KR102358607B1 - Artificial intelligence appraisal system, artificial intelligence appraisal method and storage medium - Google Patents

Abstract

The present disclosure includes a customer terminal and an emotion server that evaluates an item based on image data transmitted online from the customer terminal through a network, wherein the emotion server is trained to discriminate the model of the item by deep learning. It provides an artificial intelligence emotion system comprising an artificial intelligence-based emotion engine, wherein the emotion engine performs image signal processing on one or more point portions of the article to determine whether the article corresponds to the genuine article of the model.

Description

Artificial intelligence appraisal system, artificial intelligence appraisal method and recording medium

The present disclosure relates to an artificial intelligence emotion system, an artificial intelligence emotion method, and a recording medium.

Luxury items such as precious metals, watches, clothing, bags, shoes, wallets, and expensive items such as antiques, ceramics, calligraphy and crafts are often traded at a high price even if they are second-hand items due to their rarity. . These expensive items can be exchanged for a considerable amount of cash by depositing them as collateral at a pawnshop or the like. For this reason, counterfeits or counterfeit products are often distributed in the market for these high-priced products.

Conventionally, for the appraisal of corporeal movables including these expensive items, a professional appraiser who has been in the appraisal business for a long time resides in a pawnshop, etc. to assess the authenticity or price. Recently, even without directly visiting the pawnshop, when a photo of an item taken by the owner from a remote location is transmitted through a wired or wireless network using a terminal such as a computer or smartphone, the authenticity of the item is determined based on the transmitted photo. A remote appraisal service that provides information on market prices and the like is being studied.

However, according to the present inventor's research, the existing methods mostly depend on the personal experience of the appraiser, so there is a problem that the appraisal is not consistently performed according to a scientific and standardized way. In addition, there is a problem in that the accuracy of the emotion is greatly reduced in the existing appraisal method based on a photograph of an item. Therefore, it is required to develop a remote appraisal system that can consistently perform an appraisal in a scientific and standardized way, perform a more accurate appraisal, and present a more accurate and objective appraiser based on the appraisal result. .

The present disclosure provides an artificial intelligence emotion system, an artificial intelligence emotion method, and a recording medium that can significantly improve the problems of the prior art, including the aforementioned problems.

An embodiment of the present disclosure includes a customer terminal, and an emotion server that performs an appraisal of the article based on image data of the article transmitted from the customer terminal through a network, wherein the emotion server is an artificial intelligence learned by deep learning. an intelligence-based emotion engine, wherein the emotion engine learns data about a genuine model and a fake model for a plurality of point parts of the article, and the emotion engine, in the image data, includes the plurality of point parts By determining whether the genuine model and the fake model are included, and displaying the result of determining the degree of proximity to the genuine model and the fake model in the form of an index, It could be an artificial intelligence emotion system that performs a genuine emotion assessment that assesses the degree of proximity to a quantum.

An embodiment of the present disclosure provides an artificial intelligence appraisal method executed by an artificial intelligence appraisal system including a customer terminal and an appraisal server that performs appraisal of an article based on image data of the article transmitted from the customer terminal through a network In the following, the appraisal server performs model appraisal for determining the model of the article based on the image data, and image signal processing for a plurality of point parts of the article to determine whether the article corresponds to the genuine article of the model. Authenticity appraisal to determine whether or not, and state appraisal for determining a state index for indicating the preservation state of the article by determining one or more state factors of replacement, deformation, contamination and discoloration for the plurality of point parts of the article; , executes an appraised value determination to determine the appraised value of the article based on the state index, the emotion server includes an artificial intelligence-based emotion engine learned by deep learning, and the emotion engine includes the plurality of learns data related to a genuine model and a fake model for a point part, and the emotion engine determines whether the genuine model and the fake model are included with respect to the plurality of point parts in the image data, and the An artificial intelligence appraisal method for executing the authenticity appraisal of appreciating the degree of proximity of the article to the genuine article, the fake article, or both by displaying the result of determining the degree of proximity to the genuine model and the fake model in the form of an index can be

An embodiment of the present disclosure may be a computer-readable recording medium in which a computer program for executing the artificial intelligence emotion method is recorded.

According to the present disclosure, it is possible to provide an artificial intelligence emotion system, an artificial intelligence emotion method, and a recording medium capable of greatly improving the problems of the prior art, including the aforementioned problems.

In addition, various embodiments according to the present disclosure and their object, configuration, action, effect, advantage, etc. will be described in detail in the following description of the invention with reference to the accompanying drawings.

1 is a block diagram showing a schematic configuration of an artificial intelligence emotion system 1 according to an embodiment of the present disclosure.
2A to 2G are diagrams schematically illustrating an example of an operation screen of an emotion application executed in the customer terminal 30 .
3 is a flowchart illustrating an emotion procedure executed in the emotion server 10 .
4A and 4B are diagrams illustrating an example of a procedure for displaying model information of an article through the customer terminal 30 .
5A to 5F are diagrams for explaining an example of authenticity evaluation performed by an emotion engine mounted on the emotion server 10 .
6 is a diagram illustrating an example of a big data-based price database (DB) for calculating an appraised value.

