CN112183302A - Face information identification method, system and terminal based on evolutionary weak classifier - Google Patents
Face information identification method, system and terminal based on evolutionary weak classifier Download PDFInfo
- Publication number
- CN112183302A CN112183302A CN202011015635.0A CN202011015635A CN112183302A CN 112183302 A CN112183302 A CN 112183302A CN 202011015635 A CN202011015635 A CN 202011015635A CN 112183302 A CN112183302 A CN 112183302A
- Authority
- CN
- China
- Prior art keywords
- weak classifier
- weak
- sets
- face information
- classifier
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000001815 facial effect Effects 0.000 claims description 47
- 238000012549 training Methods 0.000 claims description 20
- 230000015654 memory Effects 0.000 claims description 15
- 230000035772 mutation Effects 0.000 claims description 12
- 230000014509 gene expression Effects 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 8
- 230000008774 maternal effect Effects 0.000 claims description 8
- 230000008775 paternal effect Effects 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 5
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 238000013136 deep learning model Methods 0.000 abstract description 7
- 230000035945 sensitivity Effects 0.000 abstract description 4
- 238000004364 calculation method Methods 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 210000000697 sensory organ Anatomy 0.000 description 4
- 238000012706 support-vector machine Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 206010063659 Aversion Diseases 0.000 description 2
- 238000007635 classification algorithm Methods 0.000 description 2
- 235000009508 confectionery Nutrition 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 230000008030 elimination Effects 0.000 description 2
- 238000003379 elimination reaction Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/178—Human faces, e.g. facial parts, sketches or expressions estimating age from face image; using age information for improving recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The face information identification method, the face information identification system and the face information identification terminal based on the evolutionary weak classifier solve the problems that in the prior art, a deep learning model is adopted to carry out face identification, and the limitation in practical application comprises large storage capacity, more power consumption, high calculation delay, high sensitivity to data imbalance among categories and the like. The face information identification method based on the evolutionary weak classifier, which is adopted by the invention, has high accuracy and lightweight attribute, is suitable for both large data sets and small data sets, can ensure that the model is more flexible under the condition of ensuring certain accuracy, can realize real-time reasoning at the edge end, and has increased compatibility and robustness.
Description
Technical Field
The invention relates to the field of face recognition, in particular to a face information recognition method, a face information recognition system and a face information recognition terminal based on an evolutionary weak classifier.
Background
Facial appearance contains a lot of valuable personal information such as expression, style, gender, age, etc. Since human face information is relatively stable, many researchers are invested in the development of face intelligent recognition technology. Face recognition is one of the problems in the field that the prospect is wide but the difficulty is high as an important branch of pattern recognition.
A typical and efficient approach to this task is to first extract facial feature values through descriptors and filters, and then use appropriate classifiers to complete the rest. Today, deep learning models, which integrate feature extraction and pattern classification into an end-to-end learning system so that the retrieval and use of potential features can be optimized through a training process, are quite competitive in classification accuracy. From the viewpoint of accuracy, the deep learning model can solve most problems in face information recognition, but brings some limitations in practical application. First, in real commercial projects, it is often impossible to get access to large amounts of tagged data. In this case, the phenomenon of overfitting often occurs when the deep neural network is trained. Secondly, the heavyweight architecture of the deep learning model directly leads to the problems of large storage capacity, large power consumption, high computation delay and the like, especially on end-user devices such as cameras and mobile phones. A final problem with deep learning methods is that they are very sensitive to imbalances in the data between classes.
Disclosure of Invention
In view of the above disadvantages of the prior art, an object of the present invention is to provide a method, a system, and a terminal for recognizing face information based on an evolutionary weak classifier, which are used to solve the limitations of the prior art in practical applications, including large storage capacity, high power consumption, high computation delay, and sensitivity to imbalance of data between classes, when a deep learning model is used for face recognition.
In order to achieve the above objects and other related objects, the present invention provides a face information recognition method based on an evolutionary weak classifier, including: extracting at least two facial features from the face image data by using a general filter; training based on the facial features to obtain a plurality of weak classifiers and obtaining identification results contained in the weak classifiers; randomly forming a plurality of weak classifier sets containing the same number of weak classifiers, and obtaining the performance indexes of the weak classifier sets related to the identification result; performing one or more generations of evolution on the weak classifier set according to the performance indexes of each weak classifier set to obtain an optimal weak classifier set; and carrying out reasoning and classification on the face image data according to the optimal weak classifier set to obtain a face information identification result.
In an embodiment of the present invention, the performing one or more generations of evolutions on the weak classifier sets according to the performance indexes of the weak classifier sets to obtain the optimal weak classifier set includes: when performing first-generation evolution on the weak classifier sets according to the performance indexes of the weak classifier sets, randomly selecting weak classifier sets with evolution threshold quantity and performing performance index sequencing to obtain a first-generation optimal weak classifier set; when the weak classifier sets are subjected to multi-generation evolution according to the performance indexes of the weak classifier sets, performing parent selection according to a descendant set of the last generation obtained by multi-time parent selection and with the evolution threshold number as a to-be-selected weak classifier set of the current generation to obtain one or more descendant sets, and taking a set with the highest performance index in the descendant sets of the current generation as an optimal weak classifier set of the current generation.
