WO2021212873A1 - Defect detection method and apparatus for four corners of certificate, and device and storage medium - Google Patents

Abstract

A defect detection method and apparatus for four corners of a certificate, and a device and a storage medium. The method comprises: when a picture including a certificate is received, acquiring a feature point coordinate of a face in the certificate included in the picture, obtaining a coordinate clustering center and the minimum bounding rectangle on the basis of the feature point coordinate, and acquiring the area of the minimum bounding rectangle; calculating and obtaining a rectangular boundary of the certificate in the picture on the basis of the area of the minimum bounding rectangle and the coordinate clustering center; and determining a rectangular region formed by the rectangular boundary of the certificate in the picture, cutting the rectangular region into a plurality of sub-pictures, detecting whether there is a missing corner in the plurality of sub-pictures, and if there is a missing corner, determining that the certificate in the received picture is a defective certificate. In the method, by means of the minimum bounding rectangle and a coordinate clustering center of a certificate in a picture, a rectangular region corresponding to the certificate is determined, and whether each of the four corners of the rectangular region is missing is detected one by one, such that the detection accuracy is improved. The method further relates to blockchain technology, and the picture of the certificate can be stored in a blockchain node.

Description

Method, device, equipment and storage medium for detecting incomplete four corners of certificate

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on April 24, 2020, the application number is CN202010336193.3, and the title is "Detection method, device, equipment and storage medium for four corners of the document", and the entire content of it is approved The reference is incorporated in this application.

Technical field

This application relates to the field of detection technology, and in particular to a method, device, equipment, and storage medium for detecting the four corners of a certificate.

Background technique

With the development of artificial intelligence, there are more and more scenarios for the analysis of documents by artificial intelligence. For example, in the process of fund online remote account opening or credential authentication or identification, the integrity of the collected credential images needs to be analyzed. If the four corners of the collected ID image are incomplete, it will adversely affect the subsequent card analysis, such as text positioning and text recognition. Therefore, how to accurately detect whether the four corners of a document image are incomplete or not has become a topic that has attracted much attention from image researchers. At present, to determine whether the four corners of the document image are incomplete, the main method used in the industry is to use a neural network similar to inceptionv3 to detect the entire document image. The inventor realizes that other objects may be mistakenly collected during the collection process of the ID image, such as the background of the environment where the ID is located when the photo is taken. As a result, it causes interference in the detection of the entire ID image, which makes the detection exist. Large errors lead to inaccurate detection results.

Summary of the invention

The embodiment of the present application provides a method for detecting defects in the four corners of a document. The method for detecting defects in the four corners of a document includes the following steps:

When a picture containing a certificate is received, obtaining the feature point coordinates of the face in the certificate included in the picture, obtaining the coordinate cluster center and the smallest bounding rectangle based on the characteristic point coordinates, and obtaining the area of the smallest bounding rectangle;

Based on the area of the smallest bounding rectangle and the coordinate clustering center, the rectangular boundary of the certificate in the picture is calculated;

Extract the rectangular area formed by the rectangular border of the certificate in the picture, and detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the certificate in the received picture is a broken certificate.

This application also provides a device for detecting defects in four corners of a document, the device for detecting defects in four corners of a document includes:

The obtaining module is used to obtain, when a picture containing a certificate is received, the coordinate of the feature point of the face in the certificate contained in the picture, obtain the coordinate clustering center and the minimum bounding rectangle based on the coordinate of the feature point, and obtain the minimum The area of the circumscribing rectangle;

A calculation module, configured to calculate the rectangular boundary of the certificate in the picture based on the area of the smallest bounding rectangle and the coordinate clustering center;

The extraction module is used to extract the rectangular area formed by the rectangular boundary of the certificate in the picture, and to detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, determine whether there are missing corners in the received picture. The certificate is incomplete.

The application also provides a four-corner defect detection device for a document. The four-corner defect detection device includes a memory, a processor, and a four-corner defect detection program that is stored on the memory and can run on the processor. When the detection program is executed by the processor, the following steps are implemented:

When a picture containing a certificate is received, obtaining the feature point coordinates of the face in the certificate included in the picture, obtaining the coordinate cluster center and the smallest bounding rectangle based on the characteristic point coordinates, and obtaining the area of the smallest bounding rectangle;

Based on the area of the smallest bounding rectangle and the coordinate clustering center, the rectangular boundary of the certificate in the picture is calculated;

Extract the rectangular area formed by the rectangular border of the certificate in the picture, and detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the certificate in the received picture is a broken certificate.

The present application also provides a storage medium on which a four-corner defect detection program is stored, and the four-corner defect detection program is executed by a processor to implement the following steps:

When a picture containing a certificate is received, obtaining the feature point coordinates of the face in the certificate included in the picture, obtaining the coordinate cluster center and the smallest bounding rectangle based on the characteristic point coordinates, and obtaining the area of the smallest bounding rectangle;

Based on the area of the smallest bounding rectangle and the coordinate clustering center, the rectangular boundary of the certificate in the picture is calculated;

Extract the rectangular area formed by the rectangular border of the certificate in the picture, and detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the certificate in the received picture is a broken certificate.

Description of the drawings

FIG. 1 is a schematic diagram of the structure of a detection device for four corners of a credential defect in a hardware operating environment involved in a solution of an embodiment of the application;

2 is a schematic flowchart of a first embodiment of a method for detecting four corners of a credential incompleteness of the application;

FIG. 3 is a schematic diagram of functional modules of a preferred embodiment of a device for detecting defects at four corners of a document in this application.

The realization, functional characteristics, and advantages of the purpose of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings.

Detailed ways

It should be understood that the specific embodiments described here are only used to explain the present application, and are not used to limit the present application.

As shown in FIG. 1, FIG. 1 is a schematic diagram of the structure of the four corners of the credential defect detection device of the hardware operating environment involved in the solution of the embodiment of the present application.

In the following description, the use of suffixes such as “module”, “part” or “unit” used to indicate elements is only for the description of the present application, and has no specific meaning in itself. Therefore, "module", "part" or "unit" can be used in a mixed manner.

The four-corner defect detection device of the examples of this application can be a PC, or a portable terminal device such as a tablet computer and a portable computer.

As shown in FIG. 1, the device for detecting defects in the four corners of the document may include a processor 1001, such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. Among them, the communication bus 1002 is used to implement connection and communication between these components. The user interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as a magnetic disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

Those skilled in the art can understand that the structure of the four-corner defect detection device shown in FIG. 1 does not constitute a limitation on the four-corner defect detection device of the document, and may include more or less components than shown in the figure, or a combination of certain components, Or different component arrangements.

As shown in FIG. 1, the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a detection program.

In the device shown in FIG. 1, the network interface 1004 is mainly used to connect to the back-end server and communicate with the back-end server; the user interface 1003 is mainly used to connect to the client (user side) and communicate with the client; and the processor 1001 can be used to call the detection program stored in the memory 1005 and perform the following operations:

When a picture containing a certificate is received, obtaining the feature point coordinates of the face in the certificate included in the picture, obtaining the coordinate cluster center and the smallest bounding rectangle based on the characteristic point coordinates, and obtaining the area of the smallest bounding rectangle;

Based on the area of the smallest bounding rectangle and the coordinate clustering center, the rectangular boundary of the certificate in the picture is calculated;

Extract the rectangular area formed by the rectangular border of the certificate in the picture, and detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the certificate in the received picture is a broken certificate.

