CN117611907A

CN117611907A - Method and device for detecting flip image and training model, electronic equipment and medium

Info

Publication number: CN117611907A
Application number: CN202311629931.3A
Authority: CN
Inventors: 沈家润
Original assignee: Agricultural Bank of China
Current assignee: Agricultural Bank of China
Priority date: 2023-11-30
Filing date: 2023-11-30
Publication date: 2024-02-27

Abstract

The embodiment of the invention discloses a method and a device for detecting a flip image and training a model, electronic equipment and a medium, wherein the method for detecting the flip image comprises the following steps: acquiring an original file to be detected; the original file to be detected comprises a target image to be detected; preprocessing the original file to be detected to obtain a normalized file to be detected; determining the target image to be detected from the normalized file to be detected; invoking a reproduction image detection service to perform reproduction detection on the target image to be detected through a reproduction image detection model of the reproduction image detection service to obtain a reproduction detection result; and screening the image of the shooting target according to the shooting detection result. The technical scheme of the embodiment of the invention can realize the automatic detection function of the flipped image, thereby improving the intelligence and the automation degree of the detection flow.

Description

Method and device for detecting flip image and training model, electronic equipment and medium

Technical Field

The embodiment of the invention relates to the technical fields of image detection, artificial intelligence and automatic flow, in particular to a method, a device, electronic equipment and a medium for detecting a flip image and training a model.

Background

Currently, business processes for many business systems typically require review involving certificates and tickets. Taking a credit card system as an example, the manual processing links of credit card examination and approval comprise a credit front plate of common examination, general line examination and approval, general line quality inspection, risk examination, merchant examination, linkage and the like, and a large amount of manual processing is needed in each operation link.

As more and more businesses involving personal finance and government affairs are transferred from offline to online, business processors often need to upload images of certificates or tickets such as identity cards, business licenses, and invoices to a platform for auditing. RPA (Robotic Process Automation ) is a process automation that is widely used in business systems in various fields by mimicking the manual operation of an end user on a computer, providing another way to automate the end user manual operation process. Correspondingly, when the quality inspection process is automatically executed through the RPA, some situations need to check whether all images are the turnups according to the service requirements. If the judgment is that the piece is turned over, the examination is not passed, otherwise, the examination is passed.

The inventors have found that the following drawbacks exist in the prior art in the process of implementing the present invention: at present, the RPA does not have the function of detecting the image by the turn-over, when the image needs to be subjected to the turn-over detection, the image often needs to be manually processed, a flow breakpoint is formed, and the automation degree of the auditing flow is reduced.

Disclosure of Invention

The embodiment of the invention provides a method, a device, electronic equipment and a medium for detecting and training a flip image, which can realize the automatic detection function of the flip image, thereby improving the intelligence and the automation degree of a detection flow.

According to an aspect of the present invention, there is provided a method for detecting a flipped image, applied to an RPA system, including:

acquiring an original file to be detected; the original file to be detected comprises a target image to be detected;

preprocessing the original file to be detected to obtain a normalized file to be detected;

determining the target image to be detected from the normalized file to be detected;

invoking a reproduction image detection service to perform reproduction detection on the target image to be detected through a reproduction image detection model of the reproduction image detection service to obtain a reproduction detection result;

and screening the image of the shooting target according to the shooting detection result.

According to another aspect of the present invention, there is provided a model training method including:

adopting an unbalanced sampling method to select a set number of positive sample image data and negative sample image data to construct a training sample set;

randomly selecting current image sample data from the training sample set;

Extracting local texture features of a sample image from the current image sample data through an balanced LBP operator;

inputting the local texture features of the sample image into a plurality of classification models, and outputting a reproduction detection result of the sample data of the current image through each classification model so as to realize a reproduction detection training process of each classification model;

screening target classification models from the classification models according to the flap detection training results of the classification models;

constructing a flipped image detection model according to the balanced LBP operator and the target classification model;

the image detection model is used for constructing a reproduction image detection service; the reproduction image detection service is applied to the reproduction image detection method described in the first aspect.