In an embodiment of the present invention, the facial features include: one or more of an expression feature, a style feature, a gender feature, an age feature, and a facial feature.
In an embodiment of the present invention, the training to obtain a plurality of weak classifiers based on the facial features and obtain the recognition results included in each weak classifier includes: based on a plurality of weak classifiers respectively trained by any two facial features; an identification result is obtained that each classifier contains a statistical probability of its corresponding facial features.
In an embodiment of the present invention, the manner of randomly forming a plurality of weak classifier sets including the same number of weak classifiers and obtaining the performance index of each weak classifier set related to the identification result is as follows: randomly obtaining a plurality of weak classifier sets containing the same number according to the weak classifiers; and deducing and obtaining a performance index for judging the classification accuracy of each weak classifier set according to the identification result of each weak classifier in each weak classifier set.
In one embodiment of the present invention, the parent selection method comprises: randomly selecting a plurality of weak classifier sets to be selected as male parent candidate sets, and randomly selecting female parent candidate sets with the same quantity as the male parent candidate sets from the weak classifier sets to be selected which are not selected as the male parent candidate sets; respectively taking the set with the maximum performance index in the male parent candidate set and the female parent candidate set as a male parent set and a female parent set; and obtaining a progeny set according to the male parent set and the female parent set.
In an embodiment of the present invention, the manner of obtaining the offspring collection according to the paternal collection and the maternal collection includes: carrying out cross and/or mutation operation on the male parent set and the female parent set to obtain a progeny set; wherein the interleaving operation comprises: randomly selecting one or more weak classifiers from the parent set and the mother set; the mutation operation includes randomly selecting one or more weak classifiers in the non-paternal set and the non-maternal set.
In an embodiment of the present invention, the manner of performing inference classification on the face image data according to the optimal weak classifier set to obtain a face information recognition result includes: based on the Bayesian probability formula, P (E ═ E)j) Carrying out normalization to obtain an improved Bayesian probability formula, and carrying out inference classification on the face image data according to the optimal weak classifier set to obtain a face information identification result; the Bayesian probability comprises:
wherein s isiDenotes the output of the classifier, E ═ EjIndicating that the known sample E is in class EjAnd T is the optimal weak classifier set and consists of K weak classifiers.
To achieve the above and other related objects, the present invention provides a face information recognition system based on an evolutionary weak classifier, the system comprising: the feature extraction module is used for extracting at least two facial features from the face image data by using the general filter; the weak classifier training module is connected with the feature extraction module and used for training and obtaining a plurality of weak classifiers based on the facial features and obtaining the identification results contained in the weak classifiers; the weak classifier set module is connected with the weak classifier training module and used for randomly forming a plurality of weak classifier sets containing the same number of weak classifiers and obtaining the performance indexes of the weak classifier sets related to the identification result; the evolution module is connected with the weak classifier set module and is used for carrying out one or more generations of evolution on the weak classifier set according to the performance indexes of each weak classifier set to obtain an optimal weak classifier set; and the reasoning and classifying module is connected with the evolution module and used for carrying out reasoning and classifying on the face image data according to the optimal weak classifier set to obtain a face information identification result.
In order to achieve the above objects and other related objects, the present invention provides a face information recognition terminal based on an evolutionary weak classifier, comprising: a memory for storing a computer program; and the processor is used for executing the face information identification method based on the evolution weak classifier.
As described above, the face information identification method, system and terminal based on the evolutionary weak classifier of the present invention have the following beneficial effects: the face information identification method based on the evolutionary weak classifier, which is adopted by the invention, has high accuracy and lightweight attribute, is suitable for both large data sets and small data sets, can ensure that the model is more flexible under the condition of ensuring certain accuracy, can realize real-time reasoning at the edge end, and has increased compatibility and robustness.
Drawings
Fig. 1 is a schematic flow chart of a face information recognition method based on an evolutionary weak classifier according to an embodiment of the present invention.
Fig. 2 is a flowchart illustrating evolution of a weak classifier set according to an embodiment of the present invention.
Fig. 3 is a schematic structural diagram of a face information recognition system based on an evolutionary weak classifier according to an embodiment of the present invention.
Fig. 4 is a schematic structural diagram of a face information recognition terminal based on an evolutionary weak classifier according to an embodiment of the present invention.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict.
It is noted that in the following description, reference is made to the accompanying drawings which illustrate several embodiments of the present invention. It is to be understood that other embodiments may be utilized and that mechanical, structural, electrical, and operational changes may be made without departing from the spirit and scope of the present invention. The following detailed description is not to be taken in a limiting sense, and the scope of embodiments of the present invention is defined only by the claims of the issued patent. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. Spatially relative terms, such as "upper," "lower," "left," "right," "lower," "below," "lower," "over," "upper," and the like, may be used herein to facilitate describing one element or feature's relationship to another element or feature as illustrated in the figures.
