method is the summation of sensor, acquisition method and the image mosaic process of finger print image.Why the image mosaic process being thought the part of line sensor, is in order to follow the definition of sensor from the conversion of pointing fingerprint image.
Fingerprint collecting image: if no special declaration abbreviates " image " as.Finger is pointed the digitizing gray level image of true finger print information by the fingerprint collecting sensor from pointing the reflection that directly collects.
Display foreground: if no special declaration abbreviates " prospect " as.Collect the formed image local of Sensor section of finger print information, reflected finger print information.
Image background: if no special declaration abbreviates " background " as.Do not collect the image local of the Sensor section formation of finger print information.
Image problem:, and, be referred to as image problem because of factors such as suitability, gatherer process exposure level, fingers deformed cause the distortion and the loss of finger print information because of finger pollutes or the noise of factors introducings such as breakage, defect sensor, digital quantization process.
Fingerprint collecting image library: if no special declaration abbreviates " fingerprint base " as.The set of image and mark thereof: this mark is classified as one group with the images acquired of identical finger, and the images acquired that difference is pointed is included into not on the same group, to discern, to refuse to discern and miss the basis for estimation of identification as Fingerprint Image Recognition Algorithms.
The fingerprint base group: one group of related fingerprint base, such as, gather certainly with a sensor.
Picture overlapping zone: if no special declaration abbreviates " overlapping region " as.Collection is from the different images of same finger, the zone corresponding to identical finger surface that its prospect part comprises jointly.
Nonlinear deformation: gather different images, be stressed and shear force causes by finger surface, with the irrelevant relative deformation of rotation translation from same finger.
Image is to problem: the mutual character that the image in the fingerprint base is right influences that algorithm effectively discerns.Mainly comprise overlapping region informational capacity size and nonlinear deformation degree, and the similarity degree of the fingerprint image of different finger collections.
Absolute handlability: various known image problems and image are to the summation of problem.
Absolute handlability space: various image problems and image are made the long-pending Euclidean space that obtains of Euclidean to the degree of problem.
Fingerprint Image Recognition Algorithms: if no special declaration abbreviates " algorithm " as.Refer to a computation process, it is right that it is input as fingerprint image, and be output as the logic determines of " same finger " or " different finger ".
Identification: the different images judgement that algorithm will pick up from same finger is " same finger ".
Refuse identification: the different images judgement that algorithm will pick up from same finger is " different finger ".
Mistake identification: the image judgement that algorithm will pick up from different fingers is " same finger ".
Refuse sincere: refuse to discern number of times and account for the ratio of same finger different images to the input number.
Know false rate: mistake is discerned number of times and is accounted for the ratio of different finger-image to the input number.
Algorithm performance is estimated: to as input, what obtain refuses sincere and the false rate level of knowledge with the image in the fingerprint base.Obtain refusing sincere and the false rate of knowledge if algorithm carries out hard threshold judgement to the similarity degree analysis result, be presented as both curves so about thresholding.
Relative handlability: the fingerprint in fingerprint base embodies the part that the target fingerprint recognition system requires during the algorithm performance that obtains as certain algorithm input is estimated.
Summary of the invention
In order to solve above-mentioned technical matters, provide a kind of handlability analyzing evaluation method and system of fingerprint base.Its purpose is, adapts to the performance evaluation demand of fingerprint collecting sensor and algorithm for recognizing fingerprint, promotes the global design of development of sensor and algorithm research and fingerprint recognition system.
The invention provides the handlability analysis and evaluation system of fingerprint collecting image library, comprising:
The fingerprint image storehouse, fingerprint collecting sensor, absolute handlability analysis module, handlability analysis module, and Fingerprint Image Recognition Algorithms module relatively;
The fingerprint collecting sensor is connected with the fingerprint image storehouse;
The fingerprint image storehouse also is connected with the Fingerprint Image Recognition Algorithms module with absolute handlability analysis module, relative handlability analysis module respectively;
The Fingerprint Image Recognition Algorithms module also connects with relative handlability analysis module;
Absolute handlability analysis module is used for the fingerprint image storehouse is analyzed, the absolute handlability in output fingerprint image storehouse;
The handlability analysis module is used for fingerprint image storehouse and fingerprint image storehouse are analyzed through the recognition performance evaluation that Fingerprint Image Recognition Algorithms obtains relatively, the relative handlability in output fingerprint image storehouse;
Described absolute handlability is the information distortion and the loss of fingerprint image, and fingerprint image between influence the summation that Fingerprint Image Recognition Algorithms is carried out the problem of fingerprint recognition;