Hereinafter, an embodiment of the present disclosure will be described in detail with reference to the accompanying drawings. In the following description, "one embodiment" means an exemplary embodiment presented to help the understanding of the present disclosure, and the present disclosure is not limited to the exemplary embodiment described below. The accompanying drawings may be illustrated by enlarging, exaggerating, briefly showing or omitting some or all of the configuration for convenience of description.

Throughout the description of the invention, when it is described that a part is "connected" or "connected" with another part, it is not only directly connected or connected, but also indirectly connected or connected through another medium or component in the middle. It may include cases where When it is described that a part "includes" a component, unless otherwise specifically indicated, other components other than the described components may be included, rather than excluded, and one or more other features or components , does not exclude the possibility of the presence or addition of parts, steps, acts, or combinations thereof.

First, the overall configuration of the artificial intelligence emotion system 1 according to an embodiment of the present disclosure will be described.

1 is a block diagram showing a schematic configuration of an artificial intelligence emotion system 1 according to an embodiment of the present disclosure.

Referring to FIG. 1 , the artificial intelligence emotion system 1 may include an emotion server 10 , a manager terminal 20 , and a customer terminal 30 . The artificial intelligence emotion system 1 may further include a data crawler 40 . The emotion server 10 , the manager terminal 20 , the customer terminal 30 and the data crawler 40 are connected to the network N, and may communicate with each other through the network N.

The emotion server 10 may be a device for managing and controlling the overall operation of the artificial intelligence emotion system. The emotion server 10 manages and controls the transmission/reception, processing, input/output, etc. of various data (including electrical signals and/or information) in accordance with the control command and/or its own operation program from the manager terminal 20 as a whole. can The emotion server 10 may include, for example, a data processing unit such as a CPU, a RAM, a data memory unit such as a ROM, a communication unit for transmitting and receiving data, and various input/output terminals for input/output of data. The emotion server 10 may include an emotion engine for performing an emotion, software such as a computer program required for system operation, and a database for storing various data used for emotion. The emotion server 10 may communicate with the manager terminal 20 and/or the data crawler 40 as necessary to perform a task required for emotion. The appraisal server 10 may receive data on the object to be appraised from the customer terminal 30 , perform appraisal based on this, and then transmit data on the appraisal result to the client terminal 30 . . To this end, the emotion server 10 may include a service server for providing a remote service to the customer terminal 30 .

The emotion engine mounted on the emotion server 10 may include an artificial intelligence-based image processing engine that determines the model, authenticity, state, etc. of an item through an image captured by the user's customer terminal 30 . The appraisal engine may further include a big data-based pricing engine that determines an appraised value according to the determined product model information. Details will be described later.

The manager terminal 20 may be a device through which the manager accesses the artificial intelligence emotion system 1 through the network N. The manager may communicate with the emotion server 10 or the data crawler 40 by accessing the network N through the manager terminal 20 . The manager terminal 20 may also communicate with the customer terminal 30 through the network N, if necessary. The manager terminal 20 may include various input/output means capable of communication, such as transmitting a control command or the like to the emotion server 10 or the data crawler 40 or receiving the processing result of data from these devices. For example, the manager terminal 20 may include a desktop computer, as well as a mobile input/output means such as a notebook computer, a tablet computer, and a smart phone. The manager terminal 20 may store and execute software or a database for managers necessary for the operation of the artificial intelligence emotion system 1 .

The customer terminal 30 may be a device for a user who wants to evaluate an item to access the artificial intelligence emotion system 1 through the network N. The customer terminal 30 may include various input/output means capable of communicating with the emotion server 10 . The customer terminal 30 may store and execute an emotion application for transmission and reception of data required for emotion. Such an emotion application may be downloaded by directly accessing the emotion server 10 or by accessing an Internet website, a web store, a mobile store, etc. that provide an application for download. The emotion application may be provided by being stored in a recording medium such as a CD, USB, SD.

The data crawler 40 is a device that collects, stores, processes and/or updates (hereinafter referred to as "crawling") data necessary for the operation of the artificial intelligence emotional system 1 through the network N. can be The data crawler 40 may be a server device such as the emotion server 10 . The data crawler 40 may build a database necessary for the emotion engine or the appraised value calculation engine of the emotion server 10 , and store and execute software required for crawling data. The data crawler 40 may crawl or input/output data from time to time and/or periodically according to a control command and/or its own operation program from the emotion server 10 or the manager terminal 20 . The data crawler 40 is connected only to the emotion server 10 if necessary, and may be configured to be operated only by the emotion server 10 .

The network (N) may include various types of wired or wireless communication networks that enable mutual communication between the emotion server 10 , the manager terminal 20 , the customer terminal 30 , and the data crawler 40 . The network N is an Internet network that provides server-client web services, and may include, for example, wired or wireless Internet networks such as LAN, WAN, 4G, 5G, and WIFI. The network (N) may be configured so that at least a portion of the emotion server 10, the manager terminal 20, and the data crawler 40 is connected only to a closed internal communication network such as a local intranet, if necessary.

