WO2021174789A1

WO2021174789A1 - Feature extraction-based image recognition method and image recognition device

Info

Publication number: WO2021174789A1
Application number: PCT/CN2020/112404
Authority: WO
Inventors: 任逍航; 陆进; 陈斌; 宋晨
Original assignee: 平安科技（深圳）有限公司
Priority date: 2020-03-04
Filing date: 2020-08-31
Publication date: 2021-09-10
Also published as: CN111339979B; CN111339979A

Abstract

The present application relates to computer vision processing in artificial intelligence, and provides a feature extraction-based image recognition method, comprising the following steps: in response to an image recognition instruction, obtaining image information of an object to be recognized; extracting a first feature from the image information, and obtaining a first candidate record from a recognition library on the basis of the first feature, wherein the first candidate record comprises a first reference feature, a second reference feature, and an object identifier, and a first similarity between the first reference feature in the first candidate record and the first feature is greater than a first threshold; extracting a second feature from the image information, and calculating a second similarity between the second reference feature in the first candidate record and the second feature; correcting the second similarity on the basis of the first similarity to obtain a corrected similarity; taking the first candidate record having the highest corrected similarity as a target candidate record; and taking the object identifier comprised in the target candidate record as the recognition result of the image information and outputting the recognition result.

Description

Image recognition method and image recognition device based on feature extraction

Cross-references to related applications

This application affirms that it enjoys the priority of the Chinese patent application with the application number 202010142297.0 and the title "Image recognition method and image recognition device based on feature extraction" filed on March 4, 2020. The entire content of the Chinese patent application is by reference The method is incorporated in this application.

Technical field

This application relates to the field of image recognition technology in artificial intelligence, and in particular to an image recognition method, device, computer equipment, and computer storage medium based on feature extraction.

Background technique

Video surveillance occupies an increasingly important position in social production and life. With the continuous improvement of people's safety requirements, higher requirements are also put forward for accurately identifying target pedestrians from video surveillance pictures. The inventor found that pedestrians themselves have a variety of external characteristics, such as appearance, body shape, clothing, carrying things, etc., and these characteristics can be used as identification features to assist pedestrian identification. Since it is difficult for the pedestrians in the video to obtain robust facial features most of the time, in the traditional pedestrian re-identification methods, most of the pedestrians’ clothing is used as the external representation of the re-recognition. However, in practical applications, the clothing of pedestrians will continue to change, such as adding or removing clothing, and frequent occlusions by buildings or other pedestrians, plus the possible color difference of different cameras, will significantly affect the accuracy of pedestrian recognition based on clothing representations. Sex and recall. Therefore, how to provide an accurate and comprehensive pedestrian recognition solution has become an urgent technical problem to be solved by those skilled in the art.

Summary of the invention

The purpose of this application is to provide an image recognition method, device, computer equipment, and computer storage medium that can accurately and quickly recognize a person in an image, so as to solve the above-mentioned defects in the prior art.

To achieve the above objective, this application provides an image recognition method based on feature extraction, which includes the following steps:

In response to the image recognition instruction, obtain the image information of the object to be recognized;

If facial features can be extracted from the image information, extract the facial features, and obtain a first candidate record from a recognition database based on the facial features, where the first candidate record contains facial reference features, Clothing reference features and object identifiers, where the first degree of similarity between the facial reference features in the first candidate record and the facial features is greater than a first threshold;

In the case that the first similarity is less than the second threshold, extracting clothing features from the image information, and calculating the second similarity between the clothing reference feature in the first candidate record and the clothing feature; Wherein the second threshold is greater than the first threshold;

Correcting the second similarity based on the first similarity to obtain a corrected similarity;

Taking the first candidate record with the highest correction similarity as the target candidate record;

The object identifier contained in the target candidate record is output as the recognition result of the image information.

In addition, in order to achieve the above object, this application also proposes an image recognition device based on feature extraction, and the image recognition device based on feature extraction includes:

The image acquisition module is adapted to acquire image information of the object to be identified in response to an image recognition instruction;

The first comparison module is suitable for extracting facial features from the image information when facial features can be extracted from the image information, and obtaining the first candidate record from the recognition database based on the facial features, wherein The first candidate record includes a face reference feature, a clothing reference feature, and an object identifier, and the first similarity between the face reference feature in the first candidate record and the facial feature is greater than a first threshold;

The second comparison module is adapted to extract clothing features from the image information when the first similarity is less than a second threshold, and calculate the second reference feature in the first candidate record and the clothing feature The second degree of similarity between;

The correction module is adapted to correct the second similarity based on the first similarity to obtain the corrected similarity;

The target confirmation module is adapted to use the first candidate record with the highest correction similarity as the target candidate record;

The recognition module is adapted to output the object identifier contained in the target candidate record as the recognition result of the image information.