Further, the step of calculating the rectangular boundary of the certificate in the picture based on the area of the smallest bounding rectangle and the coordinate clustering center includes:

Identify the type of the certificate in the picture, determine the target standard certificate corresponding to the type according to the type of the certificate, and obtain the target standard area of the target circumscribed rectangle in the target standard certificate;

Based on the proportional relationship between the target standard document and the area of the smallest circumscribed rectangle, the length and width of the certificate are calculated, wherein the proportional relationship includes the difference between the target standard length of the target standard certificate and the target standard area The first target proportional relationship between the two, and the second target proportional relationship between the target standard width of the target standard certificate and the target standard area;

Based on the relative position relationship of the target standard certificate, the length and width of the certificate, and the coordinate clustering center, the rectangular boundary of the certificate is determined, wherein the relative position relationship includes the target standard of the target standard certificate The first target relative positional relationship between the length and the target cluster center of the target standard certificate, and the second target relative between the target standard width of the target standard certificate and the target cluster center of the target standard certificate Positional relationship.

Further, the step of determining the rectangular boundary of the document based on the relative position relationship of the target standard document, the length and width of the document, and the coordinate clustering center includes:

Determine the length boundary direction of the certificate in the picture based on the relative positional relationship between the coordinate clustering center and the first target;

Determine the width boundary direction of the document in the picture based on the relative positional relationship between the coordinate clustering center and the second target;

The length boundary of the document is determined based on the length of the document and the direction of the length boundary, and the width boundary of the document is determined based on the width of the document and the direction of the width boundary. The width boundary determines the rectangular boundary of the document.

Further, after the step of extracting the rectangular area formed by the rectangular boundary of the certificate in the picture, the processor 1001 may be used to call the detection program stored in the memory 1005 and perform the following operations:

Acquiring the length and width of the rectangular area, and calculating the pixel value corresponding to the length of the rectangular area, and whether any one of the pixel values corresponding to the width of the rectangular area is greater than a preset threshold;

If any one of the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area is greater than the preset threshold, the rectangular area is divided according to the first preset segmentation method Is a plurality of sub-pictures, and the sub-pictures corresponding to the four corners of the rectangular area among the plurality of sub-pictures are used as target sub-pictures to detect whether there are defects in the four corners of the rectangular area based on each of the target sub-pictures. Horn;

If the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area are both less than or equal to the preset threshold value, the rectangular area is divided into multiple segments according to the second preset segmentation method. And use multiple sub-pictures as target sub-pictures to detect whether there are missing corners in the four corners of the rectangular area based on each of the target sub-pictures.

Further, the step of detecting whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, determining that the certificate in the received picture is a broken certificate includes:

Detect each of the target sub-pictures one by one through a preset method, and determine whether the target sub-pictures all include a complete ID corner;

If the target sub-pictures all contain a complete ID corner, it is determined that the received picture is valid;

If there is a missing corner in any one of the target sub-images in the target sub-picture, it is determined that the certificate in the received picture is a broken certificate, the received picture is determined to be invalid, and a re-collection prompt message is output.

Further, when the picture containing the credential is received, the feature point coordinates of the face in the credential included in the picture are obtained, the coordinate clustering center and the smallest bounding rectangle are obtained based on the feature point coordinates, and the smallest bounding rectangle is obtained. The steps to circumscribe the area of the rectangle include:

When a picture containing a certificate is received, extract multiple feature points of the face in the certificate through a preset neural network, and obtain feature point coordinates of the multiple feature points;

Clustering the coordinate of each feature point of the face in the certificate by a preset algorithm, and obtain the center coordinate of the coordinate of each feature point of the face in the certificate as a coordinate clustering center;

Determine the minimum circumscribed rectangle according to the position of each of the feature point coordinates of the face in the document, and detect the length and width of the minimum circumscribed rectangle, and calculate the minimum circumscribed rectangle by the length and width of the minimum circumscribed rectangle The area of the rectangle.

Further, before the step of calculating the rectangular boundary of the document in the picture based on the area of the smallest bounding rectangle and the coordinate clustering center, the processor 1001 may be used to call the detection stored in the memory 1005 Program and do the following:

Obtain the standard area, standard length, standard width, standard clustering center, and the minimum standard area of the feature circumscribed rectangle of the facial feature points in the standard certificate of each type of standard certificate;

Perform the following steps one by one for each type of the standard certificate:

Based on the standard length and the minimum standard area, generate a first proportional relationship between the standard length and the minimum standard area, and generate the standard width and the minimum standard width based on the standard width and the minimum standard area. The second proportional relationship of the standard area;

Based on the standard length and the standard cluster center, a first relative positional relationship between the standard cluster center and the standard length is generated, and based on the standard width and the standard cluster center, the standard cluster is generated The second relative positional relationship between the class center and the standard width.

In order to better understand the above technical solutions, exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the drawings show exemplary embodiments of the present disclosure, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

In order to better understand the above technical solutions, the above technical solutions will be described in detail below in conjunction with the accompanying drawings of the specification and specific implementations.

2, the first embodiment of the present application provides a schematic flowchart of a method for detecting the four corners of a credential defect. In this embodiment, the method for detecting defects in the four corners of the document includes the following steps:

Step S10, when a picture containing a certificate is received, the feature point coordinates of the face in the certificate contained in the picture are obtained, the coordinate clustering center and the minimum circumscribed rectangle are obtained based on the characteristic point coordinates, and the minimum circumscribed rectangle is obtained Area

The method for detecting incomplete four corners of a document in this embodiment is applied to a server, where the server is in communication connection with a computer, a tablet, a smart phone, and other terminals. The server is provided with an identification program and a preset neural network, and the identification program at least includes identification technology with a document For example, a program based on OCR recognition and a program based on face recognition technology such as feature point recognition are used for document recognition and face recognition, respectively. Among them, document recognition can identify various types of documents. The types of documents can be ID cards, driving licenses, marriage certificates, etc. Face recognition can identify facial feature points and extract the coordinates of the feature points. The feature points are based on the nose, Face features such as mouth, eyes, ears, etc. are generated. The preset neural network is used to detect facial feature points. The neural network can be dlib or face recognize.

Further, when a user conducts online remote account opening or identification of a certificate due to certain circumstances, he needs to take a picture containing his certificate and upload the captured picture to the terminal. Understandably, if the user uses a smart phone to take a picture, then The pictures can be uploaded directly to the server. Further, when the terminal receives the picture containing the certificate uploaded by the user, it transmits the picture to the server, so that the server can recognize the face in the picture through the recognition program and extract the coordinates of the facial feature points in the certificate. Further, the server clusters the extracted coordinates of each feature point to obtain a coordinate clustering center that characterizes the center position of each feature point of the face. Further, the server determines the smallest enclosing rectangle that includes all the feature points of the face in the picture according to the coordinates of each feature point, and further detects the length and width of the smallest enclosing rectangle to calculate the area of the smallest enclosing rectangle based on the length and width.

Further, when a picture containing a certificate is received, the characteristic point coordinates of the face in the certificate contained in the picture are obtained, the coordinate clustering center and the minimum circumscribed rectangle are obtained based on the characteristic point coordinates, and the minimum circumscribed rectangle is obtained The steps of the area include:

Step S11, when a picture containing a document is received, extract multiple feature points of the face in the document through a preset neural network, and obtain feature point coordinates of the multiple feature points;

Further, when the terminal receives the image uploaded by the user that contains the certificate, the image is sent to the server, and the server uses the edge line detection function of opencv (Open Source Computer Vision Library) to check the tilt in the image of the verification piece. The degree of perspective transformation is further used. When the document font is positive, the lower left corner is taken as the origin, and the position of the lower left corner is regarded as 0 degree, and the document is corrected to 0 degree, 90 degree, 180 degree, and 270 degree to correct the document to Parallel to the border of the picture, it is easy to identify and detect the document. Further, the server transmits the corrected picture to the preset neural network, and through the detection of the preset neural network, recognizes the face with the most feature points in the picture, extracts multiple feature points from the face, and obtains each feature The characteristic point coordinates of the point.