Described relative handlability is during a fingerprint in the fingerprint image storehouse is estimated the algorithm performance that obtains as certain Fingerprint Image Recognition Algorithms input, to embody the part that the target fingerprint recognition system requires.
Also comprise with the comprehensive module of sensor with the algorithm synthesis module;
With the algorithm synthesis module, be connected with absolute handlability analysis module with relative handlability analysis module respectively, being used for absolute handlability is that functional value obtains the algorithm performance evaluation function as the coordinate base, with relative handlability;
With the comprehensive module of sensor, connect with relative handlability analysis module, absolute handlability analysis module and with the algorithm synthesis module respectively, be used to export and be included under the absolute handlability coordinate the absolute handlability distribution density function that the absolute handlability distribution density of statistics obtains, the sensor performance evaluation function of the relative handlability distribution density function that obtains through the mapping of algorithm performance evaluation function with absolute handlability distribution density function;
Described algorithm is a Fingerprint Image Recognition Algorithms; Described sensor is the fingerprint collecting sensor.
Absolute handlability analysis module comprises:
Global statistics property analysis module is used to obtain pickup area size, resolution, resolution consistance and the gray shade scale of fingerprint image;
Partial statistics property analysis module is used to obtain the prospect background separability of fingerprint image, foreground area, and the prospect gray-scale statistical distributes, the empty statistical distribution frequently of prospect contrast statistical distribution and prospect;
The image problem analysis module is used for separating the foreground image that obtains according to prospect background and calculates prospect local frequencies intensity, and separates the level and smooth degree distribution that the foreground image that obtains calculates the crestal line field of direction according to prospect background;
Image is used for calculating the overlapping region quantity of information according to foreground image and distributes and the nonlinear deformation degree distribution the case study module; Described overlapping region refers to gather the different images from same finger, the zone corresponding to identical finger surface that its prospect part comprises jointly.
The handlability analysis module comprises relatively:
The testing evaluation module is used for based on the fingerprint image storehouse algorithm for recognizing fingerprint performance being carried out testing evaluation;
The assessment result extraction module is used for meeting pre-conditioned assessment result from the assessment result extraction.
The invention provides the handlability analyzing evaluation method of fingerprint collecting image library, comprising:
Step 51 is analyzed the fingerprint image storehouse, the absolute handlability in output fingerprint image storehouse;
Step 52 is analyzed through the recognition performance evaluation that Fingerprint Image Recognition Algorithms obtains fingerprint image storehouse and fingerprint image storehouse, the relative handlability in output fingerprint image storehouse;
Described absolute handlability is distortion of fingerprint image self-information and loss, and fingerprint image between influence the summation that Fingerprint Image Recognition Algorithms is carried out the problem of fingerprint recognition;
Described relative handlability is during a fingerprint in the fingerprint image storehouse is estimated the algorithm performance that obtains as certain Fingerprint Image Recognition Algorithms input, to embody the part that the target fingerprint recognition system requires.
Also comprise:
Step 53 is that functional value obtains the algorithm performance evaluation function with absolute handlability as the coordinate base, with relative handlability;
Step 54, the absolute handlability distribution density function that obtains with the absolute handlability distribution density of statistics under absolute handlability coordinate, with the relative handlability distribution density function that absolute handlability distribution density function is obtained through the mapping of algorithm performance evaluation function as the sensor performance evaluation function;
Described algorithm is a Fingerprint Image Recognition Algorithms; Described sensor is the fingerprint collecting sensor.
Step 51 comprises:
Step 71, pickup area size, resolution, resolution consistance and the gray shade scale of acquisition fingerprint image;
Step 72, the prospect background separability of acquisition fingerprint image, foreground area, the prospect gray-scale statistical distributes, the empty statistical distribution frequently of prospect contrast statistical distribution and prospect;
Step 73 is separated the foreground image that obtains according to prospect background and is calculated prospect local frequencies intensity, and separates the level and smooth degree distribution that the foreground image that obtains calculates the crestal line field of direction according to prospect background;
Step 74 is calculated the overlapping region quantity of information according to foreground image and is distributed and the nonlinear deformation degree distribution; Described overlapping region refers to gather the different images from same finger, the zone corresponding to identical finger surface that its prospect part comprises jointly.
Step 52 comprises:
Step 81 is carried out testing evaluation based on the fingerprint image storehouse to the algorithm for recognizing fingerprint performance;
Step 82, extraction meets pre-conditioned assessment result from assessment result.
When the present invention has avoided using proprietary fingerprint base evaluation algorithms performance, because gather the evaluation result instability that crowd's singularity and sensor characteristic cause image problem and image that the skewed popularity of problem is caused, by analyzing this evaluation result of fingerprint base correction itself.
The present invention proposes the feasible road that different fingerprint bases are merged as the algorithm performance evaluation of test set, obtain the absolute handlability coordinate of each fingerprint base by analysis, mapping is set up in algorithm performance evaluation and absolute handlability space, thereby obtain deeply comprehensively estimating algorithm performance.
The present invention assesses fingerprint base with relative handlability, and this is equivalent to evaluation and gathers the performance of the sensor of this fingerprint base with respect to target algorithm.Compare directly image is carried out the simple analysis statistics, to the more direct meaning of being built with of fingerprint recognition system.
The present invention adopts the system architecture of open classification, can receive the latest developments of algorithm research and sensor research on the one hand, make evaluating system remain availability, promote more individual, enterprise, mechanism to participate in the fingerprint base Study on Method of Accessing on the other hand.
Description of drawings
Fig. 1 is the handlability analysis and evaluation system flowchart of fingerprint collecting image library provided by the invention.
Fig. 2 is fingerprint base statistical property analysis principle figure provided by the invention.
Fig. 3 is fingerprint base image problem analysis principle figure provided by the invention.
Fig. 4 is that fingerprint base image provided by the invention is to the case study schematic diagram.
Fig. 5 is relative handlability analysis principle figure provided by the invention.
Fig. 6 is a Fingerprint Image Recognition Algorithms performance evaluation schematic diagram provided by the invention.
Fig. 7 is fingerprint collecting sensor performance analysis principle figure provided by the invention.
Embodiment
The present invention is by fingerprint base being analyzed and the handlability assessment, to reach the purpose of estimating Fingerprint Image Recognition Algorithms, living body finger print acquisition sensor and the living body finger print recognition system performance that they were constituted.
As shown in Figure 1.Necessary part for native system in the solid wire frame is an analysis part, comprise absolute handlability analysis module and relative handlability analysis module, for the native system expansion is comprehensive part, comprise in the thick dashed line frame with the algorithm synthesis module with the comprehensive module of sensor.Workflow is:
Step 1, the fingerprint image storehouse is by fingerprint collecting sensor acquisition fingerprint.
Step 2, fingerprint image storehouse are the input of absolute handlability analysis module, and absolute handlability analysis module is output as the absolute handlability in print image storehouse.
Step 3, fingerprint image storehouse and fingerprint image storehouse are the inputs of relative handlability analysis module through the recognition performance evaluation that Fingerprint Image Recognition Algorithms obtains, and the handlability analysis module is output as the relative handlability in fingerprint image storehouse relatively.
Step 4, absolute handlability, relative handlability are output as the algorithm performance evaluation as the input with the algorithm synthesis module.
Step 5, absolute handlability, relative handlability and algorithm performance evaluation function are output as the sensor performance evaluation as the input with the comprehensive module of sensor.
Absolute handlability analysis module is analyzed the intrinsic factor that influences the Fingerprint Image Recognition Algorithms performance that has nothing to do with specific algorithm of fingerprint base itself, and its treatment step is as follows:
Step a, analyze global statistics character:
Comprise pickup area size, resolution, resolution consistance, gray shade scale.The image overall statistical property that face formula collector is gathered is by the decision of sensor intrinsic property, and the image of line formula sensor acquisition is pointed scraping length, direction, exposure level changes and the influence of joining method, need add up.
Step b, analyze partial statistics character:
Comprise display foreground background separability, foreground area, the prospect gray-scale statistical distributes, prospect contrast statistical distribution, the empty statistical distribution frequently of prospect.The prospect background separability is exactly the difference degree of partial statistics character separately; Foreground area refers to the total area of effective fingerprint image; The gray-scale statistical distribution that the prospect gray-scale statistical distributes and refers to effective fingerprint image zone; Prospect contrast statistical distribution refers to that effective fingerprint image zone is with rational yardstick piecemeal, the distribution of piece contrast; The empty statistical distribution frequently of prospect refers to effective fingerprint image zone with rational yardstick piecemeal, the distribution of piece frequency distribution in short-term.
Step c, the analysis image problem:
Image problem that effect characteristics extracts has been represented the proportionate relationship of finger print information and noise in the fingerprint image to comprise prospect local frequencies intensity, the level and smooth degree distribution of the crestal line field of direction.Prospect local frequencies intensity obtains from the empty statistical distribution analysis frequently of prospect; From fingerprint image extract the crestal line field of direction and level and smooth it, the difference before and after level and smooth is exactly the level and smooth degree of the crestal line field of direction.
Steps d, analysis image are to problem:
The problem of the correlativity that the image of effect characteristics coupling is right comprises that the overlapping region quantity of information distributes the nonlinear deformation degree distribution.Carry out the fingerprint image coupling with artificial or software, matching result has just shown the overlapping region and the relative nonlinear distortion of two width of cloth images.The validity feature quality and quantity decision that the overlapping region quantity of information is comprised by overlapping region area and overlapping region.
The handlability analysis module is analyzed fingerprint base reaches certain specific algorithm on the specified performance index level relatively, reaction fingerprint recognition system performance in a certain respect.Realize dividing two steps:
Step I carries out testing evaluation based on the fingerprint image storehouse to the algorithm for recognizing fingerprint performance.
Step II, from assessment result, extract be equivalent to index that fingerprint recognition system requires the assessment result form.
With the algorithm synthesis module, the absolute handlability assessment of different fingerprint bases is assessed with the relative handlability with respect to same algorithm, comprehensive for being defined in the algorithm performance evaluation function in the absolute handlability space, make up the performance that is reached with the reflection algorithm in the face of different absolute handlability indexs.It realizes dividing two steps:
Step 1), with 4 aspects of the absolute handlability of fingerprint base as vectorial independent variable, with fingerprint base with respect to the relative handlability of certain algorithm as functional value, set up the discrete function that is defined in the absolute handlability space.
Step 2), the discrete function match being become continuous function, is exactly the performance evaluation function of algorithm.It has predicted when the input fingerprint base has certain specific absolute handlability combination, be the performance that the test set algorithm can reach.
With the comprehensive module of sensor, gather from estimate the value of this sensor with the fingerprint base of a sensor as the input of living body finger print recognition system fingerprint image acquisition by estimating.It realizes dividing two steps:
Step (1) will be carried out absolute handlability analysis respectively with the fingerprint base group that the dissimilar fingers of a sensor acquisition obtain, to add up the absolute handlability distribution density function of this sensor acquisition fingerprint base.This distribution density function is exactly the absolute performance evaluation of sensor.Absolute handlability degree of sensor acquisition image and degree of stability have been reflected.
Step (2), if the absolute handlability distribution density function of sensor acquisition fingerprint base is known, certain algorithm performance evaluation function is known simultaneously, this absolute handlability distribution density function is exactly the relative performance evaluation of sensor through the relative handlability distribution function that the mapping of algorithm evaluation function obtains.Degree that systemic the subject of knowledge and the object of knowledge reaches and degree of stability have been reflected when this sensor and selected algorithm constitute the living body finger print recognition system.
Analysis part is divided into four modules.
1. statistical property analysis module:
As shown in Figure 2, press certain path combination by basic image processing method.Face sensor global statistics character is intrinsic property, by manual input.Line sensor global statistics character obtains through software statistics.Partial statistics character, earlier with the fingerprint image piecemeal, add up gray scale, contrast and the frequency distribution of each piecemeal, select the foundation of one or more fusion as the prospect background separation according to the fingerprint base characteristic, the difference degree of prospect background on this statistical property is as the assessment of separability.After isolating foreground blocks,, calculate foreground area, intensity profile, contrast distribution, empty frequency division cloth according to statistics before.
The prospect background separability has determined the accuracy of foreground extraction, can influence the order of accuarcy of all subsequent treatment on the principle.But in view of present fingerprint collecting sensor acquisition image generally can extract prospect quite accurately, so only when display foreground background separable degree in storehouse is low, just pay attention to.
2. image problem analysis module:
As shown in Figure 3.The analysis of prospect local frequencies intensity with the empty frequency division cloth of the prospect of statistical property analysis module output as input.The assessment of the level and smooth degree of the fingerprint image crestal line field of direction, need call two basic skills in the fingerprint images processing: the judgement of the crestal line field of direction and the crestal line field of direction are level and smooth.Obtain foreground image according to the prospect background separation from fingerprint image, judge the prospect crestal line field of direction, and the field of direction is carried out smoothly, the difference degree of the level and smooth front and back crestal line field of direction is exactly the level and smooth degree of the former fingerprint image crestal line field of direction.