According to another aspect of the present invention, there is provided a device for detecting a flipped image, configured in an RPA system, including:

the original file to be detected acquisition module is used for acquiring the original file to be detected; the original file to be detected comprises a target image to be detected;

the normalized file to be detected acquisition module is used for preprocessing the original file to be detected to obtain a normalized file to be detected;

The target image to be detected determining module is used for determining the target image to be detected from the normalized file to be detected;

the device comprises a flip detection result acquisition module, a flip detection module and a display module, wherein the flip detection result acquisition module is used for calling a flip image detection service to perform flip detection on the target image to be detected through a flip image detection model of the flip image detection service to obtain a flip detection result;

and the overturning target image screening module is used for screening overturning target images according to the overturning detection result.

According to another aspect of the present invention, there is provided a model training apparatus comprising:

the training sample set construction module is used for selecting positive sample image data and negative sample image data with set quantity by adopting an unbalanced sampling method to construct a training sample set;

the current image sample data selecting module is used for randomly selecting current image sample data from the training sample set;

the image local texture feature extraction module is used for extracting sample image local texture features from the current image sample data through an balanced LBP operator;

the classification model training module is used for inputting the local texture characteristics of the sample image into a plurality of classification models, and outputting the reproduction detection result of the current image sample data through each classification model so as to realize the reproduction detection training process of each classification model;

The target classification model screening module is used for screening target classification models from the classification models according to the training results of the flap detection of the classification models;

the overturn image detection model construction module is used for constructing an overturn image detection model according to the balanced LBP operator and the target classification model;

According to another aspect of the present invention, there is provided an electronic apparatus including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the method for detecting a flipped image or the method for training a model according to any of the embodiments of the present invention.

According to another aspect of the present invention, there is provided a computer readable storage medium storing computer instructions for causing a processor to implement the method for detecting a flip image or the method for training a model according to any of the embodiments of the present invention when executed.

According to the embodiment of the invention, the RPA system is used for preprocessing an original to-be-detected file comprising the to-be-detected target image to obtain a normalized to-be-detected file, the to-be-detected target image is determined from the normalized to-be-detected file, and then the flap image detection service is called, so that the to-be-detected target image is subjected to flap detection through a flap image detection model of the flap image detection service to obtain a flap detection result, and then the flap target image is screened according to the flap detection result. Before the RPA system performs the test of the reproduction image on the original file to be tested, firstly, a set number of positive sample image data and negative sample image data are selected by adopting an unbalanced sampling method to construct a training sample set, and current image sample data are randomly selected from the training sample set, further, sample image local texture features are extracted from the current image sample data through an balanced LBP operator, so that the sample image local texture features are input into a plurality of classification models, the reproduction test result of the current image sample data is output through each classification model, so that the reproduction test training process of each classification model is realized, the target classification model is screened from each classification model according to the reproduction test training result of each classification model, and finally, the reproduction image test model is constructed according to the balanced LBP operator and the target classification model. The technical scheme can solve the problem that the process automation degree is reduced because the conventional RPA system cannot automatically detect the turnup image, and can realize the automatic detection function of the turnup image, thereby improving the intelligence and the automation degree of the detection process.

It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the invention or to delineate the scope of the invention. Other features of the present invention will become apparent from the description that follows.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of a method for detecting a flipped image according to an embodiment of the present invention;

fig. 2 is a schematic view of a deployment effect of a flip image detection model according to a first embodiment of the present invention;

FIG. 3 is a schematic diagram of a credit card handling business process according to an embodiment of the invention;

FIG. 4 is a flowchart of a model training method according to a second embodiment of the present invention;

FIG. 5 is a diagram showing an example of the original LBP calculation according to the second embodiment of the present invention;

fig. 6 is a schematic diagram of a device for detecting a flipped image according to a third embodiment of the present invention;

FIG. 7 is a schematic diagram of a model training apparatus according to a fourth embodiment of the present invention;

fig. 8 is a schematic structural diagram of an electronic device according to a fifth embodiment of the present invention.

Detailed Description

In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.

