Seal retrieval method, system, electronic device and storage medium

Abstract

The embodiment of the present invention provides a seal retrieval method, system, electronic device and storage medium, which belongs to the field of seal detection. The method includes: obtaining a seal image to be retrieved, and inputting it into a pre-built target detection model for detection; if there are seals, texts and special marks in the seal image to be retrieved, the seal area is cropped according to the corresponding coordinates to obtain a target retrieval area; extracting the corresponding feature vector and / or the feature vector corresponding to the special mark, and performing vector retrieval in the feature library to determine whether there is a matching seal image to obtain an initial matching seal; recognizing the text in the target retrieval area, and comparing the recognition result with the text in the initial matching seal, if they are consistent, it means that the match is successful. By acquiring the features of the seal from multiple dimensions, a more fine-grained seal feature extraction is achieved, thereby improving the accuracy of seal matching and avoiding false detection in the seal retrieval process.

Description

Technical Field

[0001] The present invention relates to the technical field of seal detection, and in particular to a seal retrieval method, system, electronic equipment and storage medium. Background Art

[0002] Seals are widely used as credentials in modern society's foreign exchanges, and the need for seal retrieval has also emerged. Judging whether a target seal exists in a picture in massive data is inefficient if it relies solely on manpower, so computer vision technology is often used to assist in seal recognition.

[0003] The seal retrieval method currently used generally pre-processes the seal image first, then obtains the seal area through target detection algorithms or some traditional image processing methods, and then calculates the similarity between the features of the seal area and the target seal features, and determines whether the two are consistent through the similarity, thereby completing the seal retrieval work.

[0004] The seal retrieval method currently used often only searches at the image feature level, that is, it only determines whether the image to be identified has image similarity with the target seal image. Seals of the same category often have relatively similar layouts, such as official seals, but they do not actually belong to the same seal. In addition, due to the different weights of the seals and the image loss caused by photographing and scanning, there will be a certain deviation from the original seal, so false detection will occur. Therefore, a seal similarity judgment method at a more fine-grained level is needed. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a seal retrieval method, system, electronic device and storage medium, which are used to completely or at least partially solve the technical problems existing in the above-mentioned prior art.

[0006] In order to achieve the above object, an embodiment of the present invention provides a seal retrieval method, comprising:

[0007] Acquire a seal image to be retrieved, and input the seal image to be retrieved into a pre-built target detection model for detection;

[0008] If there are seals, texts and special marks in the seal image to be searched, the seal area in the seal image to be searched is cropped according to the coordinates corresponding to the detected seals, texts and special marks to obtain the target search area;

[0009] Extracting the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, and performing vector search in the feature library based on the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, determining whether there is a matching seal image in the feature library, and obtaining an initial matching seal;

[0010] The text in the target search area is recognized, and the recognition result is compared with the text in the initial matching seal. If they are consistent, it means the match is successful.

[0011] Optionally, the process of building the target detection model includes:

[0012] Collect different types of seal images and mark them, where the seals and special marks are marked with a four-point frame to mark the coordinates, the non-curved normal text area is marked with a four-point frame to mark the coordinates, and the curved text area is marked with a sixteen-point frame;

[0013] Based on the labeled seal images of different types, a three-category target detection model is trained by using a file detection algorithm to obtain a target detection model, wherein the three-category target detection model categories are seals, texts and special marks.

[0014] Optionally, the process of marking the coordinates of a seal and a special mark with a four-point frame, marking the coordinates of a non-curved normal-shaped text area with a four-point frame, and marking the coordinates of a curved text area with a sixteen-point frame includes:

[0015] For seals, special marks and non-curved text areas, use a four-point box [x1, y1, x2, y2, x3, y3, x4, y4] to mark the coordinates, where x1 and y1 are the upper left corner coordinates of the four-point box, x2 and y2 are the upper right corner coordinates of the four-point box, x3 and y3 are the lower left corner coordinates of the four-point box, and x4 and y4 are the lower right corner coordinates of the four-point box;

[0016] The coordinates of the curved text area are marked using a sixteen-point box [x1, y1, x2, y2 ... x16, y16], wherein these sixteen points depict the text area along the shape of the curved text to form an enclosing curve.