In addition, in order to achieve the above object, this application also proposes a computer device, the computer device including a memory, a processor, and an image recognition program based on feature extraction that is stored on the memory and can run on the processor, When the processor executes the image recognition program based on feature extraction, the following steps are implemented:

In addition, in order to achieve the above-mentioned object, this application also proposes a computer-readable storage medium that stores an image recognition program based on feature extraction. When the image recognition program based on feature extraction is executed by a processor, the following is achieved: step:

The image recognition method, device, computer equipment, and computer-readable storage medium based on feature extraction provided in this application can accurately and quickly recognize tasks in a monitoring screen. On the basis of traditional clothing feature recognition, this application adds a facial feature recognition step. For images that can obtain facial features, priority is given to facial recognition, and the clothing recognition similarity is corrected according to the similarity of facial recognition. At the same time, facial features and clothing features are combined for recognition, making the recognition results more accurate and complete. In addition, this application can adjust the clothing feature model in the recognition library in real time according to the recognition results of the video images, and fully consider the increase, decrease, and change of clothes in different areas or different time periods, so that the recognition results of this application are more intelligent and humane. .

Description of the drawings

FIG. 1 is a flowchart of Embodiment 1 of the image recognition method of this application;

2 is a schematic diagram of program modules of Embodiment 1 of the image recognition device of this application;

3 is a schematic diagram of the hardware structure of Embodiment 1 of the image recognition device of this application;

4 is a flowchart of Embodiment 2 of the image recognition method of this application;

FIG. 5 is a flowchart of Embodiment 3 of the image recognition method of this application.

Detailed ways

In order to make the purpose, technical solutions, and advantages of this application clearer and clearer, the following further describes the application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the application, and are not used to limit the application. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

Example one

One of the application scenarios of this application is visitor management in office buildings. When a visitor visits an office building, he needs to register at the front desk first and keep his identity information and image information. The identity information can include ID number, name, etc., and the image information includes facial feature information and clothing feature information. The front desk staff will store the visitor’s identity information and image information as a feature record in the recognition library, which will be used as a reference for later image recognition. For example, the feature record contains (identity information, facial feature information, clothing feature information) Three-element array. When visitors enter the office building, their image information will be collected by video (image) monitoring equipment set up at different locations in the office building. When there are many visitors in the office building, this application can determine the identity of the person in the image according to the different image information collected by the monitoring equipment, so as to grasp the dynamic location of the visitor.

Referring to FIG. 1, this embodiment proposes an image recognition method based on feature extraction, which specifically includes the following steps:

S1: In response to the image recognition instruction, obtain image information of the object to be recognized.

The image recognition instruction in this application can be triggered automatically or manually. In the case of automatic triggering, it can be triggered based on time or conditions. For example, set to automatically collect an image every minute and identify the person in the image, or set to collect an image every time the optical sensor is blocked. Recognize people in images, etc.

S2: In the case where facial features can be extracted from the image information, extract the facial features, and obtain a first candidate record from a recognition database based on the first feature, where the first candidate record contains a face With reference to the reference feature, the clothing reference feature, and the object identifier, the first degree of similarity between the facial reference feature in the first candidate record and the facial feature is greater than a first threshold.

This embodiment corresponds to the situation in which facial features can be obtained from the collected image information. Generally speaking, at this time, the visitor is relatively close to the monitoring device and there is no obstruction in the middle.

After the facial features are extracted from the image information, this application compares the extracted facial features with the feature records pre-stored in the recognition library, and obtains the first candidate record whose facial feature similarity is greater than the first threshold from the pre-stored feature records . The first threshold in this application can be flexibly set according to needs, such as 60%, 70%, and so on.

According to the foregoing description, the identification database of this application stores a three-element array containing (identity information, facial feature information, and clothing feature information), and each three-element array corresponds to a feature record.

