[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

CN111639298B - Running lot detection method of biological feature recognition algorithm - Google Patents

Running lot detection method of biological feature recognition algorithm Download PDF

Info

Publication number
CN111639298B
CN111639298B CN202010416922.6A CN202010416922A CN111639298B CN 111639298 B CN111639298 B CN 111639298B CN 202010416922 A CN202010416922 A CN 202010416922A CN 111639298 B CN111639298 B CN 111639298B
Authority
CN
China
Prior art keywords
rate
individual
recognition rate
false recognition
failure
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.)
Active
Application number
CN202010416922.6A
Other languages
Chinese (zh)
Other versions
CN111639298A (en
Inventor
张烜
赵国栋
梁润淇
李学双
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.)
Holy Point Century Technology Co ltd
Original Assignee
Holy Point Century Technology 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 Holy Point Century Technology Co ltd filed Critical Holy Point Century Technology Co ltd
Priority to CN202010416922.6A priority Critical patent/CN111639298B/en
Publication of CN111639298A publication Critical patent/CN111639298A/en
Application granted granted Critical
Publication of CN111639298B publication Critical patent/CN111639298B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10Complex mathematical operations
    • G06F17/18Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/30Authentication, i.e. establishing the identity or authorisation of security principals
    • G06F21/31User authentication
    • G06F21/32User authentication using biometric data, e.g. fingerprints, iris scans or voiceprints
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02PCLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computer Security & Cryptography (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Algebra (AREA)
  • Probability & Statistics with Applications (AREA)
  • Databases & Information Systems (AREA)
  • Operations Research (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Computer Hardware Design (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

The invention relates to a run-batch inspection method of a biological characteristic recognition algorithm, which comprises the following steps: collecting characteristic elements of a plurality of individual organisms; performing running batch comparison on characteristic elements in an individual body by adopting a biological characteristic recognition algorithm, counting failure factors and the number of each failure factor, and calculating the passing rate and the failure rate of each failure factor; comparing the running batches of the characteristic elements among individuals, calculating the frequency false recognition rate and the individual false recognition rate, and further calculating the combined false recognition rate; and judging the quality of the biological feature recognition algorithm based on the number of each failure factor, the passing rate, the failure rate of each failure factor, the frequency false recognition rate, the individual false recognition rate and the combined false recognition rate. The invention can avoid the distortion of the false recognition rate evaluation algorithm caused by the change of the number of elements in an individual, accurately evaluate the performance of the algorithm, reduce the influence of the number of characteristic elements in the individual on the false recognition rate statistics, and also count the failure proportion of each factor, thereby providing the optimization direction of the algorithm and playing a guiding role for subsequent research and development and optimization.