It should be noted that the term "object" and the like in the description of the present invention and the claims and the above drawings are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

Example 1

Fig. 1 is a flowchart of a method for detecting a flipped image, which is provided in an embodiment of the present invention, and the embodiment may be suitable for a case where the flipped image detection is automatically implemented by an RPA system, where the method may be implemented by a flipped image detection device, and the device may be implemented by software and/or hardware, and may be generally integrated in an RPA system, where the RPA system may be installed and operated in an electronic device, and the electronic device may be a terminal device or a server device, and is used with a service for providing a flipped image detection function. Accordingly, as shown in fig. 1, the method includes the following operations:

s110, acquiring an original file to be detected; the original file to be detected comprises an object image to be detected.

The original file to be detected may be an unprocessed original file that needs to be inspected. The target image to be detected may be an image for which a flip image detection is required.

In an alternative embodiment of the present invention, the original document to be detected may include, but is not limited to, a credit card system document to be inspected, a house transaction application system document to be inspected, or an online payment system document to be inspected, etc.

It will be appreciated that the files to be reviewed for the online system are typically saved in the online system. Taking a credit card system as an example, various data generated in a credit card business scene of a credit card center are stored in each system, and the RPA system can simulate a user to log in, access, download, operate other operations and the like on an online system, so that examination and approval feed materials are obtained and used as original files to be detected.

S120, preprocessing the original file to be detected to obtain a normalized file to be detected.

The normalized file to be detected may be a normalized file obtained by preprocessing an original file to be detected by the RPA system.

Because the original files to be detected are often disordered, the RPA system can pretreat the original files to be detected by methods of sorting, archiving, renaming, sorting and the like after the original files to be detected are obtained, and the normalized files to be detected are obtained through sorting, so that convenience is provided for the subsequent detection steps.

S130, determining the target image to be detected from the normalized file to be detected.

Correspondingly, after the normalized file to be detected is obtained by arrangement, the RPA system can further determine the target image to be detected, which needs to be subjected to the flip detection, from the normalized file to be detected. It is understood that the number of target images to be detected in the normalized file to be detected may be one or more.

In an optional embodiment of the present invention, after the determining the target image to be detected from the normalized file to be detected, the method may further include: performing information code conversion on the target image to be detected to obtain an image conversion code; transmitting the image conversion code to the reproduction image detection service; the image reproduction detection service is used for carrying out information code conversion on the image conversion code to obtain the target image to be detected, and inputting the target image to be detected into the image reproduction detection model for reproduction detection.

The image conversion code may be code information obtained by performing information code conversion on the target image to be detected. Alternatively, the type of image conversion encoding may be BASE64 (encoding in the form of a character string). The reproduction image detection service may perform reproduction detection on the image through the reproduction image detection model.

In order to facilitate the transmission of image information, the RPA system may perform information transcoding on each target image to be detected after determining the target image to be detected, for example, convert the target image to an image transcoding code of BASE64, and then send the target image to be detected to the flip image detection service in the form of the image transcoding code. Correspondingly, after receiving the image conversion code, the image conversion code can be subjected to information coding conversion to obtain each target image to be detected, and the converted target images to be detected are input into a reproduction image detection model for reproduction detection. The advantages of this arrangement are: the data of the target image to be detected is transmitted in the form of image conversion coding, so that the efficiency and accuracy of image data transmission can be improved.

S140, calling a reproduction image detection service to carry out reproduction detection on the target image to be detected through a reproduction image detection model of the reproduction image detection service, so as to obtain a reproduction detection result.

Correspondingly, after the RPA system determines the target image to be detected, the image conversion code of the target image to be detected can be sent to the image-flipping detection service to call the image-flipping detection service, and the image-flipping detection is carried out on the target image to be detected through the image-flipping detection model of the image-flipping detection service, so that a result of the image-flipping detection is obtained.

Optionally, when the number of the target images to be detected is multiple, the RPA system may invoke multiple image detection models to perform image detection on the target images to be detected in parallel through the image detection service, so as to improve efficiency of image detection.

In an alternative embodiment of the invention, the flipped image detection model is applied to a flash application framework, and the flipped image detection model runs concurrently through the guricorn service multithreading and realizes load balancing processing through load balancing equipment.