[0017] Optionally, extracting the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, and performing vector search in a feature library based on the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, determining whether there is a matching seal image in the feature library, and obtaining an initial matching seal, includes:

[0018] The convolutional neural network model constructed using the PPLCNet algorithm is used to extract preliminary features corresponding to the search area and / or special identification, and the feature dimension is transformed through the fully connected layer of the convolutional neural network model to obtain a feature vector of fixed dimension;

[0019] The obtained fixed-dimensional feature vector is used to calculate the similarity with the feature vectors of multiple seals to be matched in the feature library, and the seal image with the highest similarity and a similarity greater than a threshold is retrieved as the initial matching seal.

[0020] Optionally, similarity is calculated according to the following formula:

[0021] ;

[0022] In the formula, represents the cosine function, They represent the feature vector of fixed dimension and the feature vector of the seal to be matched in the feature library respectively. They respectively represent the feature vector of fixed dimension and the modulus of the feature vector of the seal to be matched in the feature library.

[0023] Optionally, the text in the target search area is recognized, and the recognition result is compared with the text in the initial matching seal. If they are consistent, it means that the match is successful, including:

[0024] Sort the text areas in the target search area, sort them from left to right and from top to bottom according to the center point coordinates of the text coordinates, and stretch them into a rectangular image. If a special mark is detected, sort the coordinates of the special mark area together;

[0025] Identify the corresponding text in the stretched rectangular image to obtain the text recognition result of the seal area;

[0026] In the arranged order, the text recognition results and special markings in the seal area are compared with the text and special markings in the initially matched seal. If the two are consistent, it means they belong to the same seal and the match is successful, otherwise the match is unsuccessful.

[0027] Optionally, the seal data in the feature library includes a picture of the seal, a picture of a special mark, a feature vector and a text recognition result.

[0028] On the other hand, the present invention also provides a seal retrieval system, comprising:

[0029] An acquisition unit, used for acquiring a seal image to be retrieved, and inputting the seal image to be retrieved into a pre-built target detection model for detection;

[0030] A cropping unit, for cropping the seal area in the seal image to be searched according to the coordinates corresponding to the detected seal, text and special mark, to obtain a target search area if there are seals, text and special marks in the seal image to be searched;

[0031] An extraction unit is used to extract the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, and perform vector search in the feature library based on the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, determine whether there is a matching seal image in the feature library, and obtain an initial matching seal;

[0032] The matching unit is used to recognize the text in the target search area and compare the recognition result with the text in the initial matching seal. If they are consistent, it means that the match is successful.

[0033] On the other hand, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the seal retrieval method described above when executing the program.

[0034] On the other hand, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, and the computer program, when executed by a processor, implements the steps of the seal retrieval method described above.

[0035] Through the above technical solution, the features of the seal are obtained from multiple dimensions, and more fine-grained seal feature extraction is achieved, thereby improving the accuracy of seal matching and effectively avoiding false detection in the seal retrieval process.

[0036] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0038] Figure 1 It is an implementation flow chart of a seal retrieval method provided by an embodiment of the present invention;

[0039] Figure 2 It is a structural schematic diagram of a seal retrieval system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0041] See also Figure 1As shown, it is a flowchart of an implementation of a seal retrieval method provided by an embodiment of the present invention, comprising the following execution steps:

[0042] Step 100: Obtain a seal image to be retrieved, and input the seal image to be retrieved into a pre-built target detection model for detection.

[0043] In some embodiments, the process of constructing the target detection model includes the following steps:

[0044] S1: Collect different types of seal images and mark them. The seals and special marks are marked with a four-point frame to mark their coordinates, the non-curved normal text area is marked with a four-point frame to mark its coordinates, and the curved text area is marked with a sixteen-point frame.