Description

Running lot detection method of biological feature recognition algorithm
Technical Field
The invention belongs to the technical field of biological feature recognition in information security, and particularly relates to a run-to-run inspection method of a biological feature recognition algorithm.
Background
The biological feature recognition technology is a technology for recognizing information according to the characteristic biological features of human bodies, wherein various subject principles such as computers, biological sensing, optics and statistics are involved, and the technology comprises fingerprint recognition, finger vein recognition, face recognition, iris recognition, voice recognition and the like. The biological recognition technology is widely applied to various fields such as banks, education, security inspection, military industry, life and the like. Taking fingerprint identification and finger vein identification as an example, the invention patent with the patent number of CN109583279A discloses a fingerprint and finger vein combined identification algorithm, firstly, N candidate users are searched out by utilizing a finger vein identification method based on the combination of characteristic points and binary images by utilizing a finger vein relative position direction field similarity algorithm based on the local surrounding points of the characteristic points of the fingerprint. Experiments show that the algorithm has good stability, strong anti-interference capability and high retrieval speed based on the fingerprint feature point local surrounding point relative position direction field similarity algorithm. The finger vein recognition algorithm based on the combination of the feature points and the binary image is stable. Under the condition of a certain false recognition rate, the fingerprint and finger vein combined recognition algorithm has obviously higher recognition performance than the traditional fingerprint and finger vein single-mode recognition algorithm; but also the recognition efficiency is obviously improved.
With the needs of the age, the biological recognition industry is rapidly developed. In the process of developing and optimizing an algorithm, a running batch test link is indispensable, wherein a test index is an index for evaluating the performance of the algorithm, and the false recognition rate is one of the most important evaluation indexes. The existing false recognition rate statistical method is based on ROC curve. The method uses a database containing a large number of biological characteristic elements to conduct comparison (i.e. run-batch) and counts the ratio of error receiving times. In general, a certain number of individuals are contained in the biological characteristic databases, each individual contains an equal number of characteristic elements, the total number of times of inter-individual comparison is taken as a denominator, the number of times of error acceptance is taken as a numerator, and the number of error acceptance rate is compared with the number of times of error acceptance rate, so that the limitation of the method is that the difference of the number of individuals and the number of the characteristic elements in the individuals leads to the error acceptance rate, and the index of the error acceptance rate loses the meaning as an algorithm evaluation standard; in addition, when running batches at present, comparison is carried out in individuals, pass rate statistics is carried out only on successful comparison times, and failure factor analysis is not carried out.
Illustrating: assuming A, B two fingerprint identification companies, respectively testing algorithms of the respective companies, wherein a database of the company a comprises 9 fingers, 35 fingerprint images are each finger, running is carried out, the total number of non-repeated comparison between every two fingerprint pairs is 9 x 35 x 8 x 35 x 0.5=44100, the database of the company B comprises 50 fingers, 6 fingerprint images are each finger, and the number of non-repeated comparison is 50 x 6 x 49 x 6 x 0.5=44100. Assuming that the algorithm error acceptance times of both companies are 1, the error recognition rate is 1/44100. From this, it is found that the error rate is the same, but the running lot data amounts of company a and company B are not identical, so that it is impossible to prove that the algorithm error performance of both companies is identical. In addition, when the number of individuals is unchanged, only the number of characteristic elements in the individuals is increased, the false recognition rate is also scientifically reduced, and readers can test by themselves according to the calculation mode in examples.
Therefore, the algorithm performance evaluation is 'virtual high' because the number of data is not scientifically increased, or the number of individuals is simply increased, so that the false recognition rate is reduced (namely, the calculation denominator of the false recognition rate is increased).
Disclosure of Invention
The invention aims at solving the problems that in the traditional running batch detection method, the difference of false recognition rate is caused by the difference of the number of individuals and the number of characteristic elements in the individuals, so that the false recognition rate index loses the meaning of being an algorithm evaluation standard, the traditional running batch detection only carries out passing rate statistics on successful comparison times, and failure factor analysis is not carried out.
In order to achieve the purpose, the technical scheme provided by the invention is as follows:
the invention relates to a running batch inspection method of a biological characteristic recognition algorithm, which comprises the following steps:
1) Collecting characteristic elements of a plurality of individual creatures using a biometric device;
2) Performing running batch comparison on characteristic elements in an individual by adopting a biological characteristic recognition algorithm, counting failure factors and the number of each failure factor, and calculating the passing rate and the failure rate of each failure factor;