Fig. 2 is a schematic view of a deployment effect of a flip image detection model according to an embodiment of the present invention. In one specific example, as shown in FIG. 2, the flip image detection model may be applied to a flash application framework. The flash is a lightweight Web application framework written by using Python (a computer programming language), the flash framework is adopted, a flip image detection model obtained by model training and constructing steps is called, a service is started, and the service of the flip image detection model is realized. Gunicorn (Green Unicorn) is an unix (You Nisi operating system) based system, python based WSGI (Web Server Gateway Interface ) HTTP (Hyper Text Transfer Protocol, hypertext transfer protocol) server. By using the guricorn, the multithreading operation of the flip image detection service can be realized, and the high concurrence of the flip image detection service can be realized. And finally, connecting a plurality of devices through load balancing equipment, so that the service for detecting the flip image meets the service requirement.

In an alternative embodiment of the invention, the flipped image detection model includes an equalization LBP (Local Binary Pattern ) operator and a target classification model, which can be used to: extracting image local texture features from the target image to be detected through the balanced LBP operator; and inputting the local texture features of the image into the target classification model so as to output a flip detection result of the target image to be detected through the target classification model.

The target classification model can be one of classification models selected according to a model training result after comparing and training each classification model in the training process of the flip image detection model. It will be appreciated that the target classification model may be the most accurate model of the classification models.

A flip image is an image that is a picture taken in a screen (typically a cell phone screen). In a business scenario, a user may not submit a document to be detected using a certificate original, but rather hold a photograph of the certificate in a cell phone to provide relevant material, which is not allowed in certain provisioning specifications and therefore needs to be detected and identified. Due to the mismatch in resolution of the photographed screen and the photographing apparatus, a difference in texture occurs in the flipped image compared to the non-flipped image. Features can thus be extracted using an image processing method that can describe the texture of a picture.

In the embodiment of the invention, the flipped image detection model can be composed of an equalization LBP operator and a target classification model. The balanced LBP operator can be used for extracting image local texture features of the target image to be detected. Correspondingly, the flipped image detection model can extract image local texture features of the target image to be detected through an equalization LBP operator, the extracted image local texture features are input into the target classification model, classification and identification are carried out through the target classification model according to the image local texture features, and a flipped detection result of whether the target image to be detected is the flipped image is output.

And S150, screening the image of the shooting target according to the shooting detection result.

The image of the object to be flipped may be an image obtained by a flipping method.

Correspondingly, the RPA system can acquire the overturn detection results of the target image to be detected, which are fed back by the overturn image detection service, and further screen the overturn target image from the target image to be detected according to each overturn detection result.

Fig. 3 is a schematic diagram of a credit card setting business process provided in a first embodiment of the present invention, in a specific example, taking the credit card setting business process as an example, as shown in fig. 3, the RPA system may simulate a user to manually access and operate web pages and software at a computer end, generate, acquire and arrange relevant materials for credit card setting, and invoke a turnip image detection service to acquire a turnip detection result, so as to realize automatic full coverage of quality inspection and inspection points of a general line.

Specifically, various data generated in a credit card business scene of a credit card center are stored in each system, and the RPA flow simulates user login, access, downloading and operation, so that examination and approval of the material of the incoming part are obtained. The material of the material feeding part is often disordered, and the RPA obtains the normalized material by the methods of sorting, filing, renaming, sorting and the like, thereby providing convenience for the subsequent detection and inspection steps. In the step of detecting the turnup image, the RPA robot reads the file, converts and encodes the picture information, invokes the turnup image detection service, finally receives the response of the turnup image detection service, processes the response content and returns the final turnup detection result. And the RPA sorts the materials identified as the turnup pieces by judging the turnup detection result, and sends the materials to manual processing. And meanwhile, the materials identified as non-flap pieces can be tidied, and a subsequent examination and approval process is carried out.

Therefore, the technical scheme solves the problem of computer vision correlation of the RPA system deficiency by utilizing an artificial intelligence technology, extracts the picture characteristics by utilizing a texture-sensitive image processing method such as an equalization LBP operator, and can effectively solve the problem of detecting the flip image by combining a machine learning algorithm. Meanwhile, the RPA technology is utilized to simulate the operation of a person, realize the detection function of a turnup image, and realize the full-flow intellectualization and automation of the generation, the acquisition, the arrangement, the detection and the transfer of file materials. In other words, the RPA system in the embodiment of the invention can effectively and intelligently identify the flipped image by means of the flipped image detection model, and the breakpoint of the detection flow is opened by utilizing the mode of RPA+AI, so that the automatic full coverage of the detection flow is realized.

According to the embodiment of the invention, the RPA system is used for preprocessing an original to-be-detected file comprising the to-be-detected target image to obtain a normalized to-be-detected file, the to-be-detected target image is determined from the normalized to-be-detected file, and then the flap image detection service is called, so that the to-be-detected target image is subjected to flap detection through a flap image detection model of the flap image detection service to obtain a flap detection result, and then the flap target image is screened according to the flap detection result. The technical scheme can solve the problem that the process automation degree is reduced because the conventional RPA system cannot automatically detect the turnup image, and can realize the automatic detection function of the turnup image, thereby improving the intelligence and the automation degree of the detection process.

Example two

Fig. 4 is a flowchart of a model training method provided in a second embodiment of the present invention, where the present embodiment is applicable to a case of constructing a flipped image detection model by equalizing an LBP operator and training various classification models, and the method may be performed by a model training apparatus, where the apparatus may be implemented by software and/or hardware, and may be generally integrated into an electronic device, where the electronic device may be a terminal device or a server device, so long as the electronic device may be used to perform model training. Accordingly, as shown in fig. 4, the method includes the following operations:

S210, adopting an unbalanced sampling method to select positive sample image data and negative sample image data with set numbers to construct a training sample set.

The set number can be set according to actual requirements, and the number of the positive sample image data and the number of the negative sample image data can be different so as to meet unbalanced sampling requirements. The positive sample image data may be sample data of a flip image construction. The negative sample image data may be sample data of a non-flipped image construction.

S220, randomly selecting current image sample data from the training sample set.

Considering that in the actual service scenario, the AI (Artificial Intelligence ) technology has risks of accuracy and recognition error rate, and the positive sample hit rate and the negative sample hit rate should be preferentially ensured whether the positive sample hit rate can meet the service requirement. Therefore, when training the flip image detection model, an unbalanced sampling method can be adopted, and a set number of positive sample image data and negative sample image data are selected to construct a training sample set. For example, 17000 positive sample images and 3000 negative sample images can be selected, and the number of positive sample images accounts for 85% of the total number of samples. And finally, returning a judging result of the picture by the flip image detection model, wherein '1' represents a positive sample, judging as a flip image, and '0' represents a negative sample, and judging as a normal image. The basic sample data of the flipped image detection model can be derived from the document photographs randomly sampled and annotated for review approval. By unbalanced sampling, the positive sample hit rate can be increased.

After the training sample set is constructed, the current image sample data can be randomly selected from the training sample set and used for training the flip image detection model in the current turn.

S230, extracting local texture features of the sample image from the current image sample data through an equalization LBP operator.

LBP is an operator feature used to describe local texture features of an image. Fig. 5 is a schematic diagram of an example of original LBP calculation according to the second embodiment of the present invention. In a specific example of the original LBP calculation, as shown in fig. 5, in a black-and-white picture matrix of 3*3 pixels, different pixel values represent different shades of gray, and the surrounding 8 pixels are marked as 1 larger (or equal) than the middle pixel value and the small is marked as 0, so that a binary image (threshhold) is obtained, and then a binary string 10001111 is obtained in the clockwise direction. The pixel value of the intermediate point is thus replaced by 241. The order of calculating LBP is not hard, but only a quantization formula, the same order is maintained in the same process.

Wherein, the calculation formula of LBP is:wherein->In the above formula, P represents the number of sampling points, R represents the radius of a circular area formed by the sampling points, and s (x) represents a threshold function: g _p Represents a gray value g representing a p-th pixel except a center pixel in the window _c The gray value representing the center pixel is represented.

The balanced LBP mode is to add a specification based on the original LBP mode, and the rule is: a binary sequence changes no more than 2 times from 0 to 1 or from 1 to 0. Such as: 10100000 is 3 times and is not one uniform pattern. There are 58 unitorm patterns in total (2 for a number of 0 changes, 0 for a number of 1 changes, and 56 for a number of 2 changes) in all 8-bit binary numbers. The balanced LBP mode plays a role in reducing dimension, and is beneficial to providing an effective feature layer for training a later flip image detection model.