The order of accuarcy of the level and smooth degree analyzing of the fingerprint image crestal line field of direction depends on that fully the crestal line field of direction of being called is judged and the performance of the level and smooth module of the crestal line field of direction.Because the diversity of fingerprint image disposal route, and be in the development, native system is not specified its implementation algorithm.
3. image is to the case study module:
As shown in Figure 4.Obtain foreground image according to the prospect background separation from fingerprint image, mate, just obtain the analysis of overlapping region area and nonlinear deformation degree by the fingerprint image matching algorithm.The fingerprint image coupling is the major part of fingerprint image identification, only uses the assessment of its output to image similarity when being used for discerning.In native system, use it to the judgement of overlapping region and the judgement of nonlinear deformation.
Image depends on the performance of fingerprint image matching algorithm in overlapping region judgement and nonlinear deformation judgement of being called fully to the order of accuarcy of problem assessment.Because the diversity of fingerprint image matching method, and be in the development, native system is not specified its implementation algorithm.
4. relative handlability analysis module:
As shown in Figure 5.Earlier, as test set, obtain the performance evaluation of certain algorithm with tested fingerprint base Fingerprint Image Recognition Algorithms is carried out the method for performance evaluation based on test set.This estimates desired all evaluation informations of conformance with standard method.According to the desired performance index of fingerprint recognition system, extract corresponding evaluating data, as relative handlability.
The standard method of test is with reference to current method of testings such as FVC2006, because this functions of modules is to analyze the relative handlability of fingerprint, so will not be concerned about relevant algorithm operational efficiency in the test result.The desired performance index of final system are different in response to using, when for example the living body finger print recognition system is as lockset, require index be generally ten thousand/(or 1,000,000/) know false rate correspondence refuse sincerely, then use and just should require to extract relative handlability with this at lockset.
Comprehensive part is divided into two modules
1. same algorithm synthesis:
As shown in Figure 6.In the absolute handlability, two image problems and two images to the coordinate base of problem as absolute handlability space are respectively: prospect local frequencies intensity, the level and smooth degree of the fingerprint image crestal line field of direction, overlapping region area distributions, nonlinear deformation degree distribution.And be the function that is defined in absolute handlability coordinate with relative handlability.So different fingerprint base just becomes a function on the discrete point that is defined on the absolute handlability space with respect to the handlability assessment of certain algorithm.This discrete function is carried out the match of higher-dimension function, and the continuous function that obtains is exactly the algorithm performance evaluation function that obtains with algorithm synthesis.The meaning of this fitting function evaluation algorithms performance is, quantificational description be relation between the resulting algorithm performance of test set is estimated with the absolute handlability of fingerprint base with the fingerprint base, measurable when the absolute handlability of fingerprint base, then the performance that reveals of algorithm table is measurable.
The match of higher-dimension function is a kind of classical way, and explicit function match is mainly used in the experimental formula match, and the piecewise function match then is the conventional method in the numerical evaluation.This method is optionally, so do not specify.
2. comprehensive with sensor:
As shown in Figure 7.Adding up a large amount of fingerprint bases that certain sensor gathers distributes by the absolute handlability coordinate that analysis part obtains, this distribution is exactly the absolute performance evaluation of this sensor, has pointed out image problem that this sensor causes in gatherer process and image degree and the stability to problem.If the performance evaluation function of certain algorithm is known, and covered the absolute performance evaluation of this sensor, this absolute handlability coordinate distribution is exactly the relative performance evaluation of this sensor with respect to this algorithm by the functional value distribution that the mapping of algorithm performance evaluation function obtains so, has represented the performance evaluation when the two is combined into fingerprint recognition system.
Because the method for expressing of algorithm performance evaluation function determine by selected fit procedure, and this method for expressing directly defined mapping process, so mapping process is not specified, and selects its corresponding Function Mapping method during by selection higher-dimension function approximating method.
Those skilled in the art can also carry out various modifications to above content under the condition that does not break away from the definite the spirit and scope of the present invention of claims.Therefore scope of the present invention is not limited in above explanation, but determine by the scope of claims.