3) Performing running batch comparison on characteristic elements among individuals, calculating the frequency false recognition rate and the individual false recognition rate, and further calculating the combined false recognition rate according to the frequency false recognition rate and the individual false recognition rate;
4) The quality of the biological feature recognition algorithm is evaluated based on the number of each failure factor, the passing rate, the failure rate of each failure factor, the frequency false recognition rate, the individual false recognition rate and the combined false recognition rate.
Preferably, the step 1) records the number of individuals and the number of characteristic elements of a single individual when collecting characteristic elements of a plurality of individual organisms.
Preferably, when the number of each failure factor is counted in the step 2), if a plurality of failure factors occur in a certain element, the number is increased once under the statistics of the corresponding number of the plurality of failure factors.
Preferably, in the step 3), when the feature elements among individuals are run-to-run and aligned, a failure factor detection algorithm is used to detect a failure factor, and when the cross error acceptance among individuals occurs, the detection is only recorded once.
Preferably, the calculation formula of the frequency false recognition rate in the step 3) is as follows:
Figure BDA0002494490480000031
wherein FAR represents the frequency false recognition rate, NFA represents the frequency of false acceptance, and NIRA represents the total number of comparison between individuals.
Preferably, the individual error rate calculation formula in the step 3) is:
Figure BDA0002494490480000032
where IFAR represents individual false recognition rate, INFA represents individual false acceptance number, and NI represents total individual number.
Preferably, the calculation formula of combining the false recognition rate in the step 3) is as follows:
UTFR=T 1 ×C 1 ×FAR+T 2 ×C 2 ×IFAR×100% (3),
wherein UTFR represents the combined false recognition rate, T 1 、T 2 Represents weight ratio, and T 1 +T 2 =1,C 1 、C 2 Representing the required coefficients of uniform order of magnitude, FAR representing the frequency error rate, IFAR representing the individual error rate.
Compared with the prior art, the technical scheme provided by the invention has the following beneficial effects:
1. the invention makes optimization processing according to the actual situation of the biological recognition technology in the use process, provides an index of individual false recognition rate and a combinative false recognition rate index as evaluation criteria, wherein the individual false recognition rate is the false recognition rate taking the number of individuals as a calculation unit, combines the false recognition rate taking the existing times as the calculation unit with the false recognition rate taking the number of individuals as the calculation unit, and further obtains a new false recognition rate evaluation criteria, thereby avoiding the change of the number of elements in the individuals to a certain extent, evaluating the performance of the algorithm more accurately, reducing the influence of the number factors of the characteristic elements in the individuals on the false recognition rate statistics, and avoiding the distortion of the false recognition rate evaluation algorithm.
2. The invention provides the analysis and detection of the failure reasons for the comparison of the failed characteristic elements in the running batch algorithm, and the statistics of the failure proportion of each factor, thereby providing the optimization direction of the algorithm and playing a guiding role for the subsequent research and development and optimization.
Drawings
FIG. 1 is a flow chart of a run-to-run verification method of the biometric algorithm of the present invention.
Detailed Description
For further understanding of the present invention, the present invention will be described in detail with reference to the following examples, which are given by way of illustration of the present invention, but are not intended to limit the scope of the present invention.
Example 1
Referring to fig. 1, the invention relates to a running batch inspection method of a biological characteristic recognition algorithm, which comprises the following steps:
1) Collecting characteristic elements of a plurality of individual organisms by using a biological identification device, and recording the number of individuals and the number of characteristic elements under a single individual when the characteristic elements of the plurality of individual organisms are collected;
in this embodiment, a company collects staff in charge, and the batch data includes 2000 fingers, 8 finger vein images in each finger, and performs numbering preprocessing on the finger vein images as a database in this embodiment.
2) Comparing the running batches of characteristic elements in an individual body by adopting a biological characteristic recognition algorithm, counting the number of failure factors and each failure factor, calculating the passing rate and each failure factor failure rate, wherein one or more failure factors possibly appear in one pair of failed comparison elements, when a plurality of failure factors appear in a certain element, the number of the corresponding plurality of failure factors is increased once respectively, for example, one element has an a failure factor, the other element has a b failure factor and a c failure factor in one pair of failed comparison elements, the comparison failure elements are respectively increased once under the number statistics of the a, b and c failure factors, whether the failure elements are counted under the corresponding failure rate is checked, the accuracy of running batch results is ensured, the proportion of single failure factors to all the failure factors is counted, and the failure rate is distributed according to the proportion, so that the final failure factor failure rate is obtained;
in this embodiment, the acquired 2000-digit images are subjected to in-vivo batch running by using an algorithm a, and the batch running result is that: total comparison times are 8 x 7 x 0.5 x 2000=56000 times, wherein the comparison success times are 53234 times, namely the success rate is (53234/56000) x100% =95.06%; the failure image was detected using a failure factor detection algorithm, wherein the main failure factors are image light leakage, finger smudge, finger back pressure, finger skinning, finger axis rotation, finger horizontal rotation and other undetected failure factors, and the specific failure times, ratios, and failure factor failure rates are shown in table 1:
table 1: algorithm A in-vivo run-out failure factor failure rate statistics
Figure BDA0002494490480000041
The analysis in table 1 gives: the most main failure factors of the comparison failure are image light leakage and finger stress, failure images under each failure factor are checked, and no failure factor classification statistical error condition exists.
3) When the characteristic elements among individuals are subjected to running lot comparison, the failure factor is detected by using a failure factor detection algorithm, when the cross error acceptance among the individuals occurs, for example, the error acceptance of the characteristic element a of the individual 1 and the characteristic element b of the individual 2 is recorded once, and the error acceptance of the characteristic element a of the individual 1 and the characteristic element c of the individual 3 is performed, only the individuals 1, 2 and 3 are counted as error accepted individuals, the occurrence times are not counted, in the embodiment, the acquisition of 2000 refers to the running lot among the 8 images by using an A algorithm, and the running lot result is obtained: total number of alignments was 2000 x 8 x 1999 x 8 x 0.5= 127936000, count 147 number of failed alignments, finger 12 finger involved in false recognition alignment;
then calculating the frequency false recognition rate and the individual false recognition rate,
the calculation formula of the frequency false recognition rate is as follows:
Figure BDA0002494490480000051
wherein, FAR represents the frequency false recognition rate, NFA represents the frequency of error acceptance, NIRA represents the total frequency of inter-individual comparison, and the frequency false recognition rate FAR is about 0.00114% according to the formula;
the individual error rate calculation formula is:
Figure BDA0002494490480000052
wherein IFAR represents an individual error recognition rate, INFA represents an individual error acceptance number, NI represents a total individual number, and the individual error recognition rate ifar=0.60% is calculated according to the above formula;
further calculating a combined false recognition rate according to the frequency false recognition rate and the individual false recognition rate, wherein a calculation formula of the combined false recognition rate is as follows:
UTFR=T 1 ×C 1 ×FAR+T 2 ×C 2 ×IFAR×100% (3),
wherein UTFR represents the combined false recognition rate, based on the analysis of the batch data: the present example recommends using the weight proportion T 1 =0.38、T 2 =0.62, recommended to use a coefficient of uniform magnitude C 1 ≈65、C 2 And about 1/65, and obtaining the combination error recognition rate UTFR about 0.0339 percent.
4) The quality of the biological feature recognition algorithm is evaluated based on the number of each failure factor, the passing rate, the failure rate of each failure factor, the frequency false recognition rate, the individual false recognition rate and the combined false recognition rate.
Example two
The results of implementing one were all calculated by running the batch with the algorithm a, and the results showed that the same database was used, running the batch again with the algorithm B, and the number of times of success of the individual running batch comparison was 53149 times, and the success rate was (53149/56000) ×100% = 94.91%. The failure rate of the main failure factors is shown in Table 2.
Table 2: algorithm B in-vivo run-out failure factor failure rate statistics
Figure BDA0002494490480000061
Analysis by table 2 gives: the most main failure factors of the comparison failure are image light leakage and finger stress, failure images under each failure factor are checked, and no failure factor classification statistical error condition exists.
Performing inter-individual running batch by using a B algorithm, and counting the failure comparison times 164 times, wherein the false recognition comparison is related to the finger 14; the frequency error rate far= (164/12793600) ×100% ≡0.00128%, and the individual error rate ifar= (14/2000) ×100% = 0.70%;
according to the data with the frequency error recognition rate of 0.00128 percent and the individual error recognition rate of 0.70 percent, the weight proportion T is recommended to be used 1 =0.34、T 2 =0.66, recommended to use a coefficient of uniform magnitude C 1 ≈65、C 2 And about 1/65, and the combination error rate UTFR is about 0.0354%.
The proportion of each evaluation standard of the AB two algorithms is as follows:
a algorithm B algorithm
Pass rate of 95.06% 94.91%
Frequency error rate 0.00114% 0.00128%
Individual error rate 0.60% 0.70%
Combined error rate 0.0339% 0.0354%
According to the evaluation results of the A, B two sets of algorithms, the performance of the algorithm A is better, and the specific analysis is as follows:
the passing rate of the algorithm A is slightly higher than that of the algorithm B;
the failure factors in the individuals of the two algorithms A and B are mainly finger placement light leakage and finger stress factors, the two problems are emphasized and optimized in the later period, especially the light leakage factor of the algorithm B is obviously more serious, and the algorithm B has more other failure factors.
And 3. The frequency false recognition rate of the B algorithm is higher, and statistics is carried out on the number of false recognition individuals, so that the B algorithm has more false recognition of bodies in the actual use process. The aggregate false recognition rate of the A algorithm is lower through calculation and combination of false recognition rates.
In summary, the performance of the A algorithm is better.
The present invention has been described in detail with reference to the embodiments, but the description is only the preferred embodiments of the present invention and should not be construed as limiting the scope of the invention. All equivalent changes and modifications within the scope of the present invention should be considered as falling within the scope of the present invention.