Step S12, clustering the coordinates of each feature point of the face in the document by a preset algorithm to obtain the center coordinates of the coordinates of each feature point of the face in the document as the coordinate clustering center;

Further, the server clusters the coordinates of each feature point by using a preset algorithm. In this embodiment, the knn clustering algorithm is used to calculate the center coordinates of the coordinates of each feature point through the knn clustering algorithm, which is used to characterize the location of each feature point. The cluster center of the center point of the area. For example, the server recognizes the coordinates of facial feature points such as eyes, nose, mouth, ears, etc. through a recognition program, and calculates the center coordinates of the coordinates of the above multiple facial feature points through a clustering algorithm, and uses the center coordinates as the coordinate clustering center .

Step S13: Determine a minimum circumscribed rectangle according to the position of each of the feature point coordinates of the face in the document, and detect the length and width of the minimum circumscribed rectangle, and calculate the minimum circumscribed rectangle by the length and width of the minimum circumscribed rectangle. State the area of the smallest enclosing rectangle.

Further, the server determines the smallest enclosing rectangle that represents the feature points of all human faces according to the position of the coordinates of each feature point in the face, and detects the length and width of the smallest enclosing rectangle, which is obtained by combining the length and width with the rectangular area calculation formula The area of the smallest enclosing rectangle. For example, the server obtains the location of facial features such as eyes, mouth, ears, nose, etc., and uses a smallest circumscribed rectangle to include all the acquired facial features, and measures the length and width of the smallest circumscribed rectangle. Then calculate the area of the smallest enclosing rectangle according to the rectangle area formula.

Step S20: Based on the area of the smallest bounding rectangle and the coordinate clustering center, the rectangular boundary of the certificate in the picture is calculated;

Further, the server pre-identifies various types of standard certificates, obtains the length, width, area, and other size data of each type of standard certificate, and stores the standard certificate type and the size data of the standard certificate. The standard certificates are produced according to national standards. Such as ID card, marriage certificate, driver's license and other types of documents. Further, the server recognizes the certificate in the received picture through the recognition program, and searches for the standard certificate corresponding to the certificate type in the picture from the stored standard certificates of various types, and determines the standard certificate corresponding to the certificate type in the picture as the target Standard documents. Further, obtain the proportional relationship between the target standard area of the target standard certificate and the target standard length and target standard width included in the target standard certificate, and the target cluster center of the target standard certificate and the target standard length and target standard width respectively. The relative position relationship formed, and the target standard area of the target circumscribed rectangle containing the feature points of all faces in the target standard certificate. Further, the coordinate clustering center obtained by the above calculation and the area of the smallest circumscribed rectangle in the picture are combined with the proportional relationship, relative position relationship and the target standard area of the target circumscribed rectangle in the target standard certificate, and then the target standard area of the target circumscribed rectangle The rectangular border formed in the picture.

Step S30: Extract the rectangular area formed by the rectangular boundary of the certificate in the picture, and detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the received certificate in the picture is Incomplete documents.

Further, the coordinates of the vertices of the rectangle boundary are detected, and the coordinates of the vertices of the four corners of the rectangle are obtained. In order to ensure the correctness of the coordinate values, this embodiment is provided with an adjustment mechanism; an adjustment value a for adjustment is preset, and the adjustment value a is preferably 30 based on experience. For the x1, x2, y1, and y2 of the composition coordinates (x1, y1), (x1, y2), (x2, y2), (x2, y1), they can be adjusted by the following formula: x1-(x2-x1)/ a, x2+(x2-x1)/a, y1-(y2-y1)/a, y2+(y2-y1)/a. Further, the coordinate values of the vertices of the four corners of the rectangle are extracted, and the coordinates of the vertices of the four corners of the extracted rectangle are connected by a preset method, to obtain a rectangular area that characterizes the location of the document. Further, the rectangular area is segmented according to the first preset segmentation method or the second preset segmentation method according to different definitions, and multiple target sub-pictures are generated, and the target sub-images corresponding to the four corner areas of the rectangular area are detected. If there is a missing corner in any of the target sub-pictures, if any of the target sub-pictures is detected as a missing corner, the ID in the received image is determined to be a broken ID, and the invalid ID information is output and the user is prompted to take a new photo containing the ID picture of. It is understandable that the certificate in the received picture is incomplete, which may be due to the incompleteness of the certificate itself, or it may be due to the fact that the certificate was blocked by the customer when the picture was taken. Further, if it is detected that there are no missing corners in all target sub-pictures, it is determined that the credentials in the received pictures are valid, and the business is processed according to the user's request.

According to the method for detecting the four corners of a credential incompleteness of this embodiment, when a picture containing a credential is received, the feature point coordinates of the face in the credential contained in the picture are obtained, and the coordinate clustering center and the smallest bounding rectangle are obtained based on the feature point coordinates, And obtain the area of the smallest bounding rectangle; based on the area of the smallest bounding rectangle and the coordinate clustering center, calculate the rectangular boundary of the certificate in the picture; extract the rectangle of the certificate in the picture A rectangular area formed by the boundary, and detecting whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the certificate in the received picture is a broken certificate. Determine the rectangular area corresponding to the document through the smallest enclosing rectangle and coordinate clustering center of the document in the picture, and detect whether the four corners of the rectangular area are missing corners one by one. Compared with the detection based on the entire document picture, the detection accuracy is improved. It can accurately detect whether the certificate in the picture is incomplete.

Further, based on the first embodiment of the four-corner defect detection method of the present application, a second embodiment of the four-corner defect detection method of the present application is proposed. In the second embodiment, clustering is based on the area of the smallest bounding rectangle and the coordinate In the center, the step of calculating the rectangular boundary of the certificate in the picture includes:

Step S21: Identify the type of the certificate in the picture, determine the target standard certificate corresponding to the type according to the type of the certificate, and obtain the target standard area of the target circumscribed rectangle in the target standard certificate;

Further, the server recognizes the type of the certificate in the picture through the recognition program. For example, if it recognizes that the type of the certificate is an ID card, it determines that the target standard certificate corresponding to the type of the certificate in the picture is an ID card, and obtains that the standard ID card contains all faces. The target standard area of the target circumscribed rectangle of the feature point.

Step S22: Calculate the length and width of the document based on the proportional relationship of the target standard document and the area of the minimum circumscribed rectangle, wherein the proportional relationship includes the target standard length of the target standard document and the target A first target proportional relationship between standard areas, and a second target proportional relationship between the target standard width of the target standard certificate and the target standard area;

Further, according to the first target ratio relationship between the target standard length of the target standard certificate and the target standard area, combined with the minimum circumscribed rectangle area, the length of the certificate in the picture is calculated, and the target standard width and target standard width of the target standard certificate are calculated. The second target ratio relationship between the standard areas, combined with the minimum circumscribed rectangle area, calculates the width of the document in the picture. For example, the first target proportional relationship in the target standard document proportional relationship is a1, and the minimum circumscribed rectangle area is m, then the length of the document in the picture can be obtained as a1*m, and the second target proportional relationship is a2, and the smallest circumscribed rectangle area is m, the width of the certificate in the picture can be obtained as a2*m, and the corresponding length of the certificate is determined as a1*m, and the width of the rectangular area is a2*m.