Claims (8)

1. the handlability analysis and evaluation system of fingerprint collecting image library is characterized in that, comprising:
The fingerprint image storehouse, fingerprint collecting sensor, absolute handlability analysis module, handlability analysis module, and Fingerprint Image Recognition Algorithms module relatively;
The fingerprint collecting sensor is connected with the fingerprint image storehouse;
The fingerprint image storehouse also is connected with the Fingerprint Image Recognition Algorithms module with absolute handlability analysis module, relative handlability analysis module respectively;
The Fingerprint Image Recognition Algorithms module also connects with relative handlability analysis module;
Absolute handlability analysis module is used for the fingerprint image storehouse is analyzed, the absolute handlability in output fingerprint image storehouse;
The handlability analysis module is used for fingerprint image storehouse and fingerprint image storehouse are analyzed through the recognition performance evaluation that Fingerprint Image Recognition Algorithms obtains relatively, the relative handlability in output fingerprint image storehouse;
Described absolute handlability is the information distortion and the loss of fingerprint image, and fingerprint image between influence the summation that Fingerprint Image Recognition Algorithms is carried out the problem of fingerprint recognition;
Described relative handlability is during a fingerprint in the fingerprint image storehouse is estimated the algorithm performance that obtains as certain Fingerprint Image Recognition Algorithms input, to embody the part that the target fingerprint recognition system requires.
2. handlability analysis and evaluation as claimed in claim 1 system is characterized in that also comprising with the comprehensive module of sensor with the algorithm synthesis module;
With the algorithm synthesis module, be connected with absolute handlability analysis module with relative handlability analysis module respectively, being used for absolute handlability is that functional value obtains the algorithm performance evaluation function as the coordinate base, with relative handlability;
With the comprehensive module of sensor, connect with relative handlability analysis module, absolute handlability analysis module and with the algorithm synthesis module respectively, be used to export and be included under the absolute handlability coordinate the absolute handlability distribution density function that the absolute handlability distribution density of statistics obtains, the sensor performance evaluation function of the relative handlability distribution density function that obtains through the mapping of algorithm performance evaluation function with absolute handlability distribution density function;
Described algorithm is a Fingerprint Image Recognition Algorithms; Described sensor is the fingerprint collecting sensor.
3. handlability analysis and evaluation as claimed in claim 1 or 2 system is characterized in that, absolute handlability analysis module comprises:
Global statistics property analysis module is used to obtain pickup area size, resolution, resolution consistance and the gray shade scale of fingerprint image;
Partial statistics property analysis module is used to obtain the prospect background separability of fingerprint image, foreground area, and the prospect gray-scale statistical distributes, the empty statistical distribution frequently of prospect contrast statistical distribution and prospect;
The image problem analysis module is used for separating the foreground image that obtains according to prospect background and calculates prospect local frequencies intensity, and separates the level and smooth degree distribution that the foreground image that obtains calculates the crestal line field of direction according to prospect background;
Image is used for calculating the overlapping region quantity of information according to foreground image and distributes and the nonlinear deformation degree distribution the case study module; Described overlapping region refers to gather the different images from same finger, the zone corresponding to identical finger surface that its prospect part comprises jointly.