Hereinafter, an artificial intelligence emotion method according to an embodiment of the present disclosure using the artificial intelligence emotion system 1 including the above-described configuration will be described in detail with reference to the drawings.

First, the operation of the emotion application executed in the customer terminal 30 will be described.

2A to 2G are diagrams schematically illustrating an example of an operation screen of an emotion application executed in the customer terminal 30 .

2a to 2g, the case of using a smartphone as an example of the customer terminal 30 is exemplified, but the customer terminal 30 is not limited thereto, and data including image information such as photos and videos of goods It goes without saying that various input/output means for transmitting through the network N and receiving and confirming the result of the evaluation may be employed.

Referring to FIG. 2A , first, a user may access a mobile store that provides an application for download, etc., and may download and execute the emotion application in the customer terminal 30 .

When the emotion application is executed, the home screen 32 may be displayed. The home screen 32 may include a service icon display unit 32a. The service icon display unit 32a may display, for example, a service icon from which a user can select a desired service type on the screen. The service icon may include one or more icons for displaying services such as, for example, a transaction, a wallet, an emotion, a notification, and a setting.

The transaction icon is an online marketplace where users can place information on the list, status, price, etc. of items and trade with them, or an online shopping mall where the administrator places information on the list, condition, price, etc. of items and sells and/or rents them. Alternatively, it may be an icon for moving to a screen providing a service, such as a rental shop. The wallet icon may be an icon for moving to a screen providing a service through which users can input, store, and pay information on mobile payment methods such as cards, cash, coupons, and the like. The emotion icon may be an icon for moving to a screen that provides an artificial intelligence emotion service according to the present disclosure for an item for which the user wants to feel emotion. The notification icon may be an icon for moving to a screen displaying the content of a received message with the user as the recipient. The setting icon may be an icon for moving to a screen for specifying or changing various settings or operations of the application.

The home screen may further include a membership sign-up icon in addition to these service icons. The member signup icon may be an icon for moving to a screen where a user can join as a member by entering information such as a user's name, age, gender, address, and contact information and setting an ID and password. When the membership registration process is completed by pressing or touching the membership registration icon, an account for members may be provided. When an account for members is provided, the emotion server 10 may preferentially receive and/or process data input through the account for members. Data transmitted and received through the member's account may be continuously managed while being stored and updated in the emotion server 10 .

When the service icon is pressed or touched, a guide phrase describing the service provided by each icon may be displayed on the screen. For example, when the emotion icon is touched, a guide phrase such as “Try to change the item to the cash now” may be displayed on the upper portion of the service icon display unit 32a.

Referring to FIG. 2A, when the user presses or touches, for example, a transaction icon or an emotion icon, the home screen 32 displays a category icon indicating a category of goods handled by the transaction service or the emotion service on the category icon display unit ( 32b) can be indicated.

Category icons may include icons representing the names or images of items such as laptops, tablets, mobiles, bags, watches, wallets, and shoes. can For goods for which services are not yet provided, a guide phrase such as "coming soon" may be displayed along with an icon displaying the name or image of the product. Since the item category may be automatically determined by the emotion engine, pressing or touching the emotion icon may move directly to the next screen without displaying the category icon screen.

When the emotion icon is selected or any one of the category icons is selected, the home screen 32 may move to the emotion article photographing screen 34 as shown in FIGS. 2B to 2F .

Referring to FIG. 2B , the appraisal article photographing screen 34 may include a screen window of the camera for acquiring image signal data such as a photo or video of the article. Image signal data of the object to be assessed may be acquired by photographing the object to be assessed through the screen window of the camera.

The test item photographing screen 34 may include a photographing sequence guide 34a, an image photographing unit 34b, and a help display unit 34c.

The photographing sequence guide unit 34a may be a screen for guiding the user about the photographing order of the article, the photographing part, the photographing method, and the like. The shooting sequence guide unit 34a is displayed on the screen by the user ①, ②, ③, ?? By sequentially touching numbers such as, etc. and taking a picture according to the guidance displayed on the screen, it is possible to guide so that data necessary for appraisal of an article is acquired.

The image capture unit 34b is a screen window of the camera for photographing the object to be assessed, and may include a target window T for positioning the object to be photographed in the correct position. The target window T may include four bracket marks located at each corner of the substantially rectangular screen window, and an approximately cross-shaped target mark located at the center of the screen window to locate the center of the target. The target window T may move its position or change its size by a touch of a finger or the like. The target window (T) may be configured to recognize the photographing part of the object to be assessed by itself, and to automatically adjust the position or size of the target window (T) according to a predetermined format.

The help display unit 34c may describe a photographing method in detail or present a standard photographed image to help the user accurately photographing. For example, when the help display unit 34c is touched, a guide phrase such as "Please remove obstacles around and take a picture of the front of the product based on a clean background" is displayed, or the standard of each shooting step according to the category of the item to be assessed. Sample images can be presented.