Step S23: Determine the rectangular boundary of the document based on the relative position relationship of the target standard document, the length and width of the document, and the coordinate clustering center, wherein the relative position relationship includes the target standard document The first target relative position relationship between the target standard length of the target standard certificate and the target cluster center of the target standard certificate, and the first target standard width between the target standard width of the target standard certificate and the target cluster center of the target standard certificate 2. The relative position of the target.

Further, according to the relative positional relationship of the first target between the target standard length of the target standard certificate and the target cluster center, combined with the coordinate cluster center, the length boundary direction in the rectangular boundary is obtained. From the length of the document and the direction of the length boundary, the length boundary is obtained. Further, according to the relative positional relationship of the second target between the target standard width of the target standard certificate and the target cluster center, combined with the coordinate cluster center, the width boundary direction in the rectangular boundary is obtained. From the width of the document and the direction of the width boundary, the width boundary is obtained. Further, from the length boundary and the length of the document, as well as the width boundary and the width of the document, the rectangular boundary of the document is obtained. Specifically, based on the relative position relationship of the target standard document, the length and width of the document, and the coordinate clustering center, the step of determining the rectangular boundary of the document includes:

Step S231: Determine the length boundary direction of the document in the picture based on the relative positional relationship between the coordinate clustering center and the first target;

Further, the first target relative position relationship between the target standard length of the target standard certificate and the target cluster center is obtained, and the coordinate cluster center is taken as the center, and the target cluster center and the target are characterized by the relative position relationship of the first target. The distance relationship between the standard lengths, that is, the distance from the target cluster center to the length direction of the target standard length, determines the distance extending in the length direction perpendicular to the minimum bounding rectangle, and the end point is determined by the extended distance. Further, extending from the end point to the length direction parallel to the minimum circumscribed rectangle, the length boundary direction of the rectangle is obtained.

Step S232: Determine the width boundary direction of the document in the picture based on the relative positional relationship between the coordinate clustering center and the second target;

Further, the relative position relationship of the second target between the target standard width of the target standard certificate and the target cluster center is obtained, and the coordinate cluster center is taken as the center, and the target cluster center and the target are characterized by the relative position relationship of the second target. The distance relationship between the standard widths, that is, the distance from the target cluster center to the width direction where the target standard width is located, determines the distance extending in the width direction perpendicular to the minimum bounding rectangle, and the end point is determined by the extended distance. Further, extending from the end point to the width direction parallel to the minimum circumscribed rectangle, the width boundary direction of the rectangle is obtained.

Step S233: Determine the length boundary of the document based on the length of the document and the direction of the length boundary, and determine the width boundary of the document based on the width of the document and the direction of the width boundary, and determine the width boundary of the document based on the length of the document. And the width boundary determine the rectangular boundary of the document.

Further, the determined length boundary direction is extended to both sides and the width boundary direction is extended to both sides to obtain the intersection point of the length boundary direction and the width boundary direction, and the intersection point is determined as the vertex coordinates representing one of the four corner vertices of the document. The vertex coordinates are combined with the length boundary direction and the length of the document in the picture to determine the length boundary, and the vertex coordinates are combined with the width boundary direction and the width of the document in the picture to determine the width boundary. Further, a rectangular frame is formed based on the length boundary and the width boundary to obtain the rectangular boundary of the certificate in the picture.

In this embodiment, the area of the smallest enclosing rectangle that contains the most facial feature points of the document in the picture and the coordinate clustering center representing the center of the facial feature points are combined with the proportional relationship and relative position relationship of the target standard document to calculate that the document is in the picture. Corresponding to the rectangular boundary in, accurately obtain the rectangular area boundary corresponding to the document, so that the detection result is more accurate.

Further, based on the first embodiment or the second embodiment of the four-corner defect detection method of the present application, a third embodiment of the four-corner defect detection method of the present application is proposed. In the third embodiment, the image is extracted by the certificate The steps following the step of forming a rectangular area by the rectangular boundary include:

Step S31: Obtain the length and width of the rectangular area, and calculate the pixel value corresponding to the length of the rectangular area, and whether any one of the pixel values corresponding to the width of the rectangular area is greater than a preset threshold;

Further, the server obtains the length and width of the rectangular area corresponding to the certificate, and calculates the pixel value corresponding to the length of the rectangular area according to the length value and pixels of the rectangular area, and calculates the pixel value corresponding to the rectangular area according to the width and pixels of the rectangle. The pixel value corresponding to the width. Further, it is judged whether any one of the pixel value corresponding to the length of the rectangular area and the pixel value corresponding to the width of the rectangular area is greater than a preset threshold that characterizes sufficient or insufficient sharpness, to determine the rectangle in the picture that characterizes the document. Whether the content of the area is clear, it is convenient to determine whether the rectangular area is divided into equal parts of sub-pictures.

Step S32, if any one of the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area is greater than a preset threshold, the rectangular area is divided according to a first preset The method is divided into multiple sub-pictures, and the sub-pictures corresponding to the four corners of the rectangular area among the multiple sub-pictures are used as target sub-pictures to detect the four corners of the rectangular area based on each of the target sub-pictures. Whether there are missing corners;

Further, if any one of the calculated pixel value corresponding to the length of the rectangular area and the pixel value corresponding to the width of the rectangular area is greater than the preset threshold, the preset threshold in this embodiment is set to 2000 based on experience, It means that the content of the rectangular area is clear enough, and the rectangular area is divided into a plurality of sub-pictures according to the preset first preset segmentation method, wherein the first preset segmentation method is set according to requirements, as in this embodiment. Set the nine-equal division method to divide the rectangular area into 9 sub-pictures. Thereafter, from the segmented sub-pictures, the sub-pictures located in the four corners of the rectangular area are determined as the target sub-pictures. For example, for the above-mentioned 9 sub-pictures, the 4 sub-pictures located in the upper left corner, upper right corner, lower left corner, and lower right corner are the sub-pictures corresponding to the four corners of the rectangular area, and they are used as the target sub-pictures to facilitate Detect whether there are missing corners in the 4 sub-pictures.

Step S33: If the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area are both less than or equal to a preset threshold, the rectangular area is cut according to a second preset splitting method. Divide into a plurality of sub-pictures, and use the plurality of sub-pictures as target sub-pictures to detect whether there are missing corners in the four corners of the rectangular area based on each of the target sub-pictures.

Further, if the calculated pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area are both less than or equal to the preset threshold, it indicates that the definition of the rectangular area is insufficient, and the rectangular area is set according to the preset threshold. The second preset segmentation method is divided into a plurality of sub-pictures as target sub-pictures, and the second preset segmentation method is set according to requirements. For example, in this embodiment, a four-equal division method is preset, and the rectangular area is divided into 4 sub-pictures, and the 4 sub-pictures are used as target sub-pictures, so as to detect whether there are missing corners in the 4 target sub-pictures.