4. handlability analysis and evaluation as claimed in claim 3 system is characterized in that, the handlability analysis module comprises relatively:
The testing evaluation module is used for based on the fingerprint image storehouse algorithm for recognizing fingerprint performance being carried out testing evaluation;
The assessment result extraction module is used for meeting pre-conditioned assessment result from the assessment result extraction.
5. the handlability analyzing evaluation method of fingerprint collecting image library is characterized in that, comprising:
Step 51 is analyzed the fingerprint image storehouse, the absolute handlability in output fingerprint image storehouse;
Step 52 is analyzed through the recognition performance evaluation that Fingerprint Image Recognition Algorithms obtains fingerprint image storehouse and fingerprint image storehouse, the relative handlability in output fingerprint image storehouse;
Described absolute handlability is distortion of fingerprint image self-information and loss, and fingerprint image between influence the summation that Fingerprint Image Recognition Algorithms is carried out the problem of fingerprint recognition;
Described relative handlability is during a fingerprint in the fingerprint image storehouse is estimated the algorithm performance that obtains as certain Fingerprint Image Recognition Algorithms input, to embody the part that the target fingerprint recognition system requires.
6. handlability analyzing evaluation method as claimed in claim 5 is characterized in that, also comprises:
Step 53 is that functional value obtains the algorithm performance evaluation function with absolute handlability as the coordinate base, with relative handlability;
Step 54, the absolute handlability distribution density function that obtains with the absolute handlability distribution density of statistics under absolute handlability coordinate, with the relative handlability distribution density function that absolute handlability distribution density function is obtained through the mapping of algorithm performance evaluation function as the sensor performance evaluation function;
Described algorithm is a Fingerprint Image Recognition Algorithms; Described sensor is the fingerprint collecting sensor.
7. as claim 5 or 6 described handlability analyzing evaluation methods, it is characterized in that step 51 comprises:
Step 71, pickup area size, resolution, resolution consistance and the gray shade scale of acquisition fingerprint image;
Step 72, the prospect background separability of acquisition fingerprint image, foreground area, the prospect gray-scale statistical distributes, the empty statistical distribution frequently of prospect contrast statistical distribution and prospect;
Step 73 is separated the foreground image that obtains according to prospect background and is calculated prospect local frequencies intensity, and separates the level and smooth degree distribution that the foreground image that obtains calculates the crestal line field of direction according to prospect background;
Step 74 is calculated the overlapping region quantity of information according to foreground image and is distributed and the nonlinear deformation degree distribution; Described overlapping region refers to gather the different images from same finger, the zone corresponding to identical finger surface that its prospect part comprises jointly.
8. handlability analyzing evaluation method as claimed in claim 7 is characterized in that step 52 comprises:
Step 81 is carried out testing evaluation based on the fingerprint image storehouse to the algorithm for recognizing fingerprint performance;
Step 82, extraction meets pre-conditioned assessment result from assessment result.
CN2008102221677A 2008-09-10 2008-09-10 Method and system for analyzing and evaluating processability of finger print acquisition image library Expired - Fee Related CN101373516B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2008102221677A CN101373516B (en) 2008-09-10 2008-09-10 Method and system for analyzing and evaluating processability of finger print acquisition image library

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2008102221677A CN101373516B (en) 2008-09-10 2008-09-10 Method and system for analyzing and evaluating processability of finger print acquisition image library

Publications (2)

Publication Number Publication Date
CN101373516A CN101373516A (en) 2009-02-25
CN101373516B true CN101373516B (en) 2011-05-25

Family

ID=40447670

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2008102221677A Expired - Fee Related CN101373516B (en) 2008-09-10 2008-09-10 Method and system for analyzing and evaluating processability of finger print acquisition image library

Country Status (1)

Country Link
CN (1) CN101373516B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105205439B (en) * 2015-02-13 2017-05-03 比亚迪股份有限公司 Method for calculating area of fingerprint overlapping region and electronic device
US11281886B2 (en) 2018-02-12 2022-03-22 Huawei Technologies Co., Ltd. Fingerprint enrollment method and terminal
CN110223220B (en) * 2019-06-14 2023-03-31 北京百度网讯科技有限公司 Method and device for processing image
CN113705581A (en) * 2021-07-29 2021-11-26 中山市小兀照明科技有限公司 Light source performance characterization method and device and computer readable storage medium

Also Published As

Publication number Publication date
CN101373516A (en) 2009-02-25

Similar Documents

Publication Publication Date Title
CN110443128B (en) Finger vein identification method based on SURF feature point accurate matching
RU2302656C2 (en) Image identification system
CN102027488B (en) Fingerprint representation using gradient histograms
CN102176208B (en) Robust video fingerprint method based on three-dimensional space-time characteristics
CN110415241A (en) A kind of surface of concrete structure quality determining method based on computer vision
CN101373516B (en) Method and system for analyzing and evaluating processability of finger print acquisition image library
CN106251362B (en) A kind of sliding window method for tracking target and system based on fast correlation neighborhood characteristics point
CN101162503A (en) Method for extracting and recognizing human ear characteristic by improved Hausdorff distance
CN106557740B (en) The recognition methods of oil depot target in a kind of remote sensing images
CN106157323A (en) The insulator division and extracting method that a kind of dynamic division threshold value and block search combine
Fung et al. A review of evaluation of optimal binarization technique for character segmentation in historical manuscripts
CN101584624A (en) Guideboard recognition blind-guide device and method thereof based on DSP
CN109934278A (en) A kind of high-dimensional feature selection method of information gain mixing neighborhood rough set
CN115311507B (en) Building board classification method based on data processing
CN101582113A (en) Method for identifying face image with identity through layered comparison
CN109903053A (en) A kind of anti-fraud method carrying out Activity recognition based on sensing data
Zhang Half century for image segmentation
CN110175517A (en) A kind of Coal-rock identification method based on transform domain numerical characteristic
CN116152494A (en) Building foot point identification segmentation method based on two-stage 3D point cloud semantic segmentation
CN113378691B (en) Intelligent home management system and method based on real-time user behavior analysis
CN112330562A (en) Heterogeneous remote sensing image transformation method and system
CN109543610B (en) Vehicle detection tracking method, device, equipment and storage medium
CN109460768A (en) A kind of text detection and minimizing technology for histopathology micro-image
JP4042602B2 (en) Image processing device
CN106326882A (en) Fingerprint identification system and fingerprint identification method based on image quality assessment technology

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
DD01 Delivery of document by public notice

Addressee: Chengdu Finchos Electron Co., Ltd.

Document name: Notification to Pay the Fees

DD01 Delivery of document by public notice
DD01 Delivery of document by public notice

Addressee: CHENGDU FINCHOS ELECTRICAL Co.,Ltd.

Document name: Notification of Termination of Patent Right

DD01 Delivery of document by public notice
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20110525

Termination date: 20190910

CF01 Termination of patent right due to non-payment of annual fee