For example, when the user selects a watch as the category of the appraisal item, if the user touches the "①" icon on the shooting order guide 34a, the user can move to the front shooting screen of FIG. 2C . The front shooting screen, for example, displays the target window (T) for helping the front shooting of the watch on the image capturing unit (34b), and displays “1. A guide phrase such as "Please take a picture" and a shooting button (B) can be displayed. At this time, if the help display unit 34c is touched, an exemplary sample image obtained by correctly photographing the front of the watch may be displayed on the screen.

After the front of the watch is photographed, when the user touches the icon “②” on the photographing sequence guide 34a, the user can move to the rear photographing screen of FIG. 2D . The backside shooting screen, together with the target window (T) and the shooting button (B), can display guidance phrases such as “2. Take a picture of the inside of the watch” and a sample video of the back side of the watch. .

After the photographing of the back side of the watch is completed, when the user touches the icon “③” on the photographing sequence guide 34a, it can move to the band photographing screen of FIG. 2E . The band shooting screen, along with the target window (T) and the shooting button (B), can display guidance phrases such as “3. .

After shooting the band of the watch, if the user touches the icon "④" on the shooting order guide 34a, it can move to the clip shooting screen of FIG. 2F. The clip shooting screen, along with the target window (T) and the shooting button (B), displays guidance such as “4. Take a clip (lock part) of the watch”, and a sample video of the watch clip being shot correctly. can do.

In the appraisal application, the user selects whether to store data such as photos and videos acquired by shooting in the customer terminal 30 after the user completes the shooting of each shooting part according to the guidance of the shooting order guide unit 34a More screens can be displayed. The appraisal application may further display a screen for additionally inputting other information about the appraisal item. The emotion application may further display a screen for selecting whether to transmit data for emotion after all procedures for acquiring data required for emotion are completed. In addition, it is of course possible to further configure various screens according to the provision of a remote appraisal service or the convenience of the user.

When the user transmits the data acquired through the execution of the above-described emotion application from the customer terminal 30 to the emotion server 10 through the network N, the emotion server 10 executes an appraisal procedure based on the transmitted data and transmit the result to the customer terminal 30 . More details will be described later.

The evaluation result by the evaluation server 10 may be transmitted to the customer terminal 30 . When the emotion result is transmitted to the customer terminal 30, it may be displayed on the notification icon of the appraisal application that the appraisal result has been received. If the user touches the notification icon after executing the emotion application, the user may move to the emotion result screen 35 as shown in FIG. 2G .

Referring to FIG. 2G , the evaluation result screen 35 may display the evaluation result by the evaluation server 10 together with data such as a photo for specifying the object to be evaluated. The analysis result, for example, a guide phrase such as "assessment is complete", the authenticity of the object to be assessed, the exact model name or production year of the object to be assessed, the state of the object to be assessed, the appraised value, etc. may be displayed on the screen. The appraisal result may provide an appraisal written by the appraisal server 10 . For example, when an appraisal icon indicating an appraisal is displayed on the appraisal result screen 35 and the user touches the appraisal icon, it can be configured so that the appraisal created by the appraisal server 10 can be read. Such an appraisal may be prepared by an artificial intelligence emotion engine, which will be described later. The appraisal may simply present the appraisal result such as whether the item is genuine or the appraised value, or may provide a method of presenting the reason for the appraisal result in detail, such as the condition of the item. The appraisal may be classified as a service item provided only to users who are registered as members.

On the other hand, if the data or image about the item input by the user is not in the item list of the item handled by the artificial intelligence appraisal system 1, the image data of the item is not transmitted to the appraisal server 10, but the appraisal application itself You can judge a result such as "Not an appraisal item" and display it on the screen. In addition, as a result of the evaluation by the appraisal server 10 , the result of “appraisal impossible” may be displayed in the case of an item whose authenticity is difficult to be appraised.

Next, an emotion procedure executed in the emotion server 10 will be described in detail.

3 is a flowchart illustrating an emotion procedure executed in the emotion server 10 .

Referring to FIG. 3 , the appraisal server 10 may execute an appraisal procedure based on the image data of the item captured and transmitted by the user by executing the appraisal application mounted on the customer terminal 30 . The appraisal procedure 10 may proceed in the order of model appraisal, authenticity appraisal, state appraisal, appraiser value determination, and result transmission.

First, when data obtained by photographing an image of an article is transmitted, the appraisal server 10 may perform model appraisal for determining an accurate model of the article. The model evaluation may proceed in the order of a product group classification for classifying which product group (category) the appraisal article belongs to, a brand classification for classifying the brand of the appraisal article, and a model classification for classifying the model of the appraisal article. To this end, the appraisal server 10 may include an appraisal engine for executing article appraisal by processing image data of the article.