Further, the step of detecting whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, determining that the certificate in the received picture is a broken certificate includes;

Step S34: Detect each of the target sub-pictures one by one in a preset manner, and determine whether the target sub-pictures all include a complete document corner;

Further, this embodiment is preset with a preset method for detecting whether a picture is incomplete, such as the inceptionv3 detection method, to detect the 4 target sub-pictures obtained by segmentation one by one through the preset method, specifically, the 4 target sub-pictures are detected one by one. The sub-pictures are imported into inceptionv3 one by one for detection to determine whether the four target sub-pictures all contain a complete ID corner. Understandably, the integrity of the document corner is obtained by comparing it with the document corner of the target standard document. For example, if the document corner of the ID card is rounded, it will be detected whether the document corner contained in the target sub-picture is the same as that of the ID card. The document angle is the same.

Step S35, if the target sub-pictures all contain a complete ID corner, it is determined that the received picture is valid;

Further, if the four target sub-pictures are checked one by one, and it is determined that the four target sub-pictures all contain complete certificate corners, it means that the four corners of the certificate in the picture are complete and not damaged, then accept The certificate in the picture is complete, and the picture received is judged to be valid. Further, according to the user's request, the business processing corresponding to the user's request is performed.

Step S36, if there is a missing corner in any one of the target sub-images in the target sub-picture, determine that the certificate in the received picture is a broken certificate, determine that the received picture is invalid, and output the re-collected Prompt information.

Further, if the four target sub-pictures are detected one by one, and it is detected that the certificate corner in any one of the four target sub-pictures is incomplete, it means that the certificate in the picture is incomplete. It is determined that it is a broken document, where the broken document may be that the document itself is worn or accidentally caused by the loss of the document corner when the document is in use, or it may be that the user covers the corner of the document when the document is photographed. As a result, one of the document corners of the document in the photographed picture is missing. Further, it is determined that the received picture is invalid, and a message is output to prompt the user to take the photo of the certificate again.

In this embodiment, the server detects the pixel value corresponding to the length of the rectangular area and the pixel value corresponding to the width of the rectangular area to determine whether the content of the rectangular area that characterizes the document is sufficiently clear, and differentiates the rectangular area based on different definitions. The number of sub-pictures are segmented to more accurately detect the four corners of the rectangular area. By detecting whether the target sub-pictures selected from the sub-pictures all contain a complete ID corner, it is determined whether the received picture is valid. This embodiment realizes that the complete document angle detection is performed on the segmented pictures to determine whether the document is complete, so as to facilitate the user to perform business processing such as document recognition or remote account opening.

Further, based on the first embodiment, the second embodiment or the third embodiment of the four-corner defect detection method of the present application, a fourth embodiment of the four-corner defect detection method of the present application is proposed. In the fourth embodiment, based on the minimum Before calculating the area of the circumscribed rectangle and the coordinate clustering center, the step of obtaining the rectangular boundary of the certificate in the picture includes:

Step S40: Obtain the standard area, standard length, standard width, standard cluster center of each type of standard certificate, and the minimum standard area of the characteristic circumscribed rectangle of the facial feature points in the standard certificate;

Further, the size data of various types of standard certificates actually used are obtained in advance, and the obtained content includes the standard area, standard length, and standard width of the standard certificate, as well as the characteristic circumscribed rectangle containing the facial feature points in the standard certificate. The minimum standard area of, and the standard cluster center for clustering each face feature point, so that the server can call the size data for comparison and calculation. Understandably, considering that different types of standard documents have different outer frame sizes, or even if the outer frame sizes are the same, they may have different sizes of faces or faces in different positions, so that different types of standard documents are in the standard area. There are differences in standard length, standard width, minimum standard area, and standard clustering center. Therefore, the above-mentioned size data can be obtained for different types of standard documents. Steps S50, S60, and S70 are processed one by one.

Step S50, based on the standard length and the minimum standard area, generate a first proportional relationship between the standard length and the minimum standard area, and generate the standard width and the minimum standard area based on the standard width and the minimum standard area. The second proportional relationship of the minimum standard area;

The ratio between the standard length of the standard document and the minimum standard area is generated to generate the first proportional relationship that characterizes the relationship between the standard length and the minimum standard area. The ratio between the standard width and the minimum standard area of the standard document is used to generate the standard width and the minimum standard area. The second proportional relationship of the minimum standard area relationship.

Step S60, based on the standard length and the standard cluster center, generate a first relative positional relationship between the standard cluster center and the standard length, and generate all the standard cluster centers based on the standard width and the standard cluster centers. The second relative positional relationship between the standard cluster center and the standard width.

Further, the distance between the standard cluster center in the standard document and the standard length of the standard document is calculated. This distance represents the positional relationship of the standard cluster center relative to the standard length, and it is regarded as the standard cluster center and the standard length. The first relative positional relationship between the long. In addition, calculate the distance between the standard cluster center in the standard document and the standard width of the standard document. This distance represents the positional relationship between the standard cluster center and the standard width. The second relative position relationship between.

In this embodiment, the size data information of each type of standard certificate is obtained in advance, and based on the size data information of the standard certificate, the proportional relationship and relative position relationship of the standard certificate are generated, so that the server can call the size data of the standard certificate and the proportion of the standard certificate. Relationship and relative position relationship, calculate the length and width of the certificate in the received picture and determine the specific position of the rectangular area corresponding to the certificate.

Furthermore, this application also provides a device for detecting the four corners of the credential.

Referring to FIG. 3, FIG. 3 is a schematic diagram of the functional modules of the first embodiment of the device for detecting four corners of the credential defect of this application.

The four-corner defect detection device of the document includes:

The obtaining module 10 is configured to, when a picture containing a certificate is received, obtain the feature point coordinates of the face in the certificate contained in the picture, obtain the coordinate cluster center and the smallest bounding rectangle based on the feature point coordinates, and obtain the The area of the smallest enclosing rectangle;

The calculation module 20 is configured to calculate the rectangular boundary of the certificate in the picture based on the area of the smallest bounding rectangle and the coordinate clustering center;

The extraction module 30 is configured to extract the rectangular area formed by the rectangular boundary of the certificate in the picture, and detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, determine whether there is a missing corner in the received picture The document is incomplete.

In the four-corner defect detection device of a document of this embodiment, when a picture containing a document is received, the acquiring module 10 first obtains the coordinates of the characteristic points of the face in the document contained in the picture, and obtains the coordinate clustering center based on the coordinates of the characteristic points. And the smallest circumscribed rectangle, and obtain the area of the smallest circumscribed rectangle; then the calculation module 20 calculates the rectangular boundary of the document in the picture based on the area of the smallest circumscribed rectangle and the coordinate clustering center; The extraction module 30 extracts the rectangular area formed by the rectangular border of the certificate in the picture, and detects whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the certificate in the received picture is It is a broken document. Determine the rectangular area corresponding to the document through the smallest enclosing rectangle and coordinate clustering center of the document in the picture, and check whether the four corners of the rectangular area are missing corners one by one. Compared with the detection based on the entire document picture, the detection accuracy is improved. It can accurately detect whether the certificate in the picture is incomplete.

Further, the acquisition module 10 includes:

The first acquiring unit is configured to extract multiple feature points of the face in the document through a preset neural network when a picture containing the document is received, and acquire the feature point coordinates of the multiple feature points;

The clustering unit is used to cluster the coordinates of each feature point of the face in the certificate by using a preset algorithm to obtain the center coordinates of the coordinates of each feature point of the face in the certificate, as the coordinate clustering center;

The first determining unit is configured to determine the minimum circumscribed rectangle according to the position of each of the characteristic point coordinates of the face in the document, and detect the length and width of the minimum circumscribed rectangle, and pass the length and the sum of the minimum circumscribed rectangle The width is calculated to obtain the area of the minimum circumscribed rectangle.