Therefore, in this application, the process of comparing the extracted facial features with the feature records pre-stored in the recognition library is to obtain the corresponding facial feature information from the feature records of the recognition library, and then compare the extracted facial features with those obtained in the recognition library. The facial feature information is compared, and the first similarity x1 between the two is calculated. If the first similarity is greater than the preset first threshold, the feature record is taken as the first candidate record. Those of ordinary skill in the art understand that there may be one or more first candidate records.

S3: When the first similarity is less than the second threshold, extract clothing features from the image information, and calculate the second similarity between the clothing reference feature in the first candidate record and the clothing feature Degree x2. Wherein the second threshold is greater than the first threshold.

After one or more first candidate records have been obtained in step S2, this step is used to calculate the second difference between the clothing feature extracted from the image information and the clothing feature information stored in the first candidate record. Similarity.

S4: Based on the first similarity x1, the second similarity x2 is corrected to obtain the corrected similarity S.

The steps for calculating the revised similarity S in this application are as follows:

S41: Calculate the correction value λ based on the first degree of similarity x1, and the calculation formula of the correction value λ is:

In the above formula, TH is the first threshold, and α is the control coefficient;

S42: Multiply the second similarity x2 by the correction value λ to obtain the correction similarity S:

S=λx ₂ .

It can be seen from the above calculation formula that the present application actually corrects the second similarity x2 in a forward or reverse direction according to the relationship between the first similarity x1 and the first threshold TH. Generally speaking, when the first similarity x1 is greater than the first threshold TH, the corresponding correction value λ is greater than 1, and the correction to the second similarity x2 is positive; when the first similarity x1 is less than the first threshold TH, The corresponding correction value λ is less than 1, and the correction of the second degree of similarity x2 is reversed.

S5: Use the first candidate record with the highest correction similarity as the target candidate record.

In this step, the first candidate record with the largest corrected similarity S obtained by calculation is taken as the final selected target candidate record.

S6: Output the object identifier contained in the target candidate record as the recognition result of the image information.

In this step, the corresponding identity information is obtained from the target candidate record, and the identity information is output as the final recognition result of the image information obtained in step S1. For example, if the identity information stored in the target candidate record is "Li San, 1980xxxxxxxx", the output recognition result is "Name: Li San; ID number: 1980xxxxxxxx".

S7: Determine whether the sharpness of the image information is greater than a third threshold; if so, replace the clothing reference feature in the target candidate record with the clothing feature extracted from the image information.

This step is used to select high-quality image information of the object to be recognized, and replace the pre-stored clothing feature information in the target candidate record with the clothing feature in the image information.

Since the image information of the object to be identified in this application is the latest collected information, the clothing features in the image information are more real-time, and the clothing features in the image information of the object to be identified are replaced by the pre-stored in the target candidate record. This application can respond to visitor changes or changes in clothing in a timely manner, thereby improving the accuracy of image recognition and making the image recognition process more intelligent.

Please continue to refer to FIG. 2, which shows an image recognition device based on feature extraction. In this embodiment, the image recognition device 10 may include or be divided into one or more program modules, and one or more program modules are stored. It is stored in a storage medium and executed by one or more processors to complete the application and realize the above-mentioned image recognition method. The program module referred to in the present application refers to a series of computer program instruction segments capable of completing specific functions, and is more suitable for describing the execution process of the image recognition device 10 in the storage medium than the program itself. The following description will specifically introduce the functions of each program module in this embodiment:

The image acquisition module 11 is adapted to acquire image information of the object to be identified in response to an image recognition instruction;

The first comparison module 12 is adapted to extract the facial features when the facial features can be extracted from the image information, and obtain the first candidate record from the recognition database based on the facial features, wherein the first candidate record The candidate record contains a face reference feature, a clothing reference feature, and an object identifier, and the first similarity between the face reference feature in the first candidate record and the facial feature is greater than a first threshold;

The second comparison module 13 is adapted to extract clothing features from the image information when the first similarity is less than the second threshold, and calculate the relationship between the clothing reference feature in the first candidate record and the clothing feature The second degree of similarity of, wherein the second threshold is greater than the first threshold;

The correction module 14 is adapted to correct the second similarity based on the first similarity to obtain the corrected similarity;

The target confirmation module 15 is adapted to use the first candidate record with the highest correction similarity as the target candidate record;

The recognition module 16 is adapted to output the object identifier contained in the target candidate record as the recognition result of the image information;

The update module 17 is adapted to determine whether the sharpness of the image information is greater than a third threshold, and if so, replace the clothing reference feature in the first candidate record with the clothing feature extracted from the image information.