Throughout the specification, when a part is referred to as being "connected" to another part, this includes not only a case of being "directly connected" but also a case of being "indirectly connected" with another element interposed therebetween. In addition, when a certain part is referred to as "including" a certain component, unless otherwise stated, other components are not excluded, but it means that other components may be included.
The terms first, second, third, etc. are used herein to describe various elements, components, regions, layers and/or sections, but are not limited thereto. These terms are only used to distinguish one element, component, region, layer or section from another element, component, region, layer or section. Thus, a first element, component, region, layer or section discussed below could be termed a second element, component, region, layer or section without departing from the scope of the present invention.
Also, as used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context indicates otherwise. It will be further understood that the terms "comprises," "comprising," and/or "comprising," when used in this specification, specify the presence of stated features, operations, elements, components, items, species, and/or groups, but do not preclude the presence, or addition of one or more other features, operations, elements, components, items, species, and/or groups thereof. The terms "or" and/or "as used herein are to be construed as inclusive or meaning any one or any combination. Thus, "A, B or C" or "A, B and/or C" means "any of the following: a; b; c; a and B; a and C; b and C; A. b and C ". An exception to this definition will occur only when a combination of elements, functions or operations are inherently mutually exclusive in some way.
The embodiment of the invention provides a face information identification method based on an evolutionary weak classifier, which solves the problems of large storage capacity, more power consumption, high calculation delay, high sensitivity to data imbalance among categories and the like of the limitation of face identification in practical application by adopting a deep learning model in the prior art. The face information identification method based on the evolutionary weak classifier, which is adopted by the invention, has high accuracy and lightweight attribute, is suitable for both large data sets and small data sets, can ensure that the model is more flexible under the condition of ensuring certain accuracy, can realize real-time reasoning at the edge end, and has increased compatibility and robustness.
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings so that those skilled in the art can easily implement the embodiments of the present invention. The present invention may be embodied in many different forms and is not limited to the embodiments described herein.
As shown in fig. 1, a schematic flow chart of a face information identification method based on an evolutionary weak classifier in the embodiment of the present invention is shown.
The method comprises the following steps:
step S11: at least two facial features are extracted for the face image data using a generic filter.
Optionally, the general filter is any general filter, including but not limited to a low-pass filter, a high-pass filter, a band-stop filter, and the like.
Optionally, at least two facial features are extracted from the face image data using a gradient histogram filter.
Optionally, the facial features include: one or more of an expression feature, a style feature, a gender feature, an age feature, and a facial feature.
Specifically, the expression features include, but are not limited to, expressions such as obstinate/sad, painful, angry, fear, surprise, happiness and aversion; the style characteristics include, but are not limited to, gentle, high-cold, sweet, fashionable and other styles; the gender characteristics include male and female; age characteristics include, but are not limited to: the age stage; the features of the five sense organs include, but are not limited to: current status characteristics of the five sense organs.
Optionally, each face image data has a classification label corresponding to a facial feature.
Step S12: and training based on the facial features to obtain a plurality of weak classifiers, and obtaining the identification result contained in each weak classifier.
Optionally, the obtaining a plurality of weak classifiers based on the facial feature training and obtaining the recognition result included in each weak classifier includes: based on a plurality of weak classifiers respectively trained by any two facial features; an identification result is obtained that each classifier contains a statistical probability of its corresponding facial features.
Specifically, two facial features are arbitrarily selected from the facial features and trained to obtain a weak classifier, and a plurality of weak classifiers are obtained together; and obtaining a recognition result that each classifier contains a statistical probability of its corresponding facial features.
For example, if the facial features are angry, fear, surprise, happy, 6 weak classifiers are obtained.
Optionally, the weak classifier comprises a Support Vector Machine (SVM), which is a simple binary classifier that better separates two sample sets by hyperplane. Since the SVM classifier predicts the samples as one group or another, we call the output the left and right. For each SVM, we set two statistical tables, one table recording the statistical frequency of each class predicted to the left set of samples, and the other table recording the statistical frequency of each class predicted to the right set of samples, and we use these statistical tables for whole body learning, rather than using a priori expectations.
Step S13: and randomly forming a plurality of weak classifier sets containing the same number of weak classifiers, and obtaining the performance indexes of the weak classifier sets related to the identification result.
Optionally, the manner of randomly forming a plurality of weak classifier sets including the same number of weak classifiers and obtaining the performance index of each weak classifier set related to the identification result is as follows: randomly obtaining a plurality of weak classifier sets containing the same number according to the weak classifiers; and deducing and obtaining a performance index for judging the classification accuracy of each weak classifier set according to the identification result of each weak classifier in each weak classifier set.
Optionally, the recognition result of each weak classifier in each weak classifier set is compared with the actual result of the recognized face image data, so as to obtain the classification accuracy of each weak classifier, obtain the accuracy of each weak classifier set, and obtain the performance index. Wherein the higher the accuracy, the higher the performance index.