Further, the calculation module 20 includes:

An identification unit, configured to identify the type of the certificate in the picture, determine the target standard certificate corresponding to the type according to the type of the certificate, and obtain the target standard area of the target circumscribed rectangle in the target standard certificate;

The calculation unit is configured to calculate the length and width of the document based on the proportional relationship of the target standard document and the area of the minimum circumscribed rectangle, wherein the proportional relationship includes the target standard length of the target standard document and the area The first target proportional relationship between the target standard areas, and the second target proportional relationship between the target standard width of the target standard certificate and the target standard area;

The second determining unit is configured to determine the rectangular boundary of the document based on the relative positional relationship of the target standard document, the length and width of the document, and the coordinate clustering center, wherein the relative positional relationship includes all The first target relative position relationship between the target standard length of the target standard certificate and the target cluster center of the target standard certificate, and the target standard width of the target standard certificate and the target cluster center of the target standard certificate The relative positional relationship between the second target.

Further, the calculation module 20 further includes:

A third determining unit, configured to determine the length boundary direction of the document in the picture based on the relative positional relationship between the coordinate clustering center and the first target;

A fourth determining unit, configured to determine the width boundary direction of the document in the picture based on the relative positional relationship between the coordinate clustering center and the second target;

The fifth determining unit is configured to determine the length boundary of the document based on the length of the document and the length boundary direction, and determine the width boundary of the document based on the width of the document and the width boundary direction, and The length boundary and the width boundary determine the rectangular boundary of the document.

Further, the calculation module 20 further includes:

The second acquiring unit is used to acquire the standard area, standard length, standard width, standard clustering center, and the minimum standard area of the characteristic circumscribed rectangle of the facial feature points in the standard certificate of various types of standard certificates;

The execution unit is used to execute the following steps one by one for each type of the standard certificate:

The first generating unit is configured to generate a first proportional relationship between the standard length and the minimum standard area based on the standard length and the minimum standard area, and generate the standard width and the minimum standard area based on the standard length and the minimum standard area. The second proportional relationship between the standard width and the minimum standard area;

The second generating unit is configured to generate a first relative positional relationship between the standard cluster center and the standard length based on the standard length and the standard cluster center, and based on the standard width and the standard cluster Center, generating a second relative positional relationship between the standard cluster center and the standard width.

Further, the extraction module 30 includes:

The third acquiring unit is configured to acquire the length and width of the rectangular area, and calculate the pixel value corresponding to the length of the rectangular area, and whether any one of the pixel values corresponding to the width of the rectangular area is greater than Preset threshold;

The first segmentation unit is used for if any one of the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area is greater than a preset threshold, then the rectangular area is set according to the first A preset segmentation method is divided into a plurality of sub-pictures, and the sub-pictures corresponding to the four corners of the rectangular area among the plurality of sub-pictures are used as target sub-pictures to detect the Whether there are missing corners in the four corners of the rectangular area;

The second segmentation unit is configured to: if the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area are both less than or equal to the preset threshold, then the rectangular area is divided according to the second preset threshold. Suppose the segmentation method is divided into multiple sub-pictures, and the multiple sub-pictures are used as target sub-pictures to detect whether there are missing corners in the four corners of the rectangular area based on each of the target sub-pictures.

Further, the extraction module 30 further includes:

The detection unit is configured to detect each of the target sub-pictures one by one in a preset manner to determine whether the target sub-pictures all include a complete document corner;

The first determining unit is configured to determine that the received image is valid if all the target sub-pictures contain a complete document corner;

The second determining unit is configured to determine that the credential in the received picture is a broken credential if there is a missing corner in any one of the target sub-images, determine that the received picture is invalid and The prompt message for re-collection is output.

The application also provides a four-corner defect detection device for a document. The four-corner defect detection device includes a memory, a processor, and a four-corner defect detection program that is stored on the memory and can run on the processor. When the detection program is executed by the processor, the following steps are implemented:

When a picture containing a certificate is received, obtaining the feature point coordinates of the face in the certificate included in the picture, obtaining the coordinate cluster center and the smallest bounding rectangle based on the characteristic point coordinates, and obtaining the area of the smallest bounding rectangle;

Based on the area of the smallest bounding rectangle and the coordinate clustering center, the rectangular boundary of the certificate in the picture is calculated;

Extract the rectangular area formed by the rectangular border of the certificate in the picture, and detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the certificate in the received picture is a broken certificate.

The present application also provides a storage medium, the storage medium may be volatile or non-volatile, the storage medium stores a four-corner defect detection program of the document, and the four-corner defect detection program is executed by a processor. When implementing the following steps:

When a picture containing a certificate is received, obtaining the feature point coordinates of the face in the certificate included in the picture, obtaining the coordinate cluster center and the smallest bounding rectangle based on the characteristic point coordinates, and obtaining the area of the smallest bounding rectangle;

Based on the area of the smallest bounding rectangle and the coordinate clustering center, the rectangular boundary of the certificate in the picture is calculated;

Extract the rectangular area formed by the rectangular border of the certificate in the picture, and detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the certificate in the received picture is a broken certificate.

In the embodiments of the four-corner defect detection device and storage medium of the present application, all the technical features of each embodiment of the four-corner defect detection method of the above-mentioned document are included. This will not be exhausted.

It should be noted that in this article, the terms "include", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements not only includes those elements, It also includes other elements that are not explicitly listed, or elements inherent to the process, method, article, or system. If there are no more restrictions, the element defined by the sentence "including a..." does not exclude the existence of other identical elements in the process, method, article, or system that includes the element.

The blockchain referred to in this application is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

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

Through the description of the above implementation manners, those skilled in the art can clearly understand that the above-mentioned embodiment method can be implemented by means of software plus the necessary general hardware platform, of course, it can also be implemented by hardware, but in many cases the former is better.的实施方式。 Based on this understanding, the technical solution of this application essentially or the part that contributes to the existing technology can be embodied in the form of a software product, and the computer software product is stored in a storage medium (such as ROM/RAM) as described above. , Magnetic disks, optical disks), including several instructions to make a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) execute the method described in each embodiment of the present application.

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

Claims (20)

1. A method for detecting defects in the four corners of a document, wherein the method for detecting defects in the four corners of a document includes the following steps:

   When a picture containing a certificate is received, obtaining the feature point coordinates of the face in the certificate included in the picture, obtaining the coordinate cluster center and the smallest bounding rectangle based on the characteristic point coordinates, and obtaining the area of the smallest bounding rectangle;

   Based on the area of the smallest bounding rectangle and the coordinate clustering center, the rectangular boundary of the certificate in the picture is calculated;

   Extract the rectangular area formed by the rectangular border of the certificate in the picture, and detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the certificate in the received picture is a broken certificate.
2. The method for detecting the four corners of a credential incompleteness according to claim 1, wherein the step of calculating the rectangular boundary of the credential in the picture based on the area of the smallest bounding rectangle and the coordinate clustering center comprises:

   Identify the type of the certificate in the picture, determine the target standard certificate corresponding to the type according to the type of the certificate, and obtain the target standard area of the target circumscribed rectangle in the target standard certificate;

   Based on the proportional relationship between the target standard document and the area of the smallest circumscribed rectangle, the length and width of the certificate are calculated, wherein the proportional relationship includes the difference between the target standard length of the target standard certificate and the target standard area The first target proportional relationship between the two, and the second target proportional relationship between the target standard width of the target standard certificate and the target standard area;

   Based on the relative position relationship of the target standard certificate, the length and width of the certificate, and the coordinate clustering center, the rectangular boundary of the certificate is determined, wherein the relative position relationship includes the target standard of the target standard certificate The first target relative positional relationship between the length and the target cluster center of the target standard certificate, and the second target relative between the target standard width of the target standard certificate and the target cluster center of the target standard certificate Positional relationship.
3. The method for detecting the incompleteness of the four corners of a document according to claim 2, wherein the rectangle of the document is determined based on the relative position relationship of the target standard document, the length and width of the document, and the coordinate clustering center The boundary steps include:

   Determine the length boundary direction of the certificate in the picture based on the relative positional relationship between the coordinate clustering center and the first target;

   Determine the width boundary direction of the document in the picture based on the relative positional relationship between the coordinate clustering center and the second target;

   The length boundary of the document is determined based on the length of the document and the direction of the length boundary, and the width boundary of the document is determined based on the width of the document and the direction of the width boundary. The width boundary determines the rectangular boundary of the document.
4. The method for detecting the four corners of a credential incompleteness according to claim 1, wherein the step of extracting the rectangular area formed by the rectangular boundary of the credential in the picture afterwards comprises:

   Acquiring the length and width of the rectangular area, and calculating the pixel value corresponding to the length of the rectangular area, and whether any one of the pixel values corresponding to the width of the rectangular area is greater than a preset threshold;

   If any one of the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area is greater than the preset threshold, the rectangular area is divided according to the first preset segmentation method Is a plurality of sub-pictures, and the sub-pictures corresponding to the four corners of the rectangular area among the plurality of sub-pictures are used as target sub-pictures to detect whether there are defects in the four corners of the rectangular area based on each of the target sub-pictures. Horn;

   If the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area are both less than or equal to the preset threshold value, the rectangular area is divided into multiple segments according to the second preset segmentation method. And use multiple sub-pictures as target sub-pictures to detect whether there are missing corners in the four corners of the rectangular area based on each of the target sub-pictures.
5. The method for detecting the defect of four corners of a document according to claim 4, wherein the step of detecting whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, determining that the certificate in the received picture is a broken certificate include:

   Detect each of the target sub-pictures one by one through a preset method, and determine whether the target sub-pictures all include a complete ID corner;

   If the target sub-pictures all contain a complete ID corner, it is determined that the received picture is valid;

   If there is a missing corner in any one of the target sub-images in the target sub-picture, it is determined that the certificate in the received picture is a broken certificate, the received picture is determined to be invalid, and a re-collection prompt message is output.
6. The method for detecting the four corners of a credential defect according to claim 1, wherein when a picture containing a credential is received, the feature point coordinates of the face in the credential included in the picture are obtained, and the coordinate cluster is obtained based on the feature point coordinates. The steps of the class center and the smallest enclosing rectangle, and obtaining the area of the smallest enclosing rectangle include:

   When a picture containing a document is received, extract multiple feature points of the face in the document through a preset neural network, and obtain the feature point coordinates of the multiple feature points;

   Clustering the coordinate of each feature point of the face in the certificate by a preset algorithm, and obtain the center coordinate of the coordinate of each feature point of the face in the certificate as a coordinate clustering center;

   Determine the minimum circumscribed rectangle according to the position of each of the feature point coordinates of the face in the document, and detect the length and width of the minimum circumscribed rectangle, and calculate the minimum circumscribed rectangle by the length and width of the minimum circumscribed rectangle The area of the rectangle.
7. The method for detecting the incompleteness of four corners of a document according to any one of claims 1 to 6, wherein the rectangular boundary of the document in the picture is calculated based on the area of the smallest bounding rectangle and the coordinate clustering center The steps before include:

   Obtain the standard area, standard length, standard width, standard clustering center, and the minimum standard area of the feature circumscribed rectangle of the facial feature points in the standard certificate of each type of standard certificate;

   Perform the following steps one by one for each type of the standard certificate:

   Based on the standard length and the minimum standard area, generate a first proportional relationship between the standard length and the minimum standard area, and generate the standard width and the minimum standard width based on the standard width and the minimum standard area. The second proportional relationship of the standard area;

   Based on the standard length and the standard cluster center, a first relative positional relationship between the standard cluster center and the standard length is generated, and based on the standard width and the standard cluster center, the standard cluster is generated The second relative positional relationship between the class center and the standard width.
8. A four-corner defect detection device of a document, wherein the four-corner defect detection device of a document includes:

   The obtaining module is used to obtain, when a picture containing a certificate is received, the coordinate of the feature point of the face in the certificate contained in the picture, obtain the coordinate clustering center and the minimum bounding rectangle based on the coordinate of the feature point, and obtain the minimum The area of the circumscribing rectangle;

   A calculation module, configured to calculate the rectangular boundary of the certificate in the picture based on the area of the smallest bounding rectangle and the coordinate clustering center;

   The extraction module is used to extract the rectangular area formed by the rectangular boundary of the certificate in the picture, and to detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, determine whether there are missing corners in the received picture. The certificate is incomplete.
9. A four-corner defect detection device for a document, wherein the four-corner defect detection device includes a memory, a processor, and a four-corner defect detection program stored in the memory and running on the processor. When the detection program is executed by the processor, the following steps are implemented:

   When a picture containing a certificate is received, obtaining the feature point coordinates of the face in the certificate included in the picture, obtaining the coordinate cluster center and the smallest bounding rectangle based on the characteristic point coordinates, and obtaining the area of the smallest bounding rectangle;

   Based on the area of the smallest bounding rectangle and the coordinate clustering center, the rectangular boundary of the certificate in the picture is calculated;

   Extract the rectangular area formed by the rectangular border of the certificate in the picture, and detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the certificate in the received picture is a broken certificate.
10. The four-corner defect detection device of the document according to claim 9, wherein the step of calculating the rectangular boundary of the document in the picture based on the area of the smallest bounding rectangle and the coordinate clustering center comprises:

    Identify the type of the certificate in the picture, determine the target standard certificate corresponding to the type according to the type of the certificate, and obtain the target standard area of the target circumscribed rectangle in the target standard certificate;

    Based on the proportional relationship between the target standard document and the area of the smallest circumscribed rectangle, the length and width of the certificate are calculated, wherein the proportional relationship includes the difference between the target standard length of the target standard certificate and the target standard area The first target proportional relationship between the two, and the second target proportional relationship between the target standard width of the target standard certificate and the target standard area;

    Based on the relative position relationship of the target standard certificate, the length and width of the certificate, and the coordinate clustering center, the rectangular boundary of the certificate is determined, wherein the relative position relationship includes the target standard of the target standard certificate The first target relative positional relationship between the length and the target cluster center of the target standard certificate, and the second target relative between the target standard width of the target standard certificate and the target cluster center of the target standard certificate Positional relationship.
11. The four-corner defect detection device of the document according to claim 10, wherein the rectangle of the document is determined based on the relative position relationship of the target standard document, the length and width of the document, and the coordinate clustering center The boundary steps include:

    Determine the length boundary direction of the certificate in the picture based on the relative positional relationship between the coordinate clustering center and the first target;

    Determine the width boundary direction of the document in the picture based on the relative positional relationship between the coordinate clustering center and the second target;