The emotion engine is an artificial intelligence (AI)-based image processing engine that learns to identify objects in images by machine learning techniques, particularly deep learning techniques, and can recognize objects by itself and distinguish them from other objects based on this. . If such an emotion engine learns a large amount of image data according to product group, brand, model, condition, etc., it identifies the item in the image transmitted by the user by the already learned and evolved artificial intelligence, and the product group, brand, and model of the product are identified. Of course, it is possible to accurately determine the condition of the item as well. The learning data for deep learning of the emotion engine may collect image data according to item lists such as product groups, brands, and models through the data crawler 40, and continuous learning of the emotion engine may be performed based on this. To this end, the emotion server 10 may execute deep learning learning of the emotion engine from time to time and/or regularly through communication with the data crawler 40 .

The deep learning-based product model evaluation determines whether the product is an item handled by this service through the image captured by the user, and specifically, it can recognize the model of a specific product group and transmit the result. A deep learning engine recognizes an object learned in an image, and can utilize open sources such as Google's tensorflow and yolo.

The primary product group classification as to whether or not it corresponds to an item handled by the product appraisal service may be determined by mounting a recognition engine in the appraisal application of the customer terminal 30 . Through this, it is possible to reduce the load on the emotion server 10 by judging within the appraisal application of the customer terminal 30 without proceeding to the appraisal server 10 for items other than the handling product group.

Thereafter, in order to classify the brand and the actual model, the captured image may be transmitted and determined by the artificial intelligence emotion engine of the emotion server 10 . For example, as shown in FIG. 4A , the emotion application (mobile appraisal app) stored in the customer terminal 30 performs product group classification and handling item classification, and then transmits a brand and model confirmation request to the emotion server 10 . And, after receiving a reply from the emotion server 10 to the finally confirmed brand and/or model information and reporting it to the customer terminal 30, and/or transmitting it to the service server 12, the subsequent process can be performed have. For example, as shown in FIG. 4B , model information such as a model name, manufacturing year, serial number, and photo for specifying a model of an article may be displayed on the customer terminal 30 .

Next, the appraisal server 10 may execute a genuine appraisal for determining whether the item in the image transmitted by the user is a genuine product or a fake product of the corresponding model.

5A to 5F are diagrams for explaining an example of authenticity evaluation performed by an emotion engine mounted in the emotion server 10 .

The emotion engine extracts a specific location or part (hereinafter referred to as a point part) that can determine the characteristics of the article from the image data transmitted by the user, and processes the image signal of the image data corresponding to the point part of the article. It can be determined whether it corresponds to the genuine product according to the model information. The emotion engine may automatically extract and/or process the point portion of the model from the image data transmitted by the user and process it. Also, for example, as shown in FIG. 5A , a button, buckle, logo, etc. of a bag corresponding to a specific model may be exemplified in advance to the photographing user, so that the user may transmit an image photographed for this point portion.

For example, if the logo of the corresponding brand as shown in FIG. 5A is genuine, the ratio for each stroke and the horizontal to vertical ratio of the logo size may be constant. The authenticity of these logos can be determined by the following algorithm.

First, the Y value may be used in the YCbCr conversion equation (ITU-R BT.601) of Equation (1) below to obtain a gray image for the input image of FIG. 5A .

[Formula 1]

...(1)

...(One)

The image for Y' in Equation (1) above is the same as the gray image of FIG. 5B.

Next, a bilateral filter for removing noise appearing in the image and sharpening the logo may be applied. Equations (2) and (3) below represent the bilateral filter.

[Equation 2]

...(2)

...(2)

[Equation 3]

...(3)

...(3)

는 결과 영상을 나타내며

는 입력 영상,

는 현재 pixel의 좌표를 나타낸다.

는 filtering을 위한 window 크기를 나타내고,

는 밝기 차이와 공간적 근거리에 대한 가중치를 나타낸다.In Equation (2),

represents the resulting image.

is the input image,

represents the coordinates of the current pixel.

represents the window size for filtering,

denotes the weight for the brightness difference and spatial proximity.

Through this process, noise can be removed while maintaining an edge component having a large brightness difference, and FIG. 5C shows the result.

Next, edge detection may be performed. Here, a Gaussian difference method can be used.

[Equation 4]

...(4)

...(4)

는 다음과 같이 계산할 수 있다.By varying the sizes of two Gaussian filters, an edge component can be derived using the difference. Thereafter, contouring is performed on the detected edge component to find a bounding box of the logo portion as shown in FIG. 5D . For contouring, for example, findContour of OpenCV, which is an image processing library, may be used. When the coordinates of the bounding box found as above are p1 to p4, respectively, the ratio of the length to the width can be obtained to calculate the ratio information for the size of the logo. length ratio

can be calculated as

[Equation 5]

...(5)

...(5)

As shown in Equation (5), it is possible to determine the authenticity of the bounding box by calculating the ratio between the horizontal and vertical lengths and comparing the reference value for the genuine product with the ratio value calculated from the input image.

In addition to the above-mentioned parts, the authenticity assessment may be performed by extracting various point parts that can be identified by images, such as, for example, the shape or material of a specific part of the article, luster or color, printed or engraved letters or numbers, stitching state, and the like. These point parts can learn various image data through deep learning and the emotion engine itself can automatically extract a specific part of a specific article as a point part.