[0045] Specifically, when executing step S1, the following steps can be specifically performed: for seals, special marks and non-curved text areas, use a four-point box [x1, y1, x2, y2, x3, y3, x4, y4] to mark the coordinates, wherein x1 and y1 are the upper left corner coordinates of the four-point box, x2 and y2 are the upper right corner coordinates of the four-point box, x3 and y3 are the lower left corner coordinates of the four-point box, and x4 and y4 are the lower right corner coordinates of the four-point box; for curved text areas, use a sixteen-point box [x1, y1, x2, y2…x16, y16] to mark the coordinates, wherein these sixteen points depict the text area along the shape of the curved text to form an enclosing curve.

[0046] In some embodiments, the sixteen-point box [x1, y1, x2, y2...x16, y16] will depict the area of ​​the text along the shape of the curved text when marking the coordinates, that is, in reading order, marking from the upper left corner of the text -> marking 8 points to the upper right corner of the text -> to the lower right corner of the text -> 8 points to the lower left corner of the text, a total of 16 points, forming an enclosing curve.

[0047] S2: Based on the labeled seal images of different types, a three-category target detection model is trained using a file detection algorithm to obtain a target detection model, wherein the three-category target detection model categories are seals, texts, and special marks.

[0048] Among them, the seal can be understood as the entire seal area, the text can be understood as all the text in the seal, the special mark can be understood as some non-text patterns such as asterisks and logos in the seal, and different types of seal images such as circles, triangles, rectangles, etc. are not limited here.

[0049] Preferably, the object detection model can be constructed using the DB++ text detection algorithm.

[0050] Step 101: If there are seals, texts and special marks in the seal image to be searched, the seal area in the seal image to be searched is cropped according to the coordinates corresponding to the detected seals, texts and special marks to obtain a target search area.

[0051] Specifically, the target detection model detects whether there are stamps, texts, or special marks in the image. If there is no stamp target, the process ends. Otherwise, the stamp area in the image is cropped according to the detected coordinates. If multiple stamps are detected, each stamp needs to be compared with the stamps in the feature library to see if they are consistent. Therefore, each stamp is processed in the following steps.

[0052] Step 102: Extract the feature vector corresponding to the target retrieval area and / or the feature vector corresponding to the special mark, and perform vector search in the feature library based on the feature vector corresponding to the target retrieval area and / or the feature vector corresponding to the special mark, determine whether there is a matching seal image in the feature library, and obtain an initial matching seal.

[0053] It should be noted that the seal data in the feature library includes seal images, special logo images, feature vectors and text recognition results.

[0054] Specifically, when executing step 102, the following steps may be specifically performed:

[0055] S1020: Using the convolutional neural network model constructed by the PPLCNet algorithm, extract preliminary features corresponding to the retrieval area and / or special identification, and transform the feature dimension through the fully connected layer of the convolutional neural network model to obtain a feature vector of fixed dimension.

[0056] S1021: Calculate the similarity between the obtained fixed-dimensional feature vector and the feature vectors of multiple seals to be matched in the feature library, and retrieve the seal image with the highest similarity and a similarity greater than a threshold as the initial matching seal.

[0057] Specifically, the similarity is calculated according to the following formula:

[0058] ;

[0059] In the formula, represents the cosine function, They represent the feature vector of fixed dimension and the feature vector of the seal to be matched in the feature library respectively. They respectively represent the feature vector of fixed dimension and the modulus of the feature vector of the seal to be matched in the feature library.