Optionally, the frequency of each facial feature in the recognition result of each weak classifier in each weak classifier set is compared with the actual result of the recognized face image data, and the frequency of the facial feature corresponding to the actual result is the classification accuracy of the weak classifier, so as to obtain the accuracy of each weak classifier set and obtain the performance index. Wherein the higher the accuracy, the higher the performance index.
Step S14: and carrying out one or more generations of evolution on the weak classifier set according to the performance indexes of each weak classifier set to obtain an optimal weak classifier set.
Optionally, the performing one or more generations of evolutions on the weak classifier sets according to the performance indexes of each weak classifier set to obtain an optimal weak classifier set includes: when performing first-generation evolution on the weak classifier sets according to the performance indexes of the weak classifier sets, randomly selecting weak classifier sets with evolution threshold quantity and performing performance index sequencing to obtain a first-generation optimal weak classifier set; when the weak classifier sets are subjected to multi-generation evolution according to the performance indexes of the weak classifier sets, performing parent selection according to a descendant set of the last generation obtained by multi-time parent selection and with the evolution threshold number as a to-be-selected weak classifier set of the current generation to obtain one or more descendant sets, and taking a set with the highest performance index in the descendant sets of the current generation as an optimal weak classifier set of the current generation.
Optionally, the parent selection mode comprises: randomly selecting a plurality of weak classifier sets to be selected as male parent candidate sets, and randomly selecting female parent candidate sets with the same quantity as the male parent candidate sets from the weak classifier sets to be selected which are not selected as the male parent candidate sets; respectively taking the set with the maximum performance index in the male parent candidate set and the female parent candidate set as a male parent set and a female parent set; and obtaining a progeny set according to the male parent set and the female parent set.
For example, the parental selection approach includes: and setting the selection strength of the male parent and the female parent as M, and eliminating based on the performance indexes of the M candidate sets when the male parent is selected. That is, M sets are randomly selected, and the set with the largest performance index is selected as a male parent set. And when the female parent is selected, performing elimination based on the performance indexes of the M candidate sets. That is, M sets are randomly selected, and the set with the largest performance index is selected as the mother set.
Optionally, when the female parent is selected, the female parent is eliminated based on the performance indexes of the M candidate sets. That is, M sets are randomly selected, and the overall diversity is increased by using a Niching scheme. And comparing according to the adjusted performance indexes, and selecting the set with the maximum performance index as a parent set.
Optionally, the manner of obtaining the offspring collection according to the paternal collection and the maternal collection includes: carrying out cross and/or mutation operation on the male parent set and the female parent set to obtain a progeny set; wherein the interleaving operation comprises: randomly selecting one or more weak classifiers from the parent set and the mother set; the mutation operation includes randomly selecting one or more weak classifiers in the non-paternal set and the non-maternal set.
Alternatively, after the male parent set and female parent set are selected, their weak classifiers are divided into two sequences, and crossover and mutation are used to generate one offspring set.
In one embodiment, as shown in FIG. 2, a flow diagram for evolving a weak classifier set is shown. The first generation of N sets is formed by randomly selecting weak classifiers from a weak classifier library and combining, and each set comprises K weak classifiers. The N sets of other generations are all N sub-generation sets generated by the previous generation, and the selection strengths of the male parent and the female parent are both set to F as the optimal values. And when the male parent is selected, eliminating the male parent based on the performance indexes of the F candidate sets. That is, F sets are randomly selected, and the set with the largest performance index is selected as a father set. In selecting the female parent, Niching's protocol was used to increase overall diversity. We compare according to the adjusted performance indicators. And generating a descendant set by using cross mutation, wherein the cross means that a weak classifier is randomly selected from the parent set and the maternal set respectively, and the mutation means that the weak classifier is randomly selected from the weak classifier library with a small probability. And when the evolution is carried out to the ith generation, ending the evolution process, and selecting a set with the highest performance index in the whole process as a set for reasoning and classification.
Step S15: and carrying out reasoning and classification on the face image data according to the optimal weak classifier set to obtain a face information identification result.
Optionally, the face image data is subjected to inference classification according to the weak classifiers in the optimal weak classifier set to obtain a face information identification result.
Optionally, the manner of performing inference classification on the face image data according to the optimal weak classifier set to obtain a face information recognition result includes:
based on the Bayesian probability formula, P (E ═ E)j) Carrying out normalization to obtain an improved Bayesian probability formula, and carrying out inference classification on the face image data according to the optimal weak classifier set to obtain a face information identification result;
the Bayesian probability formula comprises:
wherein s isiDenotes the output of the classifier, E ═ EjIndicating that the known sample E is in class EjAnd T is the optimal weak classifier set and consists of K weak classifiers.
Specifically, for each weak classifier, the statistical tableThe conditional probability P(s) is recordedi＝right|C＝cj) Wherein s isiRepresents the output of the classifier, C ═ CjIndicating that the known sample C is in class CjIn (1). Since the optimal weak classifier set has made decisions independently, our goal is to find P (E ═ E)j|siE T), which can be derived by bayesian probability formula:
wherein T is the optimal weak classifier set, consisting of K weak classifiers. And during classification, the class with the maximum probability is used as a face information identification result of the optimal weak classifier set.