On the other hand, even if the product is genuine, there is a risk that a judgment error may occur if the point part extracted for authenticity is deformed by damage, discoloration, etc. and affects the evaluation. For this reason, the emotion engine can learn in advance through deep learning by collecting a large amount of data not only for the genuine model but also for the fake model.

For example, since the manufacture of counterfeit products is often performed by a specific counterfeit manufacturer, models determined to be counterfeit products may include a specific region in which the characteristics of the counterfeit product manufacturer appear. By learning the image data including the shape or characteristics of a specific part of the fake product through deep learning, the emotion engine itself can extract the point part for determining whether it is a fake model and determine whether it is a fake.

Also, in some cases, fake parts are included in the genuine model due to replacement or repair of some parts. Accordingly, the appraisal engine executes both authenticity and counterfeit discrimination for the appraised article, and the result is, for example, a genuine index indicating a degree close to the genuine model and/or a degree close to the fake model. It can be displayed in the form of a fake index to be displayed. The genuine index and/or the fake index may be displayed in a form such as numbers, letters, percentages, etc., that can be easily understood by a user at a time. For example, the authenticity index may give a score from 0 to 100 for the evaluation result, and may be displayed so that the closer to 100, the more genuine, and the closer to 0, the more fake. The same is true for the fake index. The genuine or fake index can be displayed in the range of 0 to 10, and can be displayed in various forms, such as 1 for genuine and 0 for fake.

In addition, it goes without saying that the identification algorithm can be added and updated based on image signal processing and deep learning technology for various authenticity identification methods for each item.

After the authenticity appraisal is completed, the appraisal server 10 may transmit a result of the authenticity appraisal to the customer terminal 30 in advance. The appraisal server 10 may immediately execute the state appraisal of the appraisal article without reporting the result of the authentic appraisal.

The state appraisal may be an emotion for determining the level of the preservation state of the appraisal article. The preservation state of the article may be determined by a condition factor. State factors include, for example, a replacement state indicating whether a specific part or part of an article is to be replaced, a deformed state indicating whether deformation is due to damage or breakage, a contamination state indicating whether contamination due to adhesion of foreign substances, etc., discoloration or discoloration due to use It may include a discoloration state indicating whether or not. Each article may include parts that are easily replaced, deformed, contaminated or discolored depending on the type of article or model thereof. The preservation state can be determined by extracting these parts as point parts for state evaluation.

The state emotion may be executed by the emotion engine described above. As in the authenticity appraisal, the appraisal engine learns the preservation state of the item and a large amount of image data according to the point region that can identify it by deep learning, and based on this, the preservation state of the item is determined from the image data transmitted by the user. can be discerned. The learning data regarding the preservation state of these items may be crawled frequently and/or regularly through the data crawler 40 , and continuous learning and evolution of the emotion engine may be made based thereon.

The emotion engine may extract and/or process a point portion capable of determining the state of the article from the image data transmitted by the user to determine the state of the article through image signal processing. For each point region for state determination, for example, based on the same algorithm as the image processing algorithm for authenticity determination described above, the emotion engine determines condition factors such as replacement, deformation, contamination, and/or discoloration. can be judged.

For example, as in the brand logo of FIG. 5A , a specific part that frequently deforms may be extracted as a point part for determining whether or not there is a deformation. The appraisal engine may pre-learn a large amount of data related to the deformation of the point portion and determine whether the object to be appraised is deformed based on this. Regarding replacement, contamination, discoloration, etc., the corresponding point parts can be learned in advance by the emotion engine to determine the presence or absence. Of course, other various image processing algorithms may be employed for state determination.

The appraisal engine may display the result in the form of a condition index indicating the preservation state of the article after all state determination on the appraisal article is performed. These status indexes can be displayed in a form such as numbers, letters, percentages, etc., that can be easily understood by a user at a time. For example, if there is replacement, deformation, contamination or discoloration in the point portion, 1 may be displayed, and 0 if not.

In addition, the emotion engine learns a large amount of data in advance by classifying the replacement, deformation, contamination or discoloration of the point part for status determination into "upper", "middle", and "lower", respectively, and based on this, replace, transform, In the case of contamination or discoloration status of "high", "medium", and "low", it can be displayed in the form of a score of "1", "2", and "3", respectively. The emotion engine may learn by subdividing the state of an item into three or more stages. It goes without saying that the score of the status index can be displayed in various ways, such as in the range of 1 to 10 and in the range of 1 to 100.

In the above, the two-dimensional discrimination technique based on photos has been mainly described, but the emotion server 10 reconstructs image data such as photos and videos taken by the user to perform virtual reality or augmented reality and The same three-dimensional model can be constructed. Accordingly, it is possible to construct a standardized three-dimensional model, and to perform extraction of a point portion, authenticity assessment, state assessment, and the like more accurately and easily based on this. Also, the emotion result may be three-dimensionally displayed in the form of augmented reality or the like on the three-dimensional model. Accordingly, it is possible to easily recognize the state of the article, the emotional result, and the like in a more realistic and intuitive manner.