According to the image recognition device provided by the present application, the correction module 14 includes:

The correction value unit 141 is adapted to calculate the correction value λ based on the first degree of similarity, and the calculation formula of the correction value λ is:

In the above formula, TH is the first threshold, α is the control coefficient, and x1 is the first degree of similarity;

The correction similarity unit 142 is adapted to multiply the second similarity by the correction value to obtain the corrected similarity S:

S=λx ₂ ;

In the above formula, x2 is the second degree of similarity.

This embodiment also provides a computer device, such as a smart phone, a tablet computer, a notebook computer, a desktop computer, a rack server, a blade server, a tower server, or a cabinet server (including independent servers, or A server cluster composed of multiple servers) and so on. The computer device 20 in this embodiment at least includes but is not limited to: a memory 21 and a processor 22 that can be communicatively connected to each other through a system bus, as shown in FIG. 3. It should be pointed out that FIG. 3 only shows the computer device 20 with components 21-22, but it should be understood that it is not required to implement all the illustrated components, and more or fewer components may be implemented instead.

In this embodiment, the memory 21 (ie, readable storage medium) includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), Read only memory (ROM), electrically erasable programmable read only memory (EEPROM), programmable read only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 21 may be an internal storage unit of the computer device 20, such as a hard disk or a memory of the computer device 20. In other embodiments, the memory 21 may also be an external storage device of the computer device 20, such as a plug-in hard disk, a smart media card (SMC), and a secure digital (Secure Digital, SMC) equipped on the computer device 20. SD) card, flash card (Flash Card), etc. Of course, the memory 21 may also include both an internal storage unit of the computer device 20 and an external storage device thereof. In this embodiment, the memory 21 is generally used to store an operating system and various application software installed in the computer device 20, such as the program code of the image recognition device 10 in the first embodiment, and so on. In addition, the memory 21 can also be used to temporarily store various types of data that have been output or will be output.

In some embodiments, the processor 22 may be a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. The processor 22 is generally used to control the overall operation of the computer device 20. In this embodiment, the processor 22 is used to run the program code or process data stored in the memory 21, for example, to run the image recognition device 10, so as to implement the image recognition method of the first embodiment.

Example two

This embodiment is suitable for situations where facial feature recognition is sufficiently certain. Referring to FIG. 4, this embodiment proposes an image recognition method based on feature extraction, which specifically includes the following steps:

S210: In response to the image recognition instruction, obtain image information of the object to be recognized.

S220: Extract facial features from the image information, and obtain a first candidate record from a recognition database based on the facial features; wherein the first candidate record contains a facial reference feature, a clothing reference feature, and an object identifier. The first similarity between the facial reference feature in a candidate record and the facial feature is greater than the first threshold.

The first candidate record in this application may be a three-element array containing (identity information, facial feature information, clothing feature information).

S230: Determine whether the first similarity is greater than a second threshold, where the second threshold is higher than the first threshold; if so, use the first candidate record with the highest first similarity as a target candidate record.

Those skilled in the art can understand that when the first degree of similarity obtained by comparing facial features is sufficiently large, for example, more than 95%, then we are confident enough to be sure that the identity information of the object to be recognized can be determined only through facial feature recognition. Therefore, At this time, the first candidate record corresponding to the sufficiently large similarity can be directly used as the target candidate record.

S240: Output the object identifier contained in the target candidate record as the recognition result of the image information.

In this embodiment, there is no need to compare clothing features, so that the time for image recognition can be shortened and the efficiency of image recognition can be improved.

Example three

This embodiment is suitable for situations where facial features cannot be obtained due to reasons such as a relatively long shooting distance, too low shooting pixels, or occlusion between the monitoring device and the object to be identified. Referring to FIG. 5, this embodiment proposes an image recognition method based on feature extraction, which specifically includes the following steps:

S310: In response to the image recognition instruction, obtain image information of the object to be recognized.