In the development using the python language, the texture analysis operator from the self-contained in the skin. The local_binary_pattern (image, P, R, method= 'default') function may be called to extract a feature layer after gray processing is performed on an original image. The number P of the circularly symmetric neighbor set points and the circle radius R are respectively set as shown in the following table, 4 characteristic layers are constructed as shown in the table 1, and the characteristic layers are combined together to form a final characteristic layer for providing training for a later flip image detection model.

TABLE 1 characteristic layer list

P	R	method
			8	1.0	uniform
16	2.0	uniform
			24	3.0	uniform
24	4.0	uniform

S240, inputting the local texture features of the sample image into a plurality of classification models, and outputting the reproduction detection result of the current image sample data through each classification model so as to realize the reproduction detection training process of each classification model.

S250, screening target classification models from the classification models according to the training results of the flap detection of the classification models.

Alternatively, the local texture features of the sample image may be input to multiple classification models, such as logistic regression (Logistic Regression), support vector machines (Support Vector Machine, SVM), random forest (random-forest), etc., to perform contrast training on each classification model. Each classification model is trained by adopting local texture feature characteristics of a sample image, such as the final effective feature layer obtained in the feature extraction step, training is carried out on a training set, model optimal parameters are determined by using 5-fold cross validation, and the final performance of each classification model is evaluated on a test set. Illustratively, the model preference parameters (other parameters remain default) and behavior of the three classification models are shown in Table 2 below:

table 2 Classification model vs training results List

Illustratively, as shown in Table 2, the support vector machine has the highest model accuracy. Therefore, the support vector machine can be used as the target classification model. Alternatively, the classification model obtained by training the training set obtained by adopting the positive and negative unbalance sampling method can be stored through jobilib (a software package for converting a Python code into a parallel computing mode), so as to obtain a model file.

S260, constructing a flipped image detection model according to the balanced LBP operator and the target classification model.

The image detection model is used for constructing a reproduction image detection service; the reproduction image detection service is applied to the reproduction image detection method according to any embodiment of the invention.

Correspondingly, after training to obtain a target classification model, a connection relationship between the balanced LBP operator and the target classification model can be established to construct a final flip image detection model.

Before an RPA system performs the test of a reproduction image on an original file to be tested, firstly adopting an unbalanced sampling method to select positive sample image data and negative sample image data with set quantity to construct a training sample set, randomly selecting current image sample data from the training sample set, further extracting sample image local texture features from the current image sample data through an balanced LBP operator, inputting the sample image local texture features into a plurality of classification models, outputting the reproduction test result of the current image sample data through each classification model to realize the reproduction test training process of each classification model, screening target classification models from each classification model according to the reproduction test training result of each classification model, and finally constructing and obtaining the reproduction image test model according to the balanced LBP operator and the target classification model. Correspondingly, after model training is completed, an original file to be detected including a target image to be detected can be preprocessed through an RPA system to obtain a normalized file to be detected, the target image to be detected is determined from the normalized file to be detected, and then a flap image detection service is called to perform flap detection on the target image to be detected through a flap image detection model of the flap image detection service to obtain a flap detection result, and then the flap target image is screened according to the flap detection result. The technical scheme can solve the problem that the process automation degree is reduced because the conventional RPA system cannot automatically detect the turnup image, and can realize the automatic detection function of the turnup image, thereby improving the intelligence and the automation degree of the detection process.

In the technical scheme of the disclosure, the related processes of collecting, storing, using, processing, transmitting, providing, disclosing and the like of the personal information (such as credit card information and the like) of the user accord with the regulations of related laws and regulations, and the public order is not violated.

It should be noted that any permutation and combination of the technical features in the above embodiments also belong to the protection scope of the present invention.

Example III

Fig. 6 is a schematic diagram of a device for detecting a flipped image according to a third embodiment of the present invention, where the flipped image detecting device may be configured in an RPA system, and the RPA system may be installed and operated in an electronic device, and the electronic device may be a terminal device or a server device, and is used in conjunction with a service for providing a flipped image detecting function.