Next, the emotion server 10 may execute an appraisal value determination after completing the state appraisal. The appraised value determination may be performed by a big data-based pricing engine mounted on the appraisal server 10 . The appraised value may be determined by the following appraised value calculation model.

6 is a diagram illustrating an example of a big data-based price database (DB) for calculating an appraised value.

Referring to FIG. 6 , the price DB may include a product definition table and a price table. The product definition table may record and store IDs such as product names and product numbers according to product groups, brands, models, etc. of the object to be assessed. The price table may record and store a price according to an ID of each product as a product price.

The product price recorded in the price table may be collected from time to time and/or periodically from a price reference site designated in advance by a price DB operation agent. The commodity prices recorded in these price reference sites and price tables may be automatically designated or updated by data crawling by the data crawler 40 .

If there is no corresponding product in the designated or collected price reference site, the appraisal server 10 may notify the manager terminal 20 of a signal requesting information registration for price determination by designating the ID of the corresponding product. When such a request is notified, the manager may additionally register price information in the price table directly through the manager terminal 20 . If the item is deleted from the designated or collected price reference site, you can designate the item from another site and add the designated site and collected items to the price DB operation agent.

The product price recorded in the price table may be the highest price of the corresponding ID product collected from the price reference site. That is, the price of the item with the best preservation state may be specified as the price of the corresponding ID item. Alternatively, an average price obtained by averaging the prices of the corresponding ID article collected from a plurality of price reference sites may be recorded as the price of the corresponding ID article. That is, an article in an average preservation state of the corresponding model can be specified as the price of the corresponding ID article.

The pricing engine may determine the assessed value based on the commodity price and the above-described status index.

For example, a constant weighting coefficient may be assigned to each of the factors indicating the preservation state of the article, that is, the state of replacement, deformation, contamination, and discoloration. The weighting coefficient may be a coefficient representing a proportion in the price of the appraisal article. By multiplying this weighting factor by the above-described condition index, a price discount rate indicating the degree of depreciation due to deterioration of the preservation condition of the article can be calculated. Accordingly, the appraised value may be determined by the following subtraction method.

Assessed value = commodity price - (commodity price * weight factor * condition index)

For example, assuming that the commodity price of an article is 1, a weighting factor of 0.4 for replacement, 0.3 for deformation, 0.2 for contamination, and 0.1 for discoloration may be assigned. When there is a deformation in the point portion of the article, for example, the state index becomes 1 and the weighting factor becomes 0.3, so the depreciation may be 0.3. Accordingly, the appraised value may be determined as 1-(1*0.3*1)=0.7. That is, when there is a replacement of the product, 30% of the product price may be depreciated and 70% of the product price may be determined as the assessed value.

In addition, when there are a plurality of point parts for determining each state factor such as replacement or deformation, for example, a value obtained by dividing a weighting factor of the corresponding state factor by the total number of point parts for determining the corresponding state factor is a sub weighting coefficient , and by summing all the products of the sub-weighting coefficient and the state index for each point portion of the corresponding state factor, the product of the weighting coefficient and the state index of the corresponding state factor can be obtained.

Alternatively, the appraisal engine records the lowest or average price of the corresponding ID item collected from the price reference site as the product price in the price table, and when there is no or good item corresponding to the above-mentioned depreciation, the value corresponding to the depreciation The appraised value can also be determined by the addition method.

In addition, the appraised value may be determined by reflecting the business cost or profit rate of the manager providing the appraisal service for the appraised value.

According to the emotion system and emotion method according to an embodiment of the present disclosure described above, an image of an item transmitted by a user online by the emotion server 10 equipped with an artificial intelligence-based emotion engine learned and evolved through deep learning It is possible to provide an article appraisal service that can accurately and quickly determine whether an article is a genuine product of the corresponding model in a more scientific and standard way by performing image signal processing on it.

In addition, the appraisal engine learns not only genuine data but also fake data through deep learning to perform both authenticity and counterfeit discrimination, so it is possible to provide a product appraisal service that can perform more accurate authenticity discrimination.

In addition, the appraisal engine may provide an article appraisal service capable of discriminating the price according to the preservation state of the article by performing state determination capable of determining the preservation state of the article and calculating the state index.

In addition, the price determination engine not only determines the product price by collecting a large amount of information about product price based on big data, but also calculates the depreciation of the product based on the condition index calculated by the appraisal engine, thereby discriminating according to the preservation status of the product. This possible appraised value can be calculated. Accordingly, the user can be provided with a specific appraised value determined by a more objective and consistent standard according to the preservation state of the item for which the appraisal has been requested.

The artificial intelligence appraisal system and the artificial intelligence appraisal method according to an embodiment of the present disclosure may be employed in various commerce and financial services such as online pawnshops, used transactions, and secured loans, as well as appraisal services using mobile.

In an embodiment of the present disclosure, the artificial intelligence emotion system and the artificial intelligence emotion method are produced as a control logic or program including instructions executable by a computer, such as a program module executed by a computer, and recorded in a computer-readable recording medium. It can also be implemented in a form. A computer-readable recording medium may be any available medium that can be accessed by a computer, and may include both volatile and nonvolatile media, and removable and non-removable media. In addition, the computer-readable recording medium may include a computer storage medium. Computer storage media may include both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules, or other data.