S320: In the case that facial features cannot be extracted from the image information, extract clothing features from the image information, and obtain a second candidate record from the recognition database based on the clothing features, and the second candidate record is The second degree of similarity between the clothing reference feature and the clothing feature is greater than the fourth threshold.

The process of obtaining the second candidate record from the recognition database based on the second feature includes:

The clothing feature extracted from the image information is compared with each clothing feature information stored in the recognition database for the second similarity. When the second similarity is greater than the fourth threshold, the candidate record corresponding to the first similarity is taken as The second candidate record. The fourth threshold in this application can be flexibly set according to actual needs, for example, set to 80%, 90%, and so on.

S330: Use the second candidate record with the highest second similarity as the target candidate record.

When only one recognition result needs to be output, this application uses the second candidate record with the highest second similarity as the target candidate record. Those skilled in the art can understand that sometimes the recognition result is not unique, but a plurality of different recognition results can be provided for reference. At this time, it is possible to sort from high to low according to the data of the second similarity, and use the top second candidate records as the target candidate records.

S340: Output the object identifier contained in the target candidate record as the recognition result of the image information.

This embodiment also provides a computer-readable storage medium. The computer-readable storage medium may be non-volatile or volatile, such as flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic Memory, magnetic disks, optical disks, servers, App application malls, etc., have computer programs stored thereon, and corresponding functions are realized when the programs are executed by the processor. The computer-readable storage medium of this embodiment is used to store the image recognition device 10, and when executed by a processor, it implements the image recognition method of the first embodiment.

The serial numbers of the foregoing embodiments of the present application are for description only, and do not represent the superiority or inferiority of the embodiments.

Any process or method description in the flowchart or described in other ways herein can be understood as a module, segment or part of code that includes one or more executable instructions for implementing specific logical functions or steps of the process , And the scope of the preferred embodiments of the present application includes additional implementations, which may not be in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order according to the functions involved. This should It is understood by those skilled in the art to which the embodiments of the present application belong.

A person of ordinary skill in the art can understand that all or part of the steps carried in the method of the foregoing embodiments can be implemented by a program instructing relevant hardware to complete, and the program can be stored in a computer-readable medium. When executed, it includes one of the steps of the method embodiment or a combination thereof.

In the description of this specification, descriptions with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples" or "some examples" etc. mean the specific features described in conjunction with the embodiments or examples. The structure, material or feature is included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above-mentioned terms do not necessarily refer to the same embodiment or example. Moreover, the described specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

Through the description of the above implementation manners, those skilled in the art can clearly understand that the above-mentioned embodiment method can be implemented by means of software plus the necessary general hardware platform, of course, it can also be implemented by hardware, but in many cases the former is better.的实施方式。

The above are only the preferred embodiments of the application, and do not limit the scope of the patent for this application. Any equivalent structure or equivalent process transformation made using the content of the description and drawings of the application, or directly or indirectly applied to other related technical fields , The same reason is included in the scope of patent protection of this application.

Claims

An image recognition method based on feature extraction, which includes the following steps:

In response to the image recognition instruction, obtain the image information of the object to be recognized;

If facial features can be extracted from the image information, extract the facial features, and obtain a first candidate record from a recognition database based on the facial features, where the first candidate record contains facial reference features, Clothing reference features and object identifiers, where the first degree of similarity between the facial reference features in the first candidate record and the facial features is greater than a first threshold;

In the case that the first similarity is less than the second threshold, extracting clothing features from the image information, and calculating the second similarity between the clothing reference feature in the first candidate record and the clothing feature; Wherein the second threshold is greater than the first threshold;

Correcting the second similarity based on the first similarity to obtain a corrected similarity;

Taking the first candidate record with the highest correction similarity as the target candidate record;

The object identifier contained in the target candidate record is output as the recognition result of the image information.
The image recognition method according to claim 1, wherein the step of correcting the second similarity based on the first similarity to obtain the corrected similarity comprises:

The correction value λ is calculated based on the first similarity, and the calculation formula of the correction value λ is:

In the above formula, TH is the first threshold, α is the control coefficient, and x 1 is the first degree of similarity;

Multiply the second similarity by the correction value to obtain the corrected similarity S:

S=λx 2 ;

In the above formula, x 2 is the second degree of similarity.
The image recognition method according to claim 1, wherein after the step of extracting facial features from the image information and obtaining the first candidate record from the recognition library based on the facial features, the method further comprises:

In a case where the first similarity is greater than or equal to a second threshold, use the first candidate record with the highest first similarity as the target candidate record;

The object identifier contained in the target candidate record is output as the recognition result of the image information.
The image recognition method according to claim 1 or 3, wherein after the step of outputting the object identifier contained in the target candidate record as the recognition result of the image information, the method further comprises:

Judging whether the sharpness of the image information is greater than a third threshold;

If yes, replace the clothing reference feature in the first candidate record with the clothing feature extracted from the image information.
The image recognition method according to claim 4, wherein after the step of obtaining image information of the object to be recognized in response to the image recognition instruction, the method further comprises:

In the case that the facial features cannot be extracted from the image information, the clothing features are extracted from the image information, and the second candidate record is obtained from the recognition database based on the clothing features, and the clothing in the second candidate record The second degree of similarity between the reference feature and the clothing feature is greater than the fourth threshold;

Taking the second candidate record with the highest second degree of similarity as the target candidate record;

The object identifier contained in the target candidate record is output as the recognition result of the image information.
An image recognition device based on feature extraction, which includes:

The image acquisition module is adapted to acquire image information of the object to be identified in response to an image recognition instruction;

The first comparison module is suitable for extracting facial features from the image information when facial features can be extracted from the image information, and obtaining the first candidate record from the recognition database based on the facial features, wherein The first candidate record includes a face reference feature, a clothing reference feature, and an object identifier, and the first similarity between the face reference feature in the first candidate record and the facial feature is greater than a first threshold;

The second comparison module is adapted to extract clothing features from the image information when the first similarity is less than the second threshold, and calculate the second reference feature in the first candidate record and the clothing feature The second degree of similarity between

The correction module is adapted to correct the second similarity based on the first similarity to obtain the corrected similarity;

The target confirmation module is adapted to use the first candidate record with the highest correction similarity as the target candidate record;

The recognition module is adapted to output the object identifier contained in the target candidate record as the recognition result of the image information.
The image recognition device according to claim 6, wherein the correction module comprises:

The correction value unit is adapted to calculate the correction value λ based on the first degree of similarity, and the calculation formula of the correction value λ is:

In the above formula, TH is the first threshold, α is the control coefficient, and x 1 is the first degree of similarity;

The modified similarity unit is adapted to multiply the second similarity by the modified value to obtain the modified similarity S:

S=λx 2 ;

In the above formula, x 2 is the second degree of similarity.
The image recognition device according to claim 6, further comprising:

The second target confirmation module is adapted to use the first candidate record with the highest first similarity as the target candidate record when the first similarity is greater than or equal to a second threshold;

The second recognition module is adapted to output the object identifier contained in the target candidate record as the recognition result of the image information.
The image recognition device according to claim 8, further comprising:

The update module is adapted to determine whether the sharpness of the image information is greater than a third threshold, and if so, replace the clothing reference feature in the first candidate record with the clothing feature extracted from the image information.
The image recognition device according to claim 9, further comprising:

The third comparison module is suitable for extracting clothing features from the image information when facial features cannot be extracted from the image information, and obtaining a second candidate record from the recognition database based on the clothing features, the The second degree of similarity between the clothing reference feature in the second candidate record and the clothing feature is greater than the fourth threshold;

The third target confirmation module is adapted to use the second candidate record with the highest second degree of similarity as the target candidate record;

The third recognition module is adapted to output the object identifier contained in the target candidate record as the recognition result of the image information.
A computer device includes a memory, a processor, and a computer program stored on the memory and capable of running on the processor, wherein the processor implements the following steps of an image recognition method based on feature extraction when the processor executes the computer program:

In response to the image recognition instruction, obtain the image information of the object to be recognized;

If facial features can be extracted from the image information, extract the facial features, and obtain a first candidate record from a recognition database based on the facial features, where the first candidate record contains facial reference features, Clothing reference features and object identifiers, where the first degree of similarity between the facial reference features in the first candidate record and the facial features is greater than a first threshold;

In the case that the first similarity is less than the second threshold, extracting clothing features from the image information, and calculating the second similarity between the clothing reference feature in the first candidate record and the clothing feature; Wherein the second threshold is greater than the first threshold;

Correcting the second similarity based on the first similarity to obtain a corrected similarity;

Taking the first candidate record with the highest correction similarity as the target candidate record;

The object identifier contained in the target candidate record is output as the recognition result of the image information.
The computer device according to claim 11, wherein the step of correcting the second similarity based on the first similarity to obtain the corrected similarity comprises:

The correction value λ is calculated based on the first similarity, and the calculation formula of the correction value λ is:

In the above formula, TH is the first threshold, α is the control coefficient, and x 1 is the first degree of similarity;

Multiply the second similarity by the correction value to obtain the corrected similarity S:

S=λx 2 ;

In the above formula, x 2 is the second degree of similarity.
The computer device according to claim 11, wherein, after the step of extracting facial features from the image information and obtaining the first candidate record from a recognition library based on the facial features, the method further comprises:

In a case where the first similarity is greater than or equal to a second threshold, use the first candidate record with the highest first similarity as the target candidate record;

The object identifier contained in the target candidate record is output as the recognition result of the image information.
The computer device according to claim 11 or 13, wherein after the step of outputting the object identifier contained in the target candidate record as the recognition result of the image information, the method further comprises:

Judging whether the sharpness of the image information is greater than a third threshold;

If yes, replace the clothing reference feature in the first candidate record with the clothing feature extracted from the image information.
The computer device according to claim 14, wherein after the step of obtaining image information of the object to be recognized in response to the image recognition instruction, the method further comprises:

In the case that the facial features cannot be extracted from the image information, the clothing features are extracted from the image information, and the second candidate record is obtained from the recognition database based on the clothing features, and the clothing in the second candidate record The second degree of similarity between the reference feature and the clothing feature is greater than the fourth threshold;

Taking the second candidate record with the highest second degree of similarity as the target candidate record;

The object identifier contained in the target candidate record is output as the recognition result of the image information.
A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps of an image recognition method based on feature extraction are implemented:

In response to the image recognition instruction, obtain the image information of the object to be recognized;

If facial features can be extracted from the image information, extract the facial features, and obtain a first candidate record from a recognition database based on the facial features, where the first candidate record contains facial reference features, Clothing reference features and object identifiers, where the first degree of similarity between the facial reference features in the first candidate record and the facial features is greater than a first threshold;

In the case that the first similarity is less than the second threshold, extracting clothing features from the image information, and calculating the second similarity between the clothing reference feature in the first candidate record and the clothing feature; Wherein the second threshold is greater than the first threshold;

Correcting the second similarity based on the first similarity to obtain a corrected similarity;

Taking the first candidate record with the highest correction similarity as the target candidate record;

The object identifier contained in the target candidate record is output as the recognition result of the image information.
15. The computer-readable storage medium according to claim 16, wherein the step of correcting the second similarity based on the first similarity to obtain the corrected similarity comprises:

The correction value λ is calculated based on the first similarity, and the calculation formula of the correction value λ is:

In the above formula, TH is the first threshold, α is the control coefficient, and x 1 is the first degree of similarity;

Multiply the second similarity by the correction value to obtain the corrected similarity S:

S=λx 2 ;

In the above formula, x 2 is the second degree of similarity.
16. The computer-readable storage medium according to claim 16, wherein, after the step of extracting facial features from the image information and obtaining the first candidate record from a recognition library based on the facial features, the method further comprises:

In a case where the first similarity is greater than or equal to a second threshold, use the first candidate record with the highest first similarity as the target candidate record;

The object identifier contained in the target candidate record is output as the recognition result of the image information.
The computer-readable storage medium according to claim 16 or 18, wherein after the step of outputting the object identifier contained in the target candidate record as the recognition result of the image information, the method further comprises:

Judging whether the sharpness of the image information is greater than a third threshold;

If yes, replace the clothing reference feature in the first candidate record with the clothing feature extracted from the image information.
The computer-readable storage medium according to claim 19, wherein after the step of obtaining image information of the object to be recognized in response to the image recognition instruction, the method further comprises:

In the case that the facial features cannot be extracted from the image information, the clothing features are extracted from the image information, and the second candidate record is obtained from the recognition database based on the clothing features, and the clothing in the second candidate record The second degree of similarity between the reference feature and the clothing feature is greater than the fourth threshold;

Taking the second candidate record with the highest second degree of similarity as the target candidate record;

The object identifier contained in the target candidate record is output as the recognition result of the image information.