[0060] In some embodiments, for the cropped seal area, a feature extraction model is used to calculate the feature vector of the seal area (a vector with a fixed dimension). If a special mark is detected on the seal, the feature vector of the special mark is also calculated. The HNSW algorithm is used in the feature library to perform vector retrieval of the seal area features (the HNSW algorithm is a graph structure algorithm for efficient vector retrieval, which aims to solve the similarity search problem in large-scale data sets), that is, similarity calculation is performed with the feature vectors of multiple seals to be matched in the feature library, and the seal image with a similarity greater than a threshold and the highest similarity is retrieved as the matching result. If such a seal image does not exist in the feature library, the process ends. In addition, if there is no special mark in the seal area and the matched seal image, proceed directly to the next step. If only one of the two images has a special mark, it means that there is no match, and the process ends directly. If both have special marks, the similarity of the feature vectors of the special marks of the two is calculated. If it is greater than the threshold, it means that the match is successful. When calculating the similarity, the example uses the cosine of the cos angle between the two vectors as the similarity value. Assuming that the two 256-dimensional vectors used in the example are marked respectively, the similarity value calculation formula is: . ‌Among them, the feature extraction model is a pre-trained convolutional neural network model. The example here uses the PPLCNet algorithm. The feature extraction process is to extract the preliminary features of the input image through the backbone network, and then transform the feature dimension through the fully connected layer, so as to convert the input image into a feature vector of fixed dimension. The training process of the feature extraction model is to collect pictures of various different categories, including different types of goods, vehicles, food, and different types of seal pictures, and train a classification model of tens of thousands of categories, so that pictures with high similarity are close to each other in the feature space, and pictures with low characteristics are far away from each other in the feature space, so that pictures can be distinguished by the extracted feature vectors.

[0061] Step 103: Recognize the text in the target search area, and compare the recognized text results with the text in the initial matching seal. If they are consistent, it means that the match is successful.

[0062] Specifically, when executing step 103, the following steps may be specifically performed:

[0063] S1030: Sort the text areas in the target search area from left to right and from top to bottom according to the center point coordinates of the text coordinates, and stretch them into a rectangular image. If a special mark is detected, the coordinates of the special mark area are sorted together.

[0064] S1031: Identify the corresponding text in the stretched rectangular image to obtain the text recognition result of the seal area.

[0065] Specifically, a text recognition model constructed by a CRNN text recognition algorithm may be used to recognize corresponding text in the stretched rectangular image to obtain a text recognition result of the seal area.

[0066] S1032: Compare the text recognition results and special markings in the seal area with the text and special markings in the initially matched seal in the sorted order. If the two are consistent, it means that they belong to the same seal and the match is successful. Otherwise, the match is unsuccessful.

[0067] This application combines the target detection algorithm, feature extraction algorithm and OCR recognition algorithm to obtain the features of the seal from multiple dimensions, achieve more fine-grained seal feature extraction, thereby improving the accuracy of seal matching and effectively avoiding false detection in the seal retrieval process.

[0068] See also Figure 2 FIG. 1 is a schematic diagram of a seal retrieval system provided by an embodiment of the present invention, comprising:

[0069] An acquisition unit 20 is used to acquire a seal image to be retrieved, and input the seal image to be retrieved into a pre-built target detection model for detection;

[0070] A cutting unit 21 is used for cutting the seal area in the seal image to be searched according to the coordinates corresponding to the detected seal, text and special mark to obtain a target search area if there are seals, text and special marks in the seal image to be searched;

[0071] An extraction unit 22 is used to extract the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, and perform vector search in the feature library based on the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, determine whether there is a matching seal image in the feature library, and obtain an initial matching seal;

[0072] The matching unit 23 is used to recognize the characters in the target search area and compare the recognition results with the characters in the initial matching seal. If they are consistent, it means that the match is successful.

[0073] On the other hand, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the seal retrieval method described in any one of the above embodiments when executing the program.

[0074] On the other hand, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the seal retrieval method described in any one of the above embodiments are implemented.