Since the data imbalance greatly affects the derivation result of the Bayesian probability formula, we can obtain the probability by combining P (E ═ E)j) Normalization is used for increasing the weight of the classes with fewer samples to make up for the defect that the classification algorithm is sensitive to data imbalance. However, information on class size is still important for better classification. We add this information to the individual performance indicators through an evolutionary algorithm during the training process.
Optionally, the obtained face information recognition result is entered into a database.
Similar to the principle of the embodiment, the invention provides a face information recognition system based on an evolutionary weak classifier.
Specific embodiments are provided below in conjunction with the attached figures:
fig. 3 shows a schematic structural diagram of a system of a face information identification method based on an evolutionary weak classifier in the embodiment of the present invention.
The system comprises:
a feature extraction module 31 for extracting at least two facial features from the face image data using a general filter;
a weak classifier training module 32, connected to the feature extraction module 31, configured to obtain a plurality of weak classifiers based on the facial feature training and obtain recognition results included in each of the weak classifiers;
a weak classifier set module 33 connected to the weak classifier training module 32 and configured to randomly form a plurality of weak classifier sets including the same number of weak classifiers and obtain performance indexes of the weak classifier sets related to the recognition result;
the evolution module 34 is connected to the weak classifier set module 33 and configured to perform one or more generations of evolution on the weak classifier set according to the performance index of each weak classifier set to obtain an optimal weak classifier set;
and the reasoning and classifying module 35 is connected with the evolving module 34 and is used for performing reasoning and classifying on the face image data according to the optimal weak classifier set to obtain a face information identification result.
Optionally, the feature extraction module 31 extracts at least two facial features from the face image data by using a gradient histogram filter.
Optionally, the facial features include: one or more of an expression feature, a style feature, a gender feature, an age feature, and a facial feature.
Specifically, the expression features include, but are not limited to, expressions such as obstinate/sad, painful, angry, fear, surprise, happiness and aversion; the style characteristics include, but are not limited to, gentle, high-cold, sweet, fashionable and other styles; the gender characteristics include male and female; age characteristics include, but are not limited to: the age stage; the features of the five sense organs include, but are not limited to: current status characteristics of the five sense organs.
Optionally, each face image data has a classification label corresponding to a facial feature.
Optionally, the weak classifier training module 32 is configured to train a plurality of weak classifiers based on any two facial features; an identification result is obtained that each classifier contains a statistical probability of its corresponding facial features.
Specifically, the weak classifier training module 32 trains two arbitrary facial features from the facial features to obtain a weak classifier, and obtains a plurality of weak classifiers in total; and obtaining a recognition result that each classifier contains a statistical probability of its corresponding facial features.
Optionally, the weak classifier set module 33 randomly obtains a plurality of weak classifier sets containing the same number according to the weak classifiers; and deducing and obtaining a performance index for judging the classification accuracy of each weak classifier set according to the identification result of each weak classifier in each weak classifier set.
Optionally, the weak classifier set module 33 compares the recognition result of each weak classifier in each weak classifier set with the actual result of the recognized face image data, to obtain the classification accuracy of each weak classifier, so as to obtain the accuracy of each weak classifier set, and obtain the performance index. Wherein the higher the accuracy, the higher the performance index.
Optionally, the weak classifier set module 33 compares the frequency of each facial feature in the recognition result of each weak classifier in each weak classifier set with the actual result of the recognized face image data, and the frequency of the facial feature corresponding to the actual result is the classification accuracy of the weak classifier, so as to obtain the accuracy of each weak classifier set and obtain the performance index. Wherein the higher the accuracy, the higher the performance index.
Optionally, when performing first-generation evolution on the weak classifier sets according to the performance indexes of the weak classifier sets, the evolution module 34 randomly selects weak classifier sets with an evolution threshold number and performs performance index sequencing to obtain a first-generation optimal weak classifier set; when the weak classifier sets are subjected to multi-generation evolution according to the performance indexes of the weak classifier sets, performing parent selection according to a descendant set of the last generation obtained by multi-time parent selection and with the evolution threshold number as a to-be-selected weak classifier set of the current generation to obtain one or more descendant sets, and taking a set with the highest performance index in the descendant sets of the current generation as an optimal weak classifier set of the current generation.
Optionally, the parent selection mode comprises: randomly selecting a plurality of weak classifier sets to be selected as male parent candidate sets, and randomly selecting female parent candidate sets with the same quantity as the male parent candidate sets from the weak classifier sets to be selected which are not selected as the male parent candidate sets; respectively taking the set with the maximum performance index in the male parent candidate set and the female parent candidate set as a male parent set and a female parent set; and obtaining a progeny set according to the male parent set and the female parent set.
For example, the parental selection approach includes: and setting the selection strength of the male parent and the female parent as M, and eliminating based on the performance indexes of the M candidate sets when the male parent is selected. That is, M sets are randomly selected, and the set with the largest performance index is selected as a male parent set. And when the female parent is selected, performing elimination based on the performance indexes of the M candidate sets. That is, M sets are randomly selected, and the set with the largest performance index is selected as the mother set.