    The length boundary of the document is determined based on the length of the document and the direction of the length boundary, and the width boundary of the document is determined based on the width of the document and the direction of the width boundary. The width boundary determines the rectangular boundary of the document.
12. The four-corner defect detection device of the document according to claim 9, wherein, after the step of extracting the rectangular area formed by the rectangular boundary of the certificate in the picture, the four-corner defect detection program of the document is executed by the processor It also implements the following steps:

    Acquiring the length and width of the rectangular area, and calculating the pixel value corresponding to the length of the rectangular area, and whether any one of the pixel values corresponding to the width of the rectangular area is greater than a preset threshold;

    If any one of the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area is greater than the preset threshold, the rectangular area is divided according to the first preset segmentation method Is a plurality of sub-pictures, and the sub-pictures corresponding to the four corners of the rectangular area among the plurality of sub-pictures are used as target sub-pictures to detect whether there are defects in the four corners of the rectangular area based on each of the target sub-pictures. Horn;

    If the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area are both less than or equal to the preset threshold value, the rectangular area is divided into multiple segments according to the second preset segmentation method. And use multiple sub-pictures as target sub-pictures to detect whether there are missing corners in the four corners of the rectangular area based on each of the target sub-pictures.
13. The four-corner defect detection device of the document according to claim 12, wherein the step of detecting whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, determining that the certificate in the received picture is a defective certificate include:

    Detect each of the target sub-pictures one by one through a preset method, and determine whether the target sub-pictures all include a complete ID corner;

    If the target sub-pictures all contain a complete ID corner, it is determined that the received picture is valid;

    If there is a missing corner in any one of the target sub-images in the target sub-picture, it is determined that the certificate in the received picture is a broken certificate, the received picture is determined to be invalid, and a re-collection prompt message is output.
14. The four-corner defect detection device of the document according to claim 9, wherein when the picture containing the document is received, the feature point coordinates of the face in the document contained in the picture are obtained, and the coordinate cluster is obtained based on the feature point coordinates. The steps of the class center and the smallest enclosing rectangle, and obtaining the area of the smallest enclosing rectangle include:

    When a picture containing a document is received, extract multiple feature points of the face in the document through a preset neural network, and obtain the feature point coordinates of the multiple feature points;

    Clustering the coordinate of each feature point of the face in the certificate by a preset algorithm, and obtain the center coordinate of the coordinate of each feature point of the face in the certificate as a coordinate clustering center;

    Determine the minimum circumscribed rectangle according to the position of each of the feature point coordinates of the face in the document, and detect the length and width of the minimum circumscribed rectangle, and calculate the minimum circumscribed rectangle by the length and width of the minimum circumscribed rectangle The area of the rectangle.
15. The four-corner defect detection device of the document according to any one of claims 9-14, wherein the rectangular boundary of the document in the picture is calculated based on the area of the smallest bounding rectangle and the coordinate clustering center Before the step, when the four corners of the credential defect detection program is executed by the processor, the following steps are also implemented:

    Obtain the standard area, standard length, standard width, standard clustering center, and the minimum standard area of the feature circumscribed rectangle of the facial feature points in the standard certificate of each type of standard certificate;

    Perform the following steps one by one for each type of the standard certificate:

    Based on the standard length and the minimum standard area, generate a first proportional relationship between the standard length and the minimum standard area, and generate the standard width and the minimum standard width based on the standard width and the minimum standard area. The second proportional relationship of the standard area;

    Based on the standard length and the standard cluster center, a first relative positional relationship between the standard cluster center and the standard length is generated, and based on the standard width and the standard cluster center, the standard cluster is generated The second relative positional relationship between the class center and the standard width.
16. A storage medium, wherein a four-corner defect detection program of a document is stored on the storage medium, and the following steps are implemented when the four-corner defect detection program is executed by a processor:

    When a picture containing a certificate is received, obtaining the feature point coordinates of the face in the certificate included in the picture, obtaining the coordinate cluster center and the smallest bounding rectangle based on the characteristic point coordinates, and obtaining the area of the smallest bounding rectangle;

    Based on the area of the smallest bounding rectangle and the coordinate clustering center, the rectangular boundary of the certificate in the picture is calculated;

    Extract the rectangular area formed by the rectangular border of the certificate in the picture, and detect whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, it is determined that the certificate in the received picture is a broken certificate.
17. 15. The storage medium according to claim 16, wherein the step of calculating the rectangular boundary of the certificate in the picture based on the area of the smallest bounding rectangle and the coordinate clustering center comprises:

    Identify the type of the certificate in the picture, determine the target standard certificate corresponding to the type according to the type of the certificate, and obtain the target standard area of the target circumscribed rectangle in the target standard certificate;

    Based on the proportional relationship between the target standard document and the area of the smallest circumscribed rectangle, the length and width of the certificate are calculated, wherein the proportional relationship includes the difference between the target standard length of the target standard certificate and the target standard area The first target proportional relationship between the two, and the second target proportional relationship between the target standard width of the target standard certificate and the target standard area;

    Based on the relative position relationship of the target standard certificate, the length and width of the certificate, and the coordinate clustering center, the rectangular boundary of the certificate is determined, wherein the relative position relationship includes the target standard of the target standard certificate The first target relative positional relationship between the length and the target cluster center of the target standard certificate, and the second target relative between the target standard width of the target standard certificate and the target cluster center of the target standard certificate Positional relationship.
18. The storage medium according to claim 17, wherein the step of determining the rectangular boundary of the document based on the relative positional relationship of the target standard document, the length and width of the document, and the coordinate clustering center include:

    Determine the length boundary direction of the certificate in the picture based on the relative positional relationship between the coordinate clustering center and the first target;

    Determine the width boundary direction of the document in the picture based on the relative positional relationship between the coordinate clustering center and the second target;

    The length boundary of the document is determined based on the length of the document and the direction of the length boundary, and the width boundary of the document is determined based on the width of the document and the direction of the width boundary. The width boundary determines the rectangular boundary of the document.
19. The storage medium according to claim 16, wherein, after the step of extracting the rectangular area formed by the rectangular boundary of the certificate in the picture, the program for detecting the four corners of the certificate is executed by the processor further implements the following steps :

    Acquiring the length and width of the rectangular area, and calculating the pixel value corresponding to the length of the rectangular area, and whether any one of the pixel values corresponding to the width of the rectangular area is greater than a preset threshold;

    If any one of the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area is greater than the preset threshold, the rectangular area is divided according to the first preset segmentation method Is a plurality of sub-pictures, and the sub-pictures corresponding to the four corners of the rectangular area among the plurality of sub-pictures are used as target sub-pictures to detect whether there are defects in the four corners of the rectangular area based on each of the target sub-pictures. Horn;

    If the pixel value corresponding to the length of the rectangular area and the pixel corresponding to the width of the rectangular area are both less than or equal to the preset threshold value, the rectangular area is divided into multiple segments according to the second preset segmentation method. And use multiple sub-pictures as target sub-pictures to detect whether there are missing corners in the four corners of the rectangular area based on each of the target sub-pictures.
20. The storage medium according to claim 19, wherein the step of detecting whether there are missing corners in the four corners of the rectangular area, and if there are missing corners, determining that the certificate in the received picture is a broken certificate comprises:

    Detect each of the target sub-pictures one by one through a preset method, and determine whether the target sub-pictures all include a complete ID corner;

    If the target sub-pictures all contain a complete ID corner, it is determined that the received picture is valid;

    If there is a missing corner in any one of the target sub-images in the target sub-picture, it is determined that the certificate in the received picture is a broken certificate, the received picture is determined to be invalid, and a re-collection prompt message is output.

Images