The present disclosure is not limited only to the above-described embodiments, and various modifications and changes are possible by those skilled in the art. In addition, each component or method described in each embodiment described above, each application example or a modification thereof, etc. may be implemented by being appropriately coupled to each other or separately separated as needed. Accordingly, the scope of the present disclosure should be broadly interpreted to include all modifications and other embodiments including the technical configuration and equivalents thereof described in the following claims.

10 ... appraisal server 20 ... administrator terminal
30 ... customer terminal 40 ... data crawler

Claims (10)

customer terminal;
an appraisal server for appraising the article based on the image data of the article transmitted from the customer terminal through the network;
A data crawler for performing data crawling for deep learning learning of the emotion server,
The emotion server includes an artificial intelligence-based emotion engine learned by deep learning,
The emotion engine learns data about a genuine model and a fake model for a plurality of point parts of the article,
The emotion engine, in the image data, determines whether the genuine model and the fake model are included with respect to the plurality of point parts, and also indexes the result of determining the degree of proximity to the genuine model and the fake model By displaying in the form of , authenticity assessment is performed to evaluate the degree of proximity of the article to the genuine article, a fake article, or both;
The appraisal server constructs a three-dimensional model of the article from the image data by virtual reality or augmented reality, extracts the plurality of point parts of the article from the three-dimensional model, and executes the authenticity appraisal,
The appraisal server includes a big data-based price database that registers the ID of a product according to the model of the product, and records the product price obtained from a price reference site that provides price information of the product corresponding to the ID,
The emotion engine is
Perform image signal processing for determining one or more state factors among replacement, deformation, contamination, and discoloration of the plurality of point portions of the article, and determine a state index for indicating the preservation state of the article based on this; ,
assigning to each of the one or more condition factors a weighting factor representing a weighting factor representing a weighting factor in the price of the article, and calculating the depreciation of the article by the product of the condition index and the weighting factor, based on the depreciation To determine the appraised value of the article according to the following formula (1),
Assessed Value = Merchandise Price - (Commodity Price*Weighting Factor*Status Index)...(1)
When there are a plurality of point regions for determining each state factor, a value obtained by dividing the weighting coefficient of the state factor by the total number of point regions for determining the state factor is used as a sub weighting coefficient, An artificial intelligence emotion system that calculates the product of the weight coefficient and the state index of the corresponding state factor by summing the product of the weighting coefficient and the state index. delete delete delete delete delete delete Artificial comprising: a customer terminal; an emotion server that evaluates the article based on image data of the article transmitted from the customer terminal through a network; and a data crawler for performing data crawling for deep learning learning of the emotion server An artificial intelligence emotion method executed by an intelligent emotion system, comprising:
The emotion server,
Model emotion for determining the model of the article based on the image data;
Authenticity appraisal for determining whether the article corresponds to the authenticity of the model by performing image signal processing on a plurality of point portions of the article;
a state appraisal for determining a state index for indicating a preservation state of the article by determining one or more state factors of replacement, deformation, contamination and discoloration of the plurality of point portions of the article;
performing an appraised value determination to determine the appraised value of the article based on the condition index;
The emotion server includes an artificial intelligence-based emotion engine learned by deep learning,
The emotion engine learns data about a genuine model and a fake model for the plurality of point parts of the article,
The emotion engine, in the image data, determines whether the genuine model and the fake model are included with respect to the plurality of point parts, and also indexes the result of determining the degree of proximity to the genuine model and the fake model By displaying in the form of , the authenticity assessment is performed to evaluate the degree of proximity of the article to the genuine article, the fake article, or both;
The appraisal server constructs a three-dimensional model of the article from the image data by virtual reality or augmented reality, extracts the plurality of point parts of the article from the three-dimensional model, and executes the authenticity appraisal,
The appraisal server includes a big data-based price database that registers the ID of a product according to the model of the product, and records the product price obtained from a price reference site that provides price information of the product corresponding to the ID,
The emotion engine is
Perform image signal processing for determining one or more state factors of replacement, deformation, contamination, and discoloration of the plurality of point portions of the article, and determine a state index for indicating the preservation state of the article based on this; ,
assigning to each of the one or more condition factors a weighting factor representing a weighting factor representing a weighting factor in the price of the article, and calculating the depreciation of the article by the product of the condition index and the weighting factor, based on the depreciation degree to determine the appraised value of the article according to the following formula (1),
Assessed value = commodity price - (commodity price*weighting factor*status index)...(1)
When there are a plurality of point regions for determining each state factor, a value obtained by dividing the weighting coefficient of the state factor by the total number of point regions for determining the state factor is used as a sub weighting coefficient, An artificial intelligence emotion method that calculates the product of the weight coefficient and the state index of the corresponding state factor by summing the product of the weighting coefficient and the state index. delete A computer-readable recording medium in which a computer program for executing the artificial intelligence emotion method according to claim 8 is recorded.

Images