Accordingly, as shown in fig. 6, the apparatus for detecting a flip image includes: an original to-be-detected file obtaining module 310, a normalized to-be-detected file obtaining module 320, a to-be-detected target image determining module 330, a flip detection result obtaining module 340, and a flip target image screening module 350, wherein:

an original file to be detected obtaining module 310, configured to obtain an original file to be detected; the original file to be detected comprises a target image to be detected;

A normalized file to be detected obtaining module 320, configured to pre-process the original file to be detected to obtain a normalized file to be detected;

a target image to be detected determining module 330, configured to determine the target image to be detected from the normalized file to be detected;

a flip detection result obtaining module 340, configured to invoke a flip image detection service to perform flip detection on the target image to be detected through a flip image detection model of the flip image detection service, so as to obtain a flip detection result;

and a flip target image screening module 350, configured to screen a flip target image according to the flip detection result.

Optionally, the flipped image detection model is applied to a flash application framework, and the flipped image detection model runs concurrently through a guricorn service multithread, and load balancing processing is achieved through load balancing equipment.

Optionally, the apparatus further includes an image conversion code sending module, configured to: performing information code conversion on the target image to be detected to obtain an image conversion code; transmitting the image conversion code to the reproduction image detection service; the image reproduction detection service is used for carrying out information code conversion on the image conversion code to obtain the target image to be detected, and inputting the target image to be detected into the image reproduction detection model for reproduction detection.

Optionally, the flipped image detection model includes an equalizing LBP operator and a target classification model, and the flipped image detection model is used for: extracting image local texture features from the target image to be detected through the balanced LBP operator; and inputting the local texture features of the image into the target classification model so as to output a flip detection result of the target image to be detected through the target classification model.

Optionally, the original document to be detected comprises a credit card system document to be censored.

The image detection device can execute the image detection method provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method. Technical details which are not described in detail in the present embodiment can be referred to the method for detecting a flipped image provided in any embodiment of the present invention.

Since the above-described apparatus for detecting a flip image is an apparatus capable of executing the method for detecting a flip image in the embodiment of the present invention, based on the method for detecting a flip image in the embodiment of the present invention, those skilled in the art can understand the specific implementation of the apparatus for detecting a flip image in the embodiment of the present invention and various modifications thereof, so how the apparatus for detecting a flip image implements the method for detecting a flip image in the embodiment of the present invention will not be described in detail herein. The apparatus used by those skilled in the art to implement the method for detecting a flipped image in the embodiments of the present invention is within the scope of protection intended in the present application.

Example IV

Fig. 7 is a schematic diagram of a model training apparatus according to a fourth embodiment of the present invention, as shown in fig. 7, where the apparatus includes: a training sample set construction module 410, a current image sample data selection module 420, an image local texture feature extraction module 430, a classification model training module 440, a target classification model screening module 450, and a flipped image detection model construction module 460, wherein:

A training sample set construction module 410, configured to select a set number of positive sample image data and negative sample image data to construct a training sample set by adopting an unbalanced sampling method;

a current image sample data selecting module 420, configured to randomly select current image sample data from the training sample set;

an image local texture feature extraction module 430, configured to extract a sample image local texture feature from the current image sample data by equalizing an LBP operator;

the classification model training module 440 is configured to input the local texture features of the sample image to a plurality of classification models, and output a tap detection result of the current image sample data through each classification model, so as to implement a tap detection training process for each classification model;

the target classification model screening module 450 is configured to screen a target classification model from the classification models according to the training results of the flap detection of each classification model;

a flipped image detection model construction module 460, configured to construct a flipped image detection model according to the balanced LBP operator and the target classification model;

The model training device can execute the model training method provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method. Technical details not described in detail in this embodiment may be referred to the model training method provided in any embodiment of the present invention.

Since the model training apparatus described above is an apparatus capable of executing the model training method in the embodiment of the present invention, based on the model training method described in the embodiment of the present invention, those skilled in the art can understand the specific implementation of the model training apparatus of the embodiment and various modifications thereof, so how the model training apparatus implements the model training method in the embodiment of the present invention will not be described in detail herein. The apparatus used by those skilled in the art to implement the model training method in the embodiments of the present invention is within the scope of the present application.

Example five

Fig. 8 shows a schematic diagram of the structure of an electronic device 10 that may be used to implement an embodiment of the invention. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. Electronic equipment may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the inventions described and/or claimed herein.

As shown in fig. 8, the electronic device 10 includes at least one processor 11, and a memory, such as a Read Only Memory (ROM) 12, a Random Access Memory (RAM) 13, etc., communicatively connected to the at least one processor 11, in which the memory stores a computer program executable by the at least one processor, and the processor 11 may perform various appropriate actions and processes according to the computer program stored in the Read Only Memory (ROM) 12 or the computer program loaded from the storage unit 18 into the Random Access Memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 may also be stored. The processor 11, the ROM 12 and the RAM 13 are connected to each other via a bus 14. An input/output (I/O) interface 15 is also connected to bus 14.

Various components in the electronic device 10 are connected to the I/O interface 15, including: an input unit 16 such as a keyboard, a mouse, etc.; an output unit 17 such as various types of displays, speakers, and the like; a storage unit 18 such as a magnetic disk, an optical disk, or the like; and a communication unit 19 such as a network card, modem, wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information/data with other devices via a computer network, such as the internet, and/or various telecommunication networks.

The processor 11 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various processors running machine learning model algorithms, digital Signal Processors (DSPs), and any suitable processor, controller, microcontroller, etc. The processor 11 performs the various methods and processes described above, such as a snap image detection method or a model training method.

In some embodiments, the flip image detection method or the model training method may be implemented as a computer program tangibly embodied on a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program may be loaded and/or installed onto the electronic device 10 via the ROM 12 and/or the communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the above-described flip image detection method or model training method may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform the flipped image detection method or the model training method in any other suitable way (e.g., by means of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.

A computer program for carrying out methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions/acts specified in the flowchart and/or block diagram block or blocks to be implemented. The computer program may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of the present invention, a computer-readable storage medium may be a tangible medium that can contain, or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer readable storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, the computer readable storage medium may be a machine readable signal medium. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) through which a user can provide input to the electronic device. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, speech input, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), blockchain networks, and the internet.

The computing system may include clients and servers. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty and weak service expansibility in the traditional physical hosts and VPS service are overcome.

Claims

1. The method for detecting the flip image is characterized by being applied to a robot process automation RPA system and comprising the following steps:

2. The method of claim 1, wherein the flipped image detection model is applied to a flash application framework, and wherein the flipped image detection model is concurrently run through a guricorn service multithreading and load balancing processing is implemented through a load balancing device.

3. The method of claim 1, further comprising, after said determining said target image to be detected from said normalized target file:

performing information code conversion on the target image to be detected to obtain an image conversion code;

Transmitting the image conversion code to the reproduction image detection service;

the image reproduction detection service is used for carrying out information code conversion on the image conversion code to obtain the target image to be detected, and inputting the target image to be detected into the image reproduction detection model for reproduction detection.

4. The method of claim 1, wherein the flipped image detection model comprises an equalized local binary pattern LBP operator and a target classification model, the flipped image detection model to:

extracting image local texture features from the target image to be detected through the balanced LBP operator;

and inputting the local texture features of the image into the target classification model so as to output a flip detection result of the target image to be detected through the target classification model.

5. The method of any of claims 1-4, wherein the original document to be detected comprises a credit card system document to be inspected.

6. A method of model training, comprising:

Randomly selecting current image sample data from the training sample set;

the image detection model is used for constructing a reproduction image detection service; the reproduction image detection service is applied to the reproduction image detection method according to any one of claims 1 to 5.

7. A flip image detection device, configured in an RPA system, comprising:

8. A model training device, comprising:

9. An electronic device, the electronic device comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the method of detecting a flip image as claimed in any one of claims 1 to 5 or the method of model training as claimed in claim 6.

10. A computer readable storage medium storing computer instructions for causing a processor to perform the method of detecting a flip image as claimed in any one of claims 1 to 5 or the method of model training as claimed in claim 6.