Claims (4)

1.A run-batch inspection method of a biological feature recognition algorithm is characterized in that: which comprises the following steps:
1) Collecting characteristic elements of a plurality of individual creatures using a biometric device;
2) Performing running batch comparison on characteristic elements in an individual by adopting a biological characteristic recognition algorithm, counting failure factors and the number of each failure factor, and calculating the passing rate and the failure rate of each failure factor;
3) Performing running batch comparison on characteristic elements among individuals, calculating the frequency false recognition rate and the individual false recognition rate, and further calculating the combined false recognition rate according to the frequency false recognition rate and the individual false recognition rate;
the calculation formula of the frequency false recognition rate is as follows:
Figure FDA0004192627050000011
wherein FAR represents the frequency false recognition rate, NFA represents the frequency of false acceptance, and NIRA represents the total number of comparison among individuals;
the individual error rate calculation formula is:
Figure FDA0004192627050000012
wherein IFAR represents individual false recognition rate, INFA represents individual false acceptance number, NI represents total individual number; the calculation formula combined with the false recognition rate is as follows:
UTFR=T 1 ×C 1 ×FAR+T 2 ×C 2 ×IFAR×100% (3),
wherein UTFR represents the combined false recognition rate, T 1 、T 2 Represents weight ratio, and T 1 +T 2 =1,C 1 、C 2 Representing the required coefficients of uniform orders of magnitude, wherein FAR represents the frequency false recognition rate, and IFAR represents the individual false recognition rate;
4) The quality of the biological feature recognition algorithm is evaluated based on the number of each failure factor, the passing rate, the failure rate of each failure factor, the frequency false recognition rate, the individual false recognition rate and the combined false recognition rate.
2. The run-to-run verification method of a biometric algorithm of claim 1, wherein: and when the characteristic elements of a plurality of individual organisms are acquired in the step 1), recording the number of individuals and the number of the characteristic elements of a single individual.
3. The run-to-run verification method of a biometric algorithm of claim 1, wherein: when the number of each failure factor is counted in the step 2), if a plurality of failure factors occur to a certain element, the number is increased once under the corresponding statistics of the number of the plurality of failure factors.
4. The run-to-run verification method of a biometric algorithm of claim 1, wherein: and 3) when the characteristic elements among individuals are subjected to running batch comparison, detecting the failure factor by using a failure factor detection algorithm, and recording only once when the cross error acceptance among the individuals occurs.
CN202010416922.6A 2020-05-15 2020-05-15 Running lot detection method of biological feature recognition algorithm Active CN111639298B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010416922.6A CN111639298B (en) 2020-05-15 2020-05-15 Running lot detection method of biological feature recognition algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010416922.6A CN111639298B (en) 2020-05-15 2020-05-15 Running lot detection method of biological feature recognition algorithm

Publications (2)

Publication Number Publication Date
CN111639298A CN111639298A (en) 2020-09-08
CN111639298B true CN111639298B (en) 2023-06-20

Family

ID=72328540

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010416922.6A Active CN111639298B (en) 2020-05-15 2020-05-15 Running lot detection method of biological feature recognition algorithm