[0075] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0076] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0077] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0078] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0079] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0080] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0081] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0082] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0083] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A seal retrieval method, characterized in that: include: Acquire a seal image to be retrieved, and input the seal image to be retrieved into a pre-built target detection model for detection; If there are seals, texts and special marks in the seal image to be searched, the seal area in the seal image to be searched is cropped according to the coordinates corresponding to the detected seals, texts and special marks to obtain the target search area; Extracting the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, and performing vector search in the feature library based on the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, determining whether there is a matching seal image in the feature library, and obtaining an initial matching seal; Recognize the text in the target search area, and compare the recognition result with the text in the initial matching seal. If they are consistent, it means the match is successful; The target detection model construction process includes: Collect different types of seal images and mark them, where the seals and special marks are marked with a four-point frame to mark the coordinates, the non-curved normal text area is marked with a four-point frame to mark the coordinates, and the curved text area is marked with a sixteen-point frame; Based on the labeled seal images of different types, a three-category target detection model is trained using a file detection algorithm to obtain a target detection model, wherein the three-category target detection model categories are seals, texts, and special marks; The process of marking the coordinates of seals and special marks with a four-point frame, marking the coordinates of non-curved normal text areas with a four-point frame, and marking the coordinates of curved text areas with a sixteen-point frame includes: For seals, special marks and non-curved text areas, use a four-point box [x1, y1, x2, y2, x3, y3, x4, y4] to mark the coordinates, where x1 and y1 are the upper left corner coordinates of the four-point box, x2 and y2 are the upper right corner coordinates of the four-point box, x3 and y3 are the lower left corner coordinates of the four-point box, and x4 and y4 are the lower right corner coordinates of the four-point box; The coordinates of the curved text area are marked using a sixteen-point box [x1, y1, x2, y2…x16, y16], where these sixteen points depict the text area along the shape of the curved text to form an enclosing curve; Among them, the text in the target search area is recognized, and the recognition result is compared with the text in the initial matching seal. If they are consistent, it means that the match is successful, including: Sort the text areas in the target search area, sort them from left to right and from top to bottom according to the center point coordinates of the text coordinates, and stretch them into a rectangular image; if a special mark is detected, sort the coordinates of the special mark area together; Identify the corresponding text in the stretched rectangular image to obtain the text recognition result of the seal area; In the arranged order, the text recognition results and special markings in the seal area are compared with the text and special markings in the initially matched seal. If the two are consistent, it means they belong to the same seal and the match is successful, otherwise the match is unsuccessful.

2. The seal retrieval method according to claim 1, characterized in that: Extracting the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, and performing vector search in the feature library based on the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, determining whether there is a matching seal image in the feature library, and obtaining an initial matching seal, including: The convolutional neural network model constructed using the PPLCNet algorithm is used to extract preliminary features corresponding to the search area and / or special identification, and the feature dimension is transformed through the fully connected layer of the convolutional neural network model to obtain a feature vector of fixed dimension; The obtained fixed-dimensional feature vector is used to calculate the similarity with the feature vectors of multiple seals to be matched in the feature library, and the seal image with the highest similarity and a similarity greater than a threshold is retrieved as the initial matching seal.

3. The seal retrieval method according to claim 2, characterized in that: The similarity is calculated according to the following formula: ； In the formula, represents the cosine function, They represent the feature vector of fixed dimension and the feature vector of the seal to be matched in the feature library respectively. They respectively represent the feature vector of fixed dimension and the modulus of the feature vector of the seal to be matched in the feature library.

4. The seal retrieval method according to claim 3, characterized in that: The seal data in the feature library includes a picture of the seal, a picture of a special mark, a feature vector and a text recognition result.

5. A seal retrieval system applied to the seal retrieval method according to any one of claims 1 to 4, characterized in that: include: An acquisition unit, used for acquiring a seal image to be retrieved, and inputting the seal image to be retrieved into a pre-built target detection model for detection; A cropping unit, for cropping the seal area in the seal image to be searched according to the coordinates corresponding to the detected seal, text and special mark, to obtain a target search area if there are seals, text and special marks in the seal image to be searched; An extraction unit is used to extract the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, and perform vector search in the feature library based on the feature vector corresponding to the target search area and / or the feature vector corresponding to the special mark, determine whether there is a matching seal image in the feature library, and obtain an initial matching seal; The matching unit is used to recognize the text in the target search area and compare the recognition result with the text in the initial matching seal. If they are consistent, it means that the match is successful.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the seal retrieval method according to any one of claims 1 to 4 are implemented.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the seal retrieval method according to any one of claims 1 to 4 are implemented.

Images