Optionally, the evolution module 34 eliminates the parents based on the performance indexes of the M candidate sets when selecting the parents. That is, M sets are randomly selected, and the overall diversity is increased by using a Niching scheme. And comparing according to the adjusted performance indexes, and selecting the set with the maximum performance index as a parent set.
Optionally, the evolution module 34 performs intersection and/or mutation operations on the paternal set and the maternal set to obtain a progeny set; wherein the interleaving operation comprises: randomly selecting one or more weak classifiers from the parent set and the mother set; the mutation operation includes randomly selecting one or more weak classifiers in the non-paternal set and the non-maternal set.
Optionally, the evolution module 34, after selecting the male parent set and the female parent set, divides their weak classifiers into two sequences, and generates a progeny set using crossover and mutation.
Optionally, the inference classification module 35 performs inference classification on the face image data according to the weak classifiers in the optimal weak classifier set to obtain a face information identification result.
Optionally, the inference classification module 35 bases P (E-E) in the bayesian probability formula onj) Improvements obtained by performing normalizationThe Bayesian probability formula is used for carrying out inference classification on the face image data according to the optimal weak classifier set to obtain a face information identification result;
the Bayesian probability formula comprises:
wherein s isiDenotes the output of the classifier, E ═ EjIndicating that the known sample E is in class EjAnd T is the optimal weak classifier set and consists of K weak classifiers.
Specifically, the inference classification module 35 records the conditional probability P(s) in the statistical table for each weak classifieri＝right|C＝cj) Wherein s isiRepresents the output of the classifier, C ═ CjIndicating that the known sample C is in class CjIn (1). Since the optimal weak classifier set has made decisions independently, our goal is to find P (E ═ E)j|siE T), which can be derived by bayesian probability formula:
wherein T is the optimal weak classifier set, consisting of K weak classifiers. And during classification, the class with the maximum probability is used as a face information identification result of the optimal weak classifier set.
Since the data imbalance greatly affects the derivation result of the Bayesian probability formula, we can obtain the probability by combining P (E ═ E)j) Normalization is used for increasing the weight of the classes with fewer samples to make up for the defect that the classification algorithm is sensitive to data imbalance. However, information on class size is still important for better classification. We add this information to the training process through evolutionary algorithmsIndividual performance index.
Optionally, the inference classification module 35 enters the obtained face information recognition result into a database.
Fig. 4 shows a schematic structural diagram of a face information recognition terminal 40 based on an evolutionary weak classifier in the embodiment of the present invention.
The face information recognition terminal 40 based on the evolutionary weak classifier includes: a memory 41 and a processor 42, the memory 41 being for storing computer programs; the processor 42 runs a computer program to implement the face information identification method based on the evolutionary weak classifier as shown in fig. 1.
Alternatively, the number of the memories 41 may be one or more, the number of the processors 42 may be one or more, and fig. 4 illustrates one example.
Optionally, the processor 42 in the face information recognition terminal 40 based on the evolutionary weak classifier loads one or more instructions corresponding to the process of the application program into the memory 41 according to the steps shown in fig. 1, and the processor 42 runs the application program stored in the first memory 41, so as to implement various functions in the face information recognition method based on the evolutionary weak classifier shown in fig. 1.
Optionally, the memory 41 may include, but is not limited to, a high speed random access memory, a non-volatile memory. Such as one or more magnetic disk storage devices, flash memory devices, or other non-volatile solid-state storage devices; the Processor 42 may include, but is not limited to, a Central Processing Unit (CPU), a Network Processor (NP), and the like; the Integrated Circuit may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component.
Optionally, the Processor 42 may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the Integrated Circuit may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component.
The present invention also provides a computer-readable storage medium storing a computer program, which when executed implements the method for identifying face information based on an evolutionary weak classifier as shown in fig. 1. The computer-readable storage medium may include, but is not limited to, floppy diskettes, optical disks, CD-ROMs (compact disc-read only memories), magneto-optical disks, ROMs (read-only memories), RAMs (random access memories), EPROMs (erasable programmable read only memories), EEPROMs (electrically erasable programmable read only memories), magnetic or optical cards, flash memory, or other type of media/machine-readable medium suitable for storing machine-executable instructions. The computer readable storage medium may be a product that is not accessed by the computer device or may be a component that is used by an accessed computer device.
In summary, the face information identification method, system and terminal based on the evolutionary weak classifier of the present invention are used to solve the limitations of the prior art in the practical application of face identification using a deep learning model, including large storage capacity, high power consumption, high computation delay, and sensitivity to data imbalance between classes. The face information identification method based on the evolutionary weak classifier, which is adopted by the invention, has high accuracy and lightweight attribute, is suitable for both large data sets and small data sets, can ensure that the model is more flexible under the condition of ensuring certain accuracy, can realize real-time reasoning at the edge end, and has increased compatibility and robustness. Therefore, the invention effectively overcomes various defects in the prior art and has high industrial utilization value.
The foregoing embodiments are merely illustrative of the principles of the present invention and its efficacy, and are not to be construed as limiting the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.
Claims (10)
1. A face information identification method based on an evolution weak classifier is characterized by comprising the following steps:
extracting at least two facial features from the face image data by using a general filter;
training based on the facial features to obtain a plurality of weak classifiers and obtaining identification results contained in the weak classifiers;
randomly forming a plurality of weak classifier sets containing the same number of weak classifiers, and obtaining the performance indexes of the weak classifier sets related to the identification result;
performing one or more generations of evolution on the weak classifier set according to the performance indexes of each weak classifier set to obtain an optimal weak classifier set;
and carrying out reasoning and classification on the face image data according to the optimal weak classifier set to obtain a face information identification result.
2. The method for identifying face information based on the evolved weak classifiers according to claim 1, wherein the performing one or more generations of evolutions on the weak classifier sets according to the performance indexes of each weak classifier set to obtain an optimal weak classifier set comprises:
when performing first-generation evolution on the weak classifier sets according to the performance indexes of the weak classifier sets, randomly selecting weak classifier sets with evolution threshold quantity and performing performance index sequencing to obtain a first-generation optimal weak classifier set;
when the weak classifier sets are subjected to multi-generation evolution according to the performance indexes of the weak classifier sets, performing parent selection according to a descendant set of the last generation obtained by multi-time parent selection and with the evolution threshold number as a to-be-selected weak classifier set of the current generation to obtain one or more descendant sets, and taking a set with the highest performance index in the descendant sets of the current generation as an optimal weak classifier set of the current generation.
3. The method for identifying face information based on an evolutionary weak classifier as claimed in claim 1, wherein the facial features comprise: one or more of an expression feature, a style feature, a gender feature, an age feature, and a facial feature.
4. The method for identifying face information based on evolving weak classifiers according to claim 1, wherein the training based on the face features to obtain a plurality of weak classifiers and obtaining the identification results included in each weak classifier comprises:
based on a plurality of weak classifiers respectively trained by any two facial features;
an identification result is obtained that each classifier contains a statistical probability of its corresponding facial features.
5. The method for identifying face information based on evolutionary weak classifiers according to claim 1, wherein the manner of randomly forming a plurality of weak classifier sets including the same number of weak classifiers and obtaining the performance indexes of the respective weak classifier sets related to the identification result is as follows:
randomly obtaining a plurality of weak classifier sets containing the same number according to the weak classifiers;
and deducing and obtaining a performance index for judging the classification accuracy of each weak classifier set according to the identification result of each weak classifier in each weak classifier set.
6. The method for identifying face information based on evolutionary weak classifier as claimed in claim 2, wherein the parent selection mode comprises:
randomly selecting a plurality of weak classifier sets to be selected as male parent candidate sets, and randomly selecting female parent candidate sets with the same quantity as the male parent candidate sets from the weak classifier sets to be selected which are not selected as the male parent candidate sets;
respectively taking the set with the maximum performance index in the male parent candidate set and the female parent candidate set as a male parent set and a female parent set;
and obtaining a progeny set according to the male parent set and the female parent set.
7. The method for identifying face information based on an evolutionary weak classifier as claimed in claim 6, wherein the manner of obtaining the offspring collection according to the male parent collection and the female parent collection comprises:
carrying out cross and/or mutation operation on the male parent set and the female parent set to obtain a progeny set;
wherein the interleaving operation comprises: randomly selecting one or more weak classifiers from the parent set and the mother set; the mutation operation includes randomly selecting one or more weak classifiers in the non-paternal set and the non-maternal set.
8. The method for recognizing face information based on evolutionary weak classifiers according to claim 1, wherein the manner of performing inference classification on the face image data according to the optimal weak classifier set to obtain a face information recognition result comprises:
based on the Bayesian probability formula, P (E ═ E)j) Carrying out normalization to obtain an improved Bayesian probability formula, and carrying out inference classification on the face image data according to the optimal weak classifier set to obtain a face information identification result;
the Bayesian probability formula comprises:
wherein s isiDenotes the output of the classifier, E ═ EjIndicating that the known sample E is in class EjAnd T is the optimal weak classifier set and consists of K weak classifiers.
9. A face information recognition system based on an evolutionary weak classifier, the system comprising:
the feature extraction module is used for extracting at least two facial features from the face image data by using the general filter;
the weak classifier training module is connected with the feature extraction module and used for training and obtaining a plurality of weak classifiers based on the facial features and obtaining the identification results contained in the weak classifiers;
the weak classifier set module is connected with the weak classifier training module and used for randomly forming a plurality of weak classifier sets containing the same number of weak classifiers and obtaining the performance indexes of the weak classifier sets related to the identification result;
the evolution module is connected with the weak classifier set module and is used for carrying out one or more generations of evolution on the weak classifier set according to the performance indexes of each weak classifier set to obtain an optimal weak classifier set;
and the reasoning and classifying module is connected with the evolution module and used for carrying out reasoning and classifying on the face image data according to the optimal weak classifier set to obtain a face information identification result.
10. A face information identification terminal based on an evolution weak classifier is characterized by comprising:
a memory for storing a computer program;
a processor for performing the method of evolving weak classifier based face information recognition according to any of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011015635.0A CN112183302A (en) | 2020-09-24 | 2020-09-24 | Face information identification method, system and terminal based on evolutionary weak classifier |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011015635.0A CN112183302A (en) | 2020-09-24 | 2020-09-24 | Face information identification method, system and terminal based on evolutionary weak classifier |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112183302A true CN112183302A (en) | 2021-01-05 |
Family
ID=73956968
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011015635.0A Pending CN112183302A (en) | 2020-09-24 | 2020-09-24 | Face information identification method, system and terminal based on evolutionary weak classifier |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112183302A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101075291A (en) * | 2006-05-18 | 2007-11-21 | 中国科学院自动化研究所 | Efficient promoting exercising method for discriminating human face |
| CN105160299A (en) * | 2015-07-31 | 2015-12-16 | 华南理工大学 | Human face emotion identifying method based on Bayes fusion sparse representation classifier |
| CN106529503A (en) * | 2016-11-30 | 2017-03-22 | 华南理工大学 | Method for recognizing face emotion by using integrated convolutional neural network |
| CN107273818A (en) * | 2017-05-25 | 2017-10-20 | 北京工业大学 | The selective ensemble face identification method of Genetic Algorithm Fusion differential evolution |
-
2020
- 2020-09-24 CN CN202011015635.0A patent/CN112183302A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101075291A (en) * | 2006-05-18 | 2007-11-21 | 中国科学院自动化研究所 | Efficient promoting exercising method for discriminating human face |
| CN105160299A (en) * | 2015-07-31 | 2015-12-16 | 华南理工大学 | Human face emotion identifying method based on Bayes fusion sparse representation classifier |
| CN106529503A (en) * | 2016-11-30 | 2017-03-22 | 华南理工大学 | Method for recognizing face emotion by using integrated convolutional neural network |
| CN107273818A (en) * | 2017-05-25 | 2017-10-20 | 北京工业大学 | The selective ensemble face identification method of Genetic Algorithm Fusion differential evolution |
Non-Patent Citations (2)
| Title |
|---|
| 王光焰 等: "《塔里木河流域干流水资源配置原理与实践》", 30 June 2019, 武汉大学出版社 * |
| 贾润莹 等: "基于Adaboost和遗传算法的硬盘故障预测模型优化及选择", 《计算机研究与发展》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wen et al. | Ensemble of deep neural networks with probability-based fusion for facial expression recognition | |
| Shao et al. | Feature learning for image classification via multiobjective genetic programming | |
| Liu et al. | Facial expression recognition using hybrid features of pixel and geometry | |
| Ali et al. | Boosted NNE collections for multicultural facial expression recognition | |
| H Mohamed et al. | Improved white blood cells classification based on pre-trained deep learning models | |
| Zhang et al. | Facial smile detection based on deep learning features | |
| Liu et al. | Facial age estimation using a multi-task network combining classification and regression | |
| Liu et al. | Boosting-POOF: boosting part based one vs one feature for facial expression recognition in the wild | |
| Wang et al. | S 3 d: scalable pedestrian detection via score scale surface discrimination | |
| Chergui et al. | Kinship verification through facial images using cnn-based features | |
| Tapia et al. | Clusters of features using complementary information applied to gender classification from face images | |
| Bi et al. | Genetic programming-based evolutionary deep learning for data-efficient image classification | |
| Mohammed et al. | Deep convolution neural network for facial expression recognition | |
| Fan et al. | Genetic programming for image classification: A new program representation with flexible feature reuse | |
| CN111860056A (en) | Blink-based in-vivo detection method and device, readable storage medium and equipment | |
| Zambare et al. | A theoretical study on cnn based cotton crop disease detection | |
| Varshney et al. | Madhubani Art Classification using transfer learning with deep feature fusion and decision fusion based techniques | |
| Karagoz et al. | Analysis of multiobjective algorithms for the classification of multi-label video datasets | |
| Zeng et al. | Local discriminant training and global optimization for convolutional neural network based handwritten Chinese character recognition | |
| Almuashi et al. | Siamese convolutional neural network and fusion of the best overlapping blocks for kinship verification | |
| CN112183302A (en) | Face information identification method, system and terminal based on evolutionary weak classifier | |
| Parameshachari et al. | Plant Disease Detection and Classification Using Transfer Learning Inception Technique | |
| Bharathi et al. | The significance of feature selection techniques in machine learning | |
| CN113887509A (en) | Rapid multi-modal video face recognition method based on image set | |
| Faradibah et al. | Comparison Analysis of Random Forest Classifier, Support Vector Machine, and Artificial Neural Network Performance in Multiclass Brain Tumor Classification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210105 |
|
| RJ01 | Rejection of invention patent application after publication |