Country Status (1)

Country Link
CN (1) CN111639298B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101004788A (en) * 2006-01-17 2007-07-25 公安部第一研究所 System and method of evaluating algorithm for recognizing fingerprint on certificate
CN109740457A (en) * 2018-12-20 2019-05-10 杭州当虹科技股份有限公司 A kind of face recognition algorithms evaluating method
CN110378324A (en) * 2019-07-15 2019-10-25 易诚高科(大连)科技有限公司 A kind of face recognition algorithms appraisal procedure based on quality dimensions
CN111126122A (en) * 2018-10-31 2020-05-08 浙江宇视科技有限公司 Face recognition algorithm evaluation method and device

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006221411A (en) * 2005-02-10 2006-08-24 Sharp Corp Authentication algorithm evaluation device and method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101004788A (en) * 2006-01-17 2007-07-25 公安部第一研究所 System and method of evaluating algorithm for recognizing fingerprint on certificate
CN111126122A (en) * 2018-10-31 2020-05-08 浙江宇视科技有限公司 Face recognition algorithm evaluation method and device
CN109740457A (en) * 2018-12-20 2019-05-10 杭州当虹科技股份有限公司 A kind of face recognition algorithms evaluating method
CN110378324A (en) * 2019-07-15 2019-10-25 易诚高科(大连)科技有限公司 A kind of face recognition algorithms appraisal procedure based on quality dimensions

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
BioEVA: An Evaluation Tool for Biometric Algorithms;Luiz H. R. Sucupira Jr.等;《Lecture notes in computer science》;全文 *
人脸识别应用研究;曾昊;《中国优秀硕士学位论文全文数据库》;全文 *
基于人工标定的智能交通参数算法评估;高江杰;《中国优秀硕士学位论文全文数据库》;全文 *
基于非局部梯度的图像质量评价算法;高敏娟等;《电子与信息学报》(第05期);全文 *

Also Published As

Publication number Publication date
CN111639298A (en) 2020-09-08

Similar Documents

Publication Publication Date Title
US7870151B2 (en) Fast accurate fuzzy matching
CN105590156B (en) Detection method of high-risk bank card and data processing device
CN109003266B (en) Method for counting image quality subjective evaluation results based on fuzzy clustering
CN102567993A (en) Fingerprint image quality evaluation method based on main component analysis
CN109255029A (en) A method of automatic Bug report distribution is enhanced using weighted optimization training set
CN109657011A (en) A kind of data digging method and system screening attack of terrorism criminal gang
CN113128567A (en) Abnormal electricity consumption behavior identification method based on electricity consumption data
CN113569920A (en) Second neighbor anomaly detection method based on automatic coding
CN109447153A (en) Divergence-excitation self-encoding encoder and its classification method for lack of balance data classification
CN111639298B (en) Running lot detection method of biological feature recognition algorithm
CN111340516A (en) Satisfaction evaluation system and method based on information entropy and variation coefficient fusion algorithm
CN112926577B (en) Medical bill image structuring method and device and computer readable medium
Cope et al. Classifying plant leaves from their margins using dynamic time warping
CN110020686A (en) A kind of road surface method for detecting abnormality based on intelligent perception sensing data
CN113792141B (en) Feature selection method based on covariance measurement factor
CN104778478A (en) Handwritten numeral identification method
CN116340791A (en) Noise data classification method and user classification method based on random sampling clustering
CN116051185A (en) Advertisement position data abnormality detection and screening method
CN110390309B (en) Finger vein illegal user identification method based on residual distribution
WO2022011827A1 (en) Method for optimizing loss function by means of error cause reinforcement
Le et al. Sequential pattern mining of large combinable items with values for a set-of-items recommendation
CN115242457A (en) Method and device for detecting log data, electronic equipment and storage medium
CN115237970A (en) Data prediction method, device, equipment, storage medium and program product
CN115082958A (en) Store clerk identification method, system and terminal
CN106779843A (en) A kind of competing method and apparatus for closing relationship analysis of trade company based on customer group's feature

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant