CN113971664A

CN113971664A - SSD-based service auditing method and device, electronic equipment and storage medium

Info

Publication number: CN113971664A
Application number: CN202111266601.3A
Authority: CN
Inventors: 黄琳钧; 刘鹏; 刘玉宇; 王健宗
Original assignee: Ping An Technology Shenzhen Co Ltd
Current assignee: Ping An Technology Shenzhen Co Ltd
Priority date: 2021-10-28
Filing date: 2021-10-28
Publication date: 2022-01-25

Abstract

The invention provides a service auditing method and device based on an SSD, an electronic device and a storage medium, comprising the following steps: acquiring information to be audited, wherein the information to be audited comprises a total amount to be audited, a service type to be audited, a top image and a side image; inputting the top image and the side image into an SSD model, generating a first target frame on the top image, identifying denomination information, generating a second target frame on the side image, determining a target total amount according to the denomination information and the number of the second target frame, and determining a target service type by performing character identification on the second target frame; and when the type of the service to be audited is matched with the target service type, obtaining an audit result by comparing the total amount to be audited with the target total amount. According to the technical scheme of the embodiment, the target total and the target service can be identified by utilizing the SSD model, the information required by the auditing is automatically acquired in an image identification mode, an information basis is provided for the automatic auditing, manual operation can be effectively reduced, and the service auditing efficiency and the service intellectualization are improved.

Description

SSD-based service auditing method and device, electronic equipment and storage medium

Technical Field

The invention relates to the field of Single Shot multi box detectors (SSD) of artificial intelligence, in particular to a service auditing method and device, electronic equipment and a storage medium based on the SSD.

Background

In a financial institution such as a bank, cash deposit and withdrawal is a basic service, and in the cash deposit and withdrawal service, determination of the amount of cash is the most important step. According to the traditional method, a teller bundles cash according to a specified quantity, counts and inputs the cash according to the bundle number, submits the cash to a remote authorization center for video verification, and authorizes the teller to execute subsequent operations after counting and confirmation by the remote authorization center. Although cash counting can be realized, manual operation is excessively relied on, and in order to ensure that errors do not occur, a teller and a remote authorization center need to spend a long time for confirmation, so that the efficiency of business handling is low.

Because the amount of each bundle of cash is usually determined, for example, each common 100 sheets is a bundle, and each bundle of cash is bound by a waist bar, with the development of artificial intelligence technology, some technical schemes of identifying the waist bar by using image identification technology and determining the total amount of cash by determining the number of the cash bundles appear, although the identification efficiency of the amount of cash is effectively improved, the existing scheme still needs a teller to perform subsequent processes after determining the amount of cash, and the intelligence degree is not high.

Disclosure of Invention

The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.

The embodiment of the invention provides a SSD-based service auditing method, device, electronic equipment and storage medium, which can improve the efficiency and automation degree of service auditing.

In a first aspect, an embodiment of the present invention provides a service auditing method based on an SSD, including:

acquiring information to be audited, wherein the information to be audited comprises a total amount to be audited, a service type to be audited and a cash image, and the cash image comprises a top image and a side image;

inputting the top image and the side image into a pre-trained SSD model;

carrying out digital detection on the top image through the SSD model to generate a first target frame, and identifying denomination information from the first target frame;

performing waist bar detection on the side image through the SSD model, generating a second target frame for each waist bar, determining a target total amount according to the number of the second target frames and the denomination information, and determining a target service type through character recognition on the second target frame;

and when the service type to be audited is matched with the target service type, obtaining an audit result by comparing the total amount to be audited with the target total amount.

In some embodiments, the performing, by the SSD model, waist bar detection on the side image, and generating a second target frame for each waist bar includes:

performing banknote bundle detection on the side images through the SSD model, and generating a third target frame for each banknote bundle;

and carrying out waist strip detection in each third target frame to generate the second target frame.

In some embodiments, before said performing waist bar detection in each of said third target boxes, said method further comprises:

performing image comparison on two adjacent third target frames to obtain an image overlapping ratio;

and when the image overlapping ratio is larger than a preset ratio, combining the two third target frames corresponding to the image overlapping ratio.

In some embodiments, the obtaining an audit result by comparing the total to be audited with the target total includes:

when the total amount to be audited is matched with the target total amount, determining that the auditing result is that the auditing is passed;

and when the total amount to be audited is not matched with the target total amount, auditing is stopped and prompt information is generated, and the manual auditing result obtained after the prompt information is generated is determined as the auditing result.

In some embodiments, prior to said inputting said top image and said side images to a pre-trained SSD model, said method further comprises:

performing image recognition on the top image, determining a first cash area in the top image, and cutting the top image according to the first cash area;

and performing image recognition on the side images, determining a second cash area in the side images, and cutting the top images according to the second cash area.

In some embodiments, the digitally detecting the top image by the SSD model, generating a first destination box, comprises:

acquiring a preset digital reference area;

and carrying out digital detection in the first cash area according to the digital reference area, and if a digital is detected, generating the first target frame.

In some embodiments, after the obtaining of the information to be audited, the method further includes:

acquiring a preset size threshold;

determining that the top image is valid when the size of the top image is greater than or equal to the size threshold;

and when the size of the side image is larger than or equal to the size threshold value, determining that the side image is effective.

In a second aspect, an embodiment of the present invention provides a service auditing apparatus based on an SSD, including:

the system comprises a collecting unit, a processing unit and a processing unit, wherein the collecting unit is used for obtaining information to be audited, the information to be audited comprises a total amount to be audited, a service type to be audited and a cash image, and the cash image comprises a top image and a side image;

the image input unit is used for inputting the top image and the side image into a pre-trained SSD model;

the first identification unit is used for carrying out digital detection on the top image through the SSD model, generating a first target frame and identifying denomination information from the first target frame;

the second identification unit is used for carrying out waist strip detection on the side image through the SSD model, generating a second target frame aiming at each waist strip, determining a target total amount according to the number of the second target frames and the denomination information, and determining a target service type through character identification on the second target frames;

and the checking unit is used for comparing the total amount to be checked with the target total amount to obtain a checking result when the service type to be checked is matched with the target service type.

In a third aspect, an embodiment of the present invention provides an electronic device, including: a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the SSD-based traffic auditing method when executing the computer program.

In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, which stores computer-executable instructions for performing the SSD-based service auditing method according to the first aspect.

The embodiment of the invention comprises the following steps: acquiring information to be audited, wherein the information to be audited comprises a total amount to be audited, a service type to be audited and a cash image, and the cash image comprises a top image and a side image; inputting the top image and the side image into a pre-trained SSD model; carrying out digital detection on the top image through the SSD model to generate a first target frame, and identifying denomination information from the first target frame; performing waist bar detection on the side image through the SSD model, generating a second target frame for each waist bar, determining a target total amount according to the number of the second target frames and the denomination information, and determining a target service type through character recognition on the second target frame; and when the service type to be audited is matched with the target service type, obtaining an audit result by comparing the total amount to be audited with the target total amount. According to the technical scheme of the embodiment, the target total amount can be identified by using the SSD model, the target service is determined by identifying the characters on the waist bar, and the information required by the auditing is automatically acquired in an image identification mode, so that an information basis is provided for the automatic auditing, manual operation can be effectively reduced, and the efficiency and the intellectualization of the service auditing are improved.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Drawings

The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the example serve to explain the principles of the invention and not to limit the invention.

FIG. 1 is a flow chart of a SSD-based service auditing method provided by one embodiment of the present invention;

FIG. 2 is a diagram of an exemplary SSD model according to another embodiment of the present invention;

FIG. 3 is a diagram of a second target box provided by another embodiment of the present invention;

FIG. 4 is a flow chart of generating a second target box provided by another embodiment of the present invention;

FIG. 5 is a schematic diagram of a third target box provided by another embodiment of the present invention;

FIG. 6 is a flow chart of merging third target blocks provided by another embodiment of the present invention;

FIG. 7 is a flow chart of determining audit results provided by another embodiment of the present invention;

FIG. 8 is a flow diagram of cropping an image provided by another embodiment of the present invention;

FIG. 9 is a schematic illustration of a first cash section and a second cash section provided by another embodiment of the present invention;

FIG. 10 is a flow chart of generating a first target box provided by another embodiment of the present invention;

FIG. 11 is a flow chart for determining an image validity based on a size threshold provided by another embodiment of the present invention;

fig. 12 is a block diagram of an SSD-based service auditing apparatus according to another embodiment of the present invention;

fig. 13 is a device diagram of an electronic apparatus according to another embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

It should be noted that although functional blocks are partitioned in a schematic diagram of an apparatus and a logical order is shown in a flowchart, in some cases, the steps shown or described may be performed in a different order than the partitioning of blocks in the apparatus or the order in the flowchart. The terms "first," "second," and the like in the description, in the claims, or in the drawings described above, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.

The invention provides a service auditing method and device based on an SSD (solid State disk), electronic equipment and a storage medium, wherein the service auditing method based on the SSD comprises the following steps: acquiring information to be audited, wherein the information to be audited comprises a total amount to be audited, a service type to be audited and a cash image, and the cash image comprises a top image and a side image; inputting the top image and the side image into a pre-trained SSD model; carrying out digital detection on the top image through the SSD model to generate a first target frame, and identifying denomination information from the first target frame; performing waist bar detection on the side image through the SSD model, generating a second target frame for each waist bar, determining a target total amount according to the number of the second target frames and the denomination information, and determining a target service type through character recognition on the second target frame; and when the service type to be audited is matched with the target service type, obtaining an audit result by comparing the total amount to be audited with the target total amount. According to the technical scheme of the embodiment, the target total amount can be identified by using the SSD model, the target service is determined by identifying the characters on the waist bar, and the information required by the auditing is automatically acquired in an image identification mode, so that an information basis is provided for the automatic auditing, manual operation can be effectively reduced, and the efficiency and the intellectualization of the service auditing are improved.

The embodiment of the application can acquire and process related data based on an artificial intelligence technology. Among them, Artificial Intelligence (AI) is a theory, method, technique and application system that simulates, extends and expands human Intelligence using a digital computer or a machine controlled by a digital computer, senses the environment, acquires knowledge and uses the knowledge to obtain the best result.

The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.

It should be noted that the data in the embodiment of the present invention may be stored in a server, and the server may be an independent server, or may be a cloud server that provides basic cloud computing services such as cloud service, a cloud database, cloud computing, a cloud function, cloud storage, Network service, cloud communication, middleware service, domain name service, security service, Content Delivery Network (CDN), big data, and an artificial intelligence platform.

As shown in fig. 1, fig. 1 is a method for auditing a service based on an SSD according to an embodiment of the present invention, which includes, but is not limited to, the following steps:

step S110, obtaining information to be audited, wherein the information to be audited comprises a total amount to be audited, a service type to be audited and a cash image, and the cash image comprises a top image and a side image;

step S120, inputting the top image and the side image into a pre-trained SSD model;

step S130, carrying out digital detection on the top image through an SSD model to generate a first target frame, and identifying denomination information from the first target frame;

step S140, carrying out waist strip detection on the side surface image through an SSD model, generating a second target frame aiming at each waist strip, determining a target total amount according to the number and the denomination information of the second target frames, and determining a target service type by carrying out character recognition on the second target frames;

and S150, when the service type to be audited is matched with the target service type, obtaining an audit result by comparing the total amount to be audited with the target total amount.

It should be noted that, in order to implement service auditing, the SSD-based service auditing method of the present embodiment may be applied to a remote authorization center, and certainly, may also be applied to a cloud server, and only has a permission to audit a cash service, and a skilled person in the art has a motivation to adjust a specific carrier according to an actual situation, which is not limited herein, and if not described specifically, the following embodiments are described by taking a remote authorization center as an example.

It should be noted that the message to be audited may be obtained through a terminal device, for example, a terminal device is in communication connection with a remote authorization center, in the process of handling cash business, the teller inputs the message to be audited in the terminal device, and uploads the message to the remote authorization center through a network, and the terminal device may be a common computer, a common mobile device, and the like, and can input information, which is not limited herein. It can be understood that the message to be audited may be directly generated by the terminal device, or may be obtained by issuing a pre-configured information template by the remote authorization center, and the teller inputs corresponding information according to the information template at the terminal device, for example, the information template includes a plurality of input options, which respectively correspond to the total amount to be audited, the type of the service to be audited, the top image and the side image, the teller manually inputs the total amount to be audited and the type of the service to be audited, and then shoots the top image and the side image of the cash to be processed by the shooting device, for example, shoots the cash by a high-speed shooting instrument commonly used by a financial institution, and uploads the corresponding image through the information template, thereby completing the reporting of the information to be audited. It can be understood that the information template can effectively ensure that the input information to be audited conforms to the specification, for example, the type of the service to be audited can be selected in the form of a pull-down menu, so as to ensure that the remote authorization center can be correctly matched with the corresponding service, and the layout and the input mode of the information template can be adjusted according to the actual requirement, which is not limited in this embodiment.

It should be noted that, since the technical solution of this embodiment is applied to a scenario of a financial institution, cash is usually bundled according to a certain standard, for example, each 100 cash is usually a bundle, and of course, if an actual requirement is met, the amount of each bundle of cash can also be adjusted, and the amount of each bundle of cash is reported through the information template described in the above embodiment, so that the total amount is calculated in the remote authorization center by taking each bundle of cash as a unit, for example, 50 cash is taken as a bundle, corresponding information is correspondingly input in the information template and reported to the remote authorization center, so that the remote authorization center can recognize that the audit is performed by taking 50 cash as a bundle for total amount calculation, and other amounts are analogized and are not described herein.

It should be noted that the front and back sides of the cash are printed with the value of the amount of money, and the bundle of cash in the financial institution must have the same denomination, so that the top image can be taken by a high-speed camera, so that the denomination of the cash corresponding to the service can be determined by digital detection, for example, the denomination of the bundle of cash is 100 yuan if the number obtained by digital detection and recognition is 100 yuan. In addition, for most financial institutions, the waist bar for binding cash is usually printed with a service identifier corresponding to the cash, such as deposit or withdrawal, and the service identifier is usually located on the side of the waist bar, so that the acquired side image includes the service identifier, thereby providing an image basis for character recognition. It is worth noting that in the field of image processing based on artificial intelligence, it is likely that the collected image does not have a target to be detected, and the image is generally required to be preprocessed or pre-judged, but because the operation specification in the financial field is strict, and the image collection is completed by a teller operating high-speed shooting instrument, the situation that the top image does not have a denomination and the side image does not have a service identifier is difficult to occur, so that the step of judging whether the image has the image to be detected or not can be omitted in the embodiment, the calculation step of the SSD model can be reduced, and the detection efficiency can be improved.

As can be understood by those skilled in the art, the SSD technology is a target detection algorithm for directly predicting the coordinates and the category of a target frame, has no process of generating a candidate frame, can perform target detection by using a multi-scale feature map, improves the target detection precision, and keeps higher efficiency. In addition, the main network structure of the SSD model is usually VGG16, two full connection layers FC6 and FC7 of the VGG-16 are replaced by convolution layers conv6 and conv7, and then 4 convolution layers are added to construct a network structure.

Referring to fig. 2, in the network structure of the SSD model, the feature layer is selected from convolutional layers conv4_3, conv7, conv8_2, conv9_2, conv10_2 and conv11_2 of VGG-16, where a convolution kernel of conv4_3 may be 3 × 3 × 1024, a convolution kernel of conv7 may be 1 × 01 × 11024, a convolution kernel of conv8_2 may be 1 × 1 × 256, a convolution kernel of conv9_2 may be 1 × 1 × 128, a convolution kernel of conv10_2 may be 1 × 1 × 128, a convolution kernel of conv11_2 may be 1 × 1 × 128, and when the size of the input image is 300 based on the above 6 feature layers for target detection, the 6 feature layer scales may be scale [38,19,10,5,3,1, 10, 3,1, 3, 10, 1, 3,1, 3, 10, 3, 2, and 3, respectively]. In addition, the SSD model may be configured with a prior box that may generate multiple anchor boxes for each pixel of the selected feature layer, firstThe length-width ratio of the test frame can be selected from h/w ═ 1,2,1/2,3,1/3]And a width to height ratio of 1 for two dimensions, where h is height and w is width; different feature layers correspond to different prior frame sizes, smin is 0.2, smax is 0.9 and conv4_2 layer sizes are calculated independently relative to the size of an original image, and other size calculations are obtained by the following formula:

wherein k is ∈ [1, m ]]M is the number of selected characteristic layers; in addition, the loss function of the SSD model may include a category loss and a LOC loss, wherein the category loss may use a softmax function, the selected portion of the anchor is used as the loss, and the LOC loss may use a positive case of the anchor containing the target as the regression loss.

It should be noted that, since the SSD model has the characteristics of fast detection speed and high accuracy for detecting small objects, the SSD model can be used to quickly complete digital detection from the top image, so as to generate a first target frame in the region corresponding to the denomination of cash, determine denomination information through simple image recognition, recognize each waist bar from the side image, and generate a second target frame for each waist bar, where the number of the second target frames is the number of the waist bars, so that the total amount can be calculated based on this, for example, in the side image shown in fig. 3, the number of the recognized second target frames 300 is 10, the denomination recognized from the first target frame is 100, and the number of each bundle of banknotes is 100, and the total amount is 100 × 10 — 100000.

It should be noted that, based on the above description, the text information in the second target frame is the service identifier, so that the target service type can be determined through simple text recognition, and a specific recognition manner is not an improvement made in this embodiment, and is not described herein any more, and text recognition can be implemented in a specific area.

It is worth noting that the target total amount and the target service type are obtained by cash image recognition, so that the recognition process of the SSD model is equivalent to a process of applying an artificial intelligence technology to replace manual confirmation of a remote control center in the prior art, and the speed of SSD model detection greatly exceeds the speed of manual confirmation, so that the auditing process can be effectively improved. It can be understood that, since the total amount to be audited and the type of the service to be audited are obtained through teller input, when the type of the service to be audited is consistent with the target service type, the auditing result can be determined by judging whether the total amount to be audited is consistent with the target total amount obtained through identification, and for the financial field, if the total amount to be audited is consistent with the target total amount, the auditing is passed, which is not described in detail herein.

In addition, in an embodiment, referring to fig. 4, step S140 of the embodiment shown in fig. 1 further includes, but is not limited to, the following steps:

step S410, performing money bundle detection on the side images through the SSD model, and generating a third target frame for each money bundle;

step S420, performing waist bar detection in each third target frame to generate a second target frame.

Note that, the banknote bundles are usually stacked and placed in the manner shown in fig. 5, and the positions of the waist strips of each banknote bundle are not necessarily the same, so in order to ensure the accuracy of waist strip detection, bundle detection may be performed from the side images to generate a third target frame for each bundle, and the bundle detection may be determined by a distance between the cash, for example, the distance between two pieces of cash is greater than a certain threshold, it may be determined that the two pieces of cash are respectively assigned to different money bundles, or the upper and lower boundaries of the waist bar may be identified, the upper boundary of the waist bar is determined as the upper boundary of the third target frame, and the lower boundary of the waist bar is determined as the lower boundary of the third target frame, as shown in fig. 5, through the bank note bundle detection, 10 third target frames 510 can be identified, the specific detection mode can be adjusted according to actual requirements, and the bank note bundle detection can be realized.

It is to be noted that after the third target frame is obtained, the detection of the square image in the third target frame may be performed, so as to implement the detection of the waist stripe, and the specific manner may be adjusted according to actual requirements.

In addition, in an embodiment, referring to fig. 6, before performing step S420 of the embodiment shown in fig. 4, the following steps are further included, but not limited to:

step S610, performing image comparison on two adjacent third target frames to obtain an image overlapping ratio;

step S620, merging the two third target frames corresponding to the image overlap ratio when the image overlap ratio is greater than the preset ratio.

It should be noted that, since cash is paper money, even if the paper money is placed in a stacking manner, a situation of inclination may occur, and at this time, a situation that the money bundles corresponding to the two third target frames are the same money bundle is likely to occur, and when the money bundle is inclined, the two adjacent third target frames are likely to overlap with each other, so as to avoid that the cash placement inclination affects the accuracy of detection, image comparison may be performed on the adjacent third target frames after the third target frames are generated, and when the obtained image overlap ratio is greater than the preset ratio, the two adjacent third target frames actually correspond to the same money bundle, and the two third target frames may be combined.

It should be noted that the numerical value of the preset proportion may be adjusted according to the timing requirement, for example, the preset proportion may be appropriately increased when the size of the third target frame is large, and those skilled in the art know how to adjust the preset proportion, so that the SSD model can accurately identify the banknote bundle overlapping caused by the cash placing inclination.

In addition, in an embodiment, referring to fig. 7, step S150 of the embodiment shown in fig. 1 further includes, but is not limited to, the following steps:

step S710, when the total amount to be audited is matched with the target total amount, determining that the auditing result is that the auditing is passed;

and S720, stopping auditing and generating prompt information when the total amount to be audited is not matched with the target total amount, and determining the manual auditing result obtained after the prompt information is generated as the auditing result.

It should be noted that, for the financial field, the main point of the cash business audit lies in the amount verification, so that when the total amount to be audited matches the target total amount, the total amount to be audited input by the teller can be determined to be correct, thereby determining that the audit result is that the audit is passed, without manual confirmation again, effectively reducing the audit process and improving the audit efficiency.

It should be noted that, when the total amount to be audited is not matched with the target total amount, an error may occur in the identification of the SSD model, or an error may occur in the total amount to be audited, in this case, two consecutive identifications may be performed by the SSD model, and when both the two identifications are matched, it may be determined that the first unmatched root is the error of the SSD model, and it is determined that the audit result is the audit pass; of course, based on the rigor degree of the financial industry, the automatic auditing process can also be stopped, the auditing personnel is prompted to intervene by generating the prompt information, the manual auditing result is input in the remote authorization center in a manual auditing mode, and the auditing process is completed by taking the manual auditing result as the final auditing result, so as to ensure the accuracy of the business auditing.

In addition, in an embodiment, referring to fig. 8, before performing step S420 of the embodiment shown in fig. 4, the following steps are further included, but not limited to:

step S810, performing image recognition on the top image, determining a first cash area in the top image, and cutting the top image according to the first cash area;

step S820, performing image recognition on the side image, determining a second cash area in the side image, and cropping the top image according to the second cash area.

It should be noted that since the shooting range of the high-speed scanner is generally large, images other than cash are likely to be shot, for example, as shown in fig. 9, the input cash image includes a top image 910 and a side image 920, and the cash bundle occupies only a partial area in both images, and also includes an external environment, and since the SSD model does not include a candidate frame, the external environment has a large influence on the accuracy, after the cash image is obtained, the cash area can be determined by recognizing the boundary of the cash, for example, in fig. 9, the boundary of the cash is recognized in the top image 910 to obtain a first cash area 911, the image recognition of the cash is performed in the side image 920 to obtain a second cash area 921, and the top image 910 and the side image 920 are cropped so that the first cash area 911 remains in the top image 910, and the second cash area 921 remains in the side image 920, the first cash area 911 and the second cash area 921 are used as the input of the SSD model, so that the interference of the complex environment can be effectively reduced, the calculation amount is reduced, and the accuracy of the target detection is improved.

It should be noted that the first cash area or the second cash area can be obtained by a simple image recognition technology, and those skilled in the art know how to recognize paper money with a fixed and known shape in an image, and no redundant description is given to the image recognition technology.

It should be noted that, in order to avoid cutting out the part of the banknote bundle by mistake, appropriate redundancy may be introduced when determining the first cash area and the second cash area, and the target frames of the first cash area and the second cash area are enlarged, and those skilled in the art know how to configure redundancy, so as to avoid cutting out the image of the banknote bundle, and it is not described herein to be repeated.

In addition, in an embodiment, referring to fig. 10, step S130 of the embodiment shown in fig. 1 further includes, but is not limited to, the following steps:

step S1010, acquiring a preset digital reference area;

step S1020, performing number detection in the first cash area according to the number reference area, and if a number is detected, generating the first target frame.

It should be noted that, unlike the common artificial intelligence image recognition, the image recognition target for the cash business is cash, and the digital area of the cash is known, so a digital reference area can be preset, and the efficiency of the digital detection can be improved, for example, in the top image 910 shown in fig. 9, the digital area on the back side of the 100-yuan cash of the rmb is located at two top corners on the upper side and in the left area on the lower side, and the corresponding area can be configured as the digital reference area, so that the digital detection is directly performed in the digital reference area, and the efficiency of the digital detection is effectively improved.

It should be noted that, because there are many currency types of cash, the determination may also be performed according to the information template described in the embodiment shown in fig. 1, for example, handling a dollar deposit service, inputting the currency type as dollar deposit in the information template, and the reference area may improve the detection efficiency.

In addition, in an embodiment, referring to fig. 11, after step S110 of the embodiment shown in fig. 1 is executed, the following steps are included, but not limited to:

step S1110, acquiring a first size threshold and a second size threshold that are set in advance;

step S1120, determining that the top image is valid when the size of the top image is greater than or equal to the first size threshold;

in step S1130, when the size of the side image is greater than or equal to the second size threshold, it is determined that the side image is valid.

It should be noted that, for the SSD model shown in fig. 1, since the convolution layers are all used, although it is allowed to input pictures with different sizes, the number of final anchor boxes is different due to different picture sizes, and the detection accuracy is affected to a certain extent, therefore, a first size threshold and a second size threshold may be set in advance, and after the cash image is acquired, the size of the image is determined, so as to avoid the detection accuracy from being affected by too small of the image, for example, based on the SSD model shown in fig. 1, the first size threshold may be set to 300 × 300, that is, the top image needs to satisfy more than or equal to 300 × 300, and based on the embodiment shown in fig. 8, the cash region to be cut needs to be cut, that is, the first cash region and the second cash region are both more than or equal to 300 × 300. It is understood that the values of the first size threshold and the second size threshold may be the same, or may be different according to the requirement, and are not limited herein.

In addition, in an embodiment, referring to fig. 12, the present invention further provides an SSD-based service auditing apparatus 1200, which includes but is not limited to the following apparatuses:

the system comprises an acquisition unit 1210, a processing unit and a processing unit, wherein the acquisition unit 1210 is used for acquiring information to be audited, the information to be audited comprises a total amount to be audited, a service type to be audited and a cash image, and the cash image comprises a top image and a side image;

an image input unit 1220, configured to input the top image and the side image to a pre-trained SSD model;

a first identifying unit 1230, configured to perform digital detection on the top image through the SSD model, generate a first target frame, and identify denomination information from the first target frame;

the second identification unit 1240 is configured to perform waist bar detection on the side image through the SSD model, generate a second target frame for each waist bar, determine a target total amount according to the number of the second target frames and the denomination information, and determine a target service type by performing character identification on the second target frame;

the auditing unit 1250 is configured to, when the service type to be audited matches the target service type, obtain an auditing result by comparing the total amount to be audited with the target total amount.

In addition, referring to fig. 13, an embodiment of the present invention further provides an electronic device 1300, where the electronic device 1300 includes: memory 1310, processor 1320, and computer programs stored on memory 1310 and executable on processor 1320.

The processor 1320 and memory 1310 may be connected by a bus or other means.

Non-transitory software programs and instructions required to implement the SSD-based service auditing method of the above-described embodiment are stored in the memory 1310, and when executed by the processor 1320, perform the SSD-based service auditing method applied to the device of the above-described embodiment, for example, perform the above-described method steps S110 to S150 in fig. 1, method steps S410 to S420 in fig. 4, method steps S610 to S620 in fig. 6, method steps S710 to S720 in fig. 7, method steps S710 to S820 in fig. 8, method steps S1010 to S1020 in fig. 10, and method steps S1110 to S1130 in fig. 11.

The above-described embodiments of the apparatus are merely illustrative, wherein the units illustrated as separate components may or may not be physically separate, i.e. may be located in one place, or may also be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.

Furthermore, an embodiment of the present invention further provides a computer-readable storage medium, where the computer-readable storage medium stores computer-executable instructions, which are executed by a processor or a controller, for example, by a processor in the above-mentioned embodiment of the electronic device, and enable the processor to execute the SSD-based service auditing method applied to the electronic device in the above-mentioned embodiment, for example, execute the above-mentioned method steps S110 to S150 in fig. 1, method steps S410 to S420 in fig. 4, method steps S610 to S620 in fig. 6, method steps S710 to S720 in fig. 7, method steps S710 to S820 in fig. 8, method steps S1010 to S1020 in fig. 10, and method steps S1110 to S1130 in fig. 11. It will be understood by those of ordinary skill in the art that all or some of the steps, means, and methods disclosed above may be implemented as software, firmware, hardware, or suitable combinations thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable storage media, which may include computer storage media (or non-transitory storage media) and communication storage media (or transitory storage media). The term computer storage media includes volatile and nonvolatile, removable and non-removable storage media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as is well known to those of ordinary skill in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, Digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other storage medium which can be used to store the desired information and which can be accessed by a computer. In addition, communication storage media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery storage media as is well known to those of ordinary skill in the art.

The application is operational with numerous general purpose or special purpose computing system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet-type devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like. The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

While the preferred embodiments of the present invention have been described in detail, it will be understood by those skilled in the art that the foregoing and various other changes, omissions and deviations in the form and detail thereof may be made without departing from the scope of this invention.

Claims

1. A SSD-based service auditing method is characterized by comprising the following steps:

inputting the top image and the side image into a pre-trained SSD model;

2. The SSD-based transaction auditing method of claim 1, wherein the performing, by the SSD model, waist bar detection on the side images and generating a second target box for each waist bar comprises:

3. The SSD-based traffic auditing method of claim 2, wherein prior to the waist bar detection in each of the third destination boxes, the method further comprises:

4. The SSD-based service auditing method of claim 1, wherein obtaining an auditing result by comparing the total to be audited with the target total comprises:

5. The SSD-based transaction auditing method of claim 1, wherein prior to the inputting the top and side images to a pre-trained SSD model, the method further comprises:

6. The SSD-based business auditing method of claim 5, wherein the digitally detecting the top image through the SSD model to generate a first target box comprises:

acquiring a preset digital reference area;

7. The SSD-based service auditing method according to claim 1 or 5, characterized in that after said obtaining information to be audited, the method further comprises:

acquiring a preset first size threshold and a preset second size threshold;

determining that the top image is valid when the size of the top image is greater than or equal to the first size threshold;

and when the size of the side image is larger than or equal to the second size threshold value, determining that the side image is effective.

8. A SSD-based service auditing apparatus, comprising:

9. An electronic device, comprising: memory, processor and computer program stored on the memory and executable on the processor, characterized in that the processor implements the SSD-based traffic auditing method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing computer-executable instructions for performing the SSD-based traffic auditing method of any one of claims 1-7.