Method for identifying authenticity of scanned seal based on multi-dimensional deep convolution kernel detection

Abstract

The invention discloses a scanning copy seal authenticity identification method based on multi-dimensional deep convolution kernel detection, and relates to the technical field of seal authenticity identification. The method comprises the following steps: 1, scanning data submitted by a project department through a scanner, and generating a contract scanning electronic file; 2, calling an API (Application Program Interface) for execution, and enhancing the definition through Rea l-ESRGAN; 3, OCR character recognition, wherein the OCR technology can directly recognize paper documents, texts of photos, scene photos or some texts suitable for being obtained from some visual devices. According to the method, the classification function of the kmeans algorithm is combined, real stamps and forged stamps can be quickly distinguished, particularly in the building industry, the authenticity of the stamps is crucial for the legal efficacy of contracts, and therefore the risk of forged stamps can be remarkably reduced through introduction of the method, and the enterprise information of the building industry is crawled, so that the efficiency of the stamps is improved. Information completion can be carried out on the extracted data, and the accuracy and integrity of the recognition result are further ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of seal authenticity recognition, and in particular to a seal authenticity recognition method based on multi-dimensional deep convolution kernel detection scanned copy. Background Art

[0002] Real-ESRGAN [Training Real-World Bl ind Super-Resolution with PureSynthetic Data] uses pure synthetic data for real-world blind super-resolution training, which is used to improve the quality of images and videos. Optical Character Recognition (OCR) uses a variety of algorithms and technologies for text recognition. Natural Language Processing (NLP) is an important direction in the fields of computer science and artificial intelligence. It studies various theories and methods that can achieve effective communication between humans and computers using natural language. The k-means clustering algorithm is an iterative clustering analysis algorithm. Its steps are to pre-divide the data into K groups, randomly select K objects as the initial cluster centers, and then calculate the distance between each object and each seed cluster center, and assign each object to the cluster center closest to it. The cluster centers and the objects assigned to them represent a cluster.

[0003] The construction industry has a large number of paper seal documents due to on-site convenience and network security. Seal authenticity identification helps companies avoid risks

[0004] Therefore, a method for detecting the authenticity of scanned seals based on multi-dimensional deep convolution kernel detection is proposed. Summary of the invention

[0005] The purpose of the present invention is to solve the problems raised in the above-mentioned background technology, and provide a method for identifying the authenticity of a scanned seal based on multi-dimensional deep convolution kernel detection.

[0006] In order to achieve the above-mentioned purpose, the present invention specifically adopts the following technical solutions:

[0007] A method for identifying the authenticity of a seal scanned copy based on multi-dimensional deep convolution kernel detection comprises the following steps:

[0008] Step 1: Scan the information submitted by the project department with a scanner to generate an electronic scanned file of the contract;

[0009] Step 2: Call API to execute and enhance the clarity through Real-ESRGAN;

[0010] Step 3: OCR text recognition. OCR technology can directly recognize text from paper documents, photos, scene photos, or text obtained from some suitable visual devices;

[0011] Step 4: NLP natural language processing, using NLP content extraction to enable computers to understand, process and generate natural language text or speech;

[0012] Step 5: Crawling the enterprise information of the construction industry and completing the extracted data;

[0013] Step 6: kmeans algorithm classification, grouping the data according to similarity, making the data in the same group as similar as possible, and the data in different groups as different as possible;

[0014] Step 7: Categorize and score the data;

[0015] Step 8: Output the report that can be published.

[0016] Furthermore, the Real-ESRGAN clarity enhancement mainly includes the following steps:

[0017] Predicting high-level distribution and low-level details, Real-ESRGAN first predicts the high-level distribution of the image, and then enhances the details by estimating the kernel function related to the low-level details and adding noise;

[0018] Using Generative Adversarial Network (GAN), Real-ESRGAN uses GAN to upgrade the original low-resolution images. The Generator is responsible for generating high-resolution images, and the Discriminator is responsible for determining whether an image is generated by the generator or a real high-resolution image.

[0019] Redesigning the noise injection module and enhanced downsampling layer: The paper proposes two new architectural improvements, including decoupling the noise injection module and enhanced downsampling layer, so that low-level details can be effectively captured and have a positive effect on super-resolution results;

[0020] New loss function: Real-ESRGAN proposes an improved perceptual oss and Related Average GAN adversarial loss, which enables the model to achieve better results;

[0021] Optimization for practical application scenarios, unlike many traditional SR methods, Real-ESRGAN pays more attention to the super-encoding rate of real-world images, including images with complex textures and different noise levels.

[0022] Furthermore, the OCR text recognition mainly includes the following steps:

[0023] Preprocessing usually includes background noise removal, grayscale processing, binarization, and character segmentation. Background noise removal is to retain only text information as much as possible. Grayscale and binarization are to convert color or grayscale images into black and white images that are easier to process. Character segmentation is for subsequent recognition of individual characters.

[0024] Feature extraction, obtaining effective feature values ​​is the key to improving recognition rate. The purpose of this step is to find the features that distinguish each character, such as character edges, character stroke intersections, etc.

[0025] Character recognition, after completing feature extraction, the last step is character recognition, this step usually uses some machine learning methods, such as support vector machine (SVM), neural network, etc.

[0026] Furthermore, the NLP natural language processing mainly includes the following steps:

[0027] Semantic analysis, which involves understanding the meaning of words in a specific context, usually by extracting roots and suffixes, or mapping words to predetermined concepts;

[0028] Part-of-speech tagging, the process of assigning the correct phonetic label to each word in a text;

[0029] Syntactic analysis, which involves determining the syntactic structure of a sentence and generating a tree or graph that represents the sentence's syntactic structure;

[0030] Semantic comprehension, which requires understanding word meanings and the output of syntactic analysis to understand the exact meaning of the sentence;

[0031] Context recognition and ambiguity resolution: Computers need to understand the specific meaning of words in context, which is very difficult because many words have different meanings in different contexts.

[0032] Sentiment analysis, which uses natural language processing to determine the emotional tone of text and whether a review is positive or negative;

[0033] Sentence generation, constructing new sentences through natural language processing, generating responses or descriptions related to the input.

[0034] Furthermore, the kmeans algorithm classification mainly includes the following steps:

[0035] Initialization: randomly select K points as the initial cluster centers. There are also other optimized initialization methods, such as K-means++.

[0036] Clustering: For each sample, calculate its distance to each cluster center and classify it into the corresponding category based on the principle of the closest distance.

[0037] Update: For each category, calculate the mean of all samples in the class and update the cluster center of the category to the newly obtained mean.

[0038] Iterate and repeat the clustering and updating until the cluster center can no longer be updated, or the update amplitude is less than a given threshold, or the preset maximum number of iterations is reached.

[0039] Furthermore, Scrapy is used when crawling the construction industry enterprise information. Scrapy is an open source web data crawling framework written in Python, which provides users with a complete set of solutions for web crawling, data extraction, and data storage.

[0040] Furthermore, the Scrapy crawler framework includes a Scrapy engine, a Scheduler, a Downloader, Spiders, an Item Pipeline project pipeline, Downloader Middlewares and Spider middleware.

[0041] The beneficial effects of the present invention are as follows:

[0042] Upload the file image through scanning equipment or computer, connect to the system API for identification and return the result, combine with specific construction industry data for data proofreading, and use the clarity enhancement algorithm and image classification algorithm to quickly identify the newly added forged seals;

[0043] Through automated processes, including scanning, clarity enhancement, OCR text recognition, NLP processing and classification scoring, the entire recognition process can be greatly accelerated, thereby improving the efficiency of audit work. Using the Real-ESRGAN clarity enhancement algorithm, the image quality of scanned documents can be optimized and the accuracy of OCR and seal recognition can be improved. In addition, NLP processing can further refine text information and enhance the accuracy and completeness of data. Combined with the classification function of the kmeans algorithm, real seals can be quickly distinguished from forged seals. Especially in the construction industry, the authenticity of the seal is crucial to the legal effect of the contract. Therefore, the introduction of this method can significantly reduce the risk of forged seals. By crawling the construction industry enterprise information, the extracted data can be supplemented to further ensure the accuracy and completeness of the recognition results. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a flow chart of the present invention when used;

[0045] Figure 2 is a flow chart of the API identification execution system of the present invention;

[0046] Figure 3 is a flowchart of Real-ESRGAN clarity enhancement of the present invention;

[0047] Figure 4 is a flow chart of OCR text recognition of the present invention;

[0048] Figure 5 It is a flow chart of NLP natural language processing of the present invention;

[0049] Figure 6 It is a flow chart of the kmeans algorithm classification of the present invention;

[0050] Figure 7 It is a schematic diagram of the report format of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0052] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. In addition, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0054] In the description of the embodiments of the present invention, it should be noted that the terms "inside", "outside", "upper", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, or are the orientations or positional relationships in which the inventive product is usually placed when used. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0055] like Figures 1 to 7 As shown, a method for identifying the authenticity of a seal scanned copy based on multi-dimensional deep convolution kernel detection includes the following steps:

[0056] Step 1: Scan the information submitted by the project department with a scanner to generate an electronic scanned file of the contract;

[0057] Step 2: Call API to execute and enhance the clarity through Real-ESRGAN;

[0058] Step 3: OCR text recognition. OCR technology can directly recognize text from paper documents, photos, scene photos, or text obtained from some suitable visual devices;

[0059] Step 4: NLP natural language processing, using NLP content extraction to enable computers to understand, process and generate natural language text or speech;

[0060] Step 5: Crawling the enterprise information of the construction industry and completing the extracted data;

[0061] Step 6: kmeans algorithm classification, grouping the data according to similarity, making the data in the same group as similar as possible, and the data in different groups as different as possible;

[0062] Step 7: Categorize and score the data;

[0063] Step 8: Output the report that can be published.

[0064] Real-ESRGAN clarity enhancement mainly includes the following steps:

[0065] Predicting high-level distribution and low-level details, Real-ESRGAN first predicts the high-level distribution of the image, and then enhances the details by estimating the kernel function related to the low-level details and adding noise;

[0066] Using Generative Adversarial Network (GAN), Real-ESRGAN uses GAN to upgrade the original low-resolution images. The Generator is responsible for generating high-resolution images, and the Discriminator is responsible for determining whether an image is generated by the generator or a real high-resolution image.

[0067] Redesigning the noise injection module and enhanced downsampling layer: The paper proposes two new architectural improvements, including decoupling the noise injection module and enhanced downsampling layer, so that low-level details can be effectively captured and have a positive effect on super-resolution results;

[0068] New loss function: Real-ESRGAN proposes an improved perceptual oss and Related Average GAN adversarial loss, which enables the model to achieve better results;

[0069] Optimization for practical application scenarios, unlike many traditional SR methods, Real-ESRGAN pays more attention to the super-encoding rate of real-world images, including images with complex textures and different noise levels.

[0070] OCR text recognition mainly includes the following steps:

[0071] Preprocessing: usually includes steps such as background noise removal, grayscale processing, binarization, and character segmentation. Background noise removal is to retain only text information as much as possible. Grayscale and binarization are to convert color or grayscale images into black and white images that are easier to process. Character segmentation is for subsequent recognition of individual characters.

[0072] Feature extraction: Obtaining effective feature values ​​is the key to improving recognition rate. The purpose of this step is to find the features that distinguish each character, such as character edges, character stroke intersections, etc.

[0073] Character recognition: After completing feature extraction, the last step is character recognition, which usually uses some machine learning methods, such as support vector machine (SVM), neural network, etc.

[0074] NLP natural language processing mainly includes the following steps:

[0075] Semantic analysis, which involves understanding the meaning of words in a specific context, usually by extracting roots and suffixes, or mapping words to predetermined concepts;

[0076] Part-of-speech tagging, the process of assigning the correct phonetic label to each word in a text;

[0077] Syntactic analysis, which involves determining the syntactic structure of a sentence and generating a tree or graph that represents the sentence's syntactic structure;

[0078] Semantic comprehension, which requires understanding word meanings and the output of syntactic analysis to understand the exact meaning of the sentence;

[0079] Context recognition and ambiguity resolution: Computers need to understand the specific meaning of words in context, which is very difficult because many words have different meanings in different contexts.

[0080] Sentiment analysis, which uses natural language processing to determine the emotional tone of text and whether a review is positive or negative;

[0081] Sentence generation, constructing new sentences through natural language processing, generating responses or descriptions related to the input.

[0082] The kmeans algorithm classification mainly includes the following steps:

[0083] Initialization: randomly select K points as the initial cluster centers. There are also other optimized initialization methods, such as K-means++.

[0084] Clustering: For each sample, calculate its distance to each cluster center and classify it into the corresponding category based on the principle of the closest distance.

[0085] Update: For each category, calculate the mean of all samples in the class and update the cluster center of the category to the newly obtained mean.

[0086] Iterate and repeat the clustering and updating until the cluster center can no longer be updated, or the update amplitude is less than a given threshold, or the preset maximum number of iterations is reached.

[0087] Scrapy is used when crawling corporate information in the construction industry. Scrapy is an open source web data crawling framework written in Python. It provides users with a complete set of solutions for web crawling, data extraction, and data storage.

[0088] The Scrapy crawler framework includes the Scrapy engine, Scheduler, Downloader, Spiders crawler, Item Pipeline project pipeline, Downloader Middlewares download middleware and Spider middleware. More specifically, the Scrapy engine is responsible for controlling the flow of data in all components of the system and triggering events when corresponding actions occur. The Scheduler is responsible for accepting requests from the engine and adding them to the queue, providing them to the engine when the engine requests again. The Downloader is responsible for obtaining page data and returning it to the engine for subsequent processing. The Spiders define how to crawl a website, including crawling actions and how to extract structured data from web page content. The Item Pipeline is responsible for processing the items extracted by the spider and performing tasks such as collection, verification, and persistence. The Downloader Middleware is a hook framework located between the Scrapy engine and the downloader. It mainly handles requests and responses between the Scrapy engine and the downloader. The Spider Middleware is a hook framework located between the Scrapy engine and the spider. Its main job is to process the spider. der's response input and request output.

[0089] In summary: upload file images through scanning equipment or computers, connect to the system API for identification and return results, combine with specific construction industry data for data proofreading, use clarity enhancement algorithm and image classification algorithm to quickly identify newly added forged seals; through automated processes, including scanning, clarity enhancement, OCR text recognition, NLP processing and classification scoring, the entire recognition process can be greatly accelerated, thereby improving the efficiency of audit work, using the Real-ESRGAN clarity enhancement algorithm, the image quality of scanned copies can be optimized, and the accuracy of OCR and seal recognition can be improved. In addition, NLP processing can further refine text information and enhance the accuracy and completeness of data. Combined with the classification function of the kmeans algorithm, real seals can be quickly distinguished from forged seals. Especially in the construction industry, the authenticity of the seal is crucial to the legal effect of the contract. Therefore, the introduction of this method can significantly reduce the risk of forged seals. By crawling the construction industry enterprise information, the extracted data can be supplemented to further ensure the accuracy and completeness of the recognition results.

[0090] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for identifying the authenticity of a seal scanned copy based on multi-dimensional deep convolution kernel detection, characterized in that: The following steps are involved: Step 1: Scan the information submitted by the project department with a scanner to generate an electronic scanned file of the contract; Step 2: Call the API to execute and enhance the clarity through Real-ESRGAN; Step 3: OCR text recognition. OCR technology can directly recognize text from paper documents, photos, scene photos, or text obtained from some suitable visual devices; Step 4: NLP natural language processing, using NLP content extraction to enable computers to understand, process and generate natural language text or speech; Step 5: Crawling the enterprise information of the construction industry and completing the extracted data; Step 6: kmeans algorithm classification, grouping the data according to similarity, making the data in the same group as similar as possible, and the data in different groups as different as possible; Step 7: Categorize and score the data; Step 8: Output the report that can be published.

2. According to claim 1, a method for identifying the authenticity of a seal based on a multi-dimensional deep convolution kernel detection scan, characterized in that: The Real-ESRGAN clarity enhancement mainly includes the following steps: Predicting high-level distribution and low-level details,Real-ESRGAN first predicts the high-level distribution of the image, and then enhances the details by adding noise by estimating the kernel function related to the low-level details; Using Generative Adversarial Network (GAN), Real-ESRGAN uses GAN to upgrade the original low-resolution images. The Generator is responsible for generating high-resolution images, and the Discriminator is responsible for determining whether an image is generated by the generator or a real high-resolution image. Redesigning the noise injection module and enhanced downsampling layer: The paper proposes two new architectural improvements, including decoupling the noise injection module and enhanced downsampling layer, so that low-level details can be effectively captured and have a positive effect on super-resolution results; New loss function: Real-ESRGAN proposes improved perceptual loss and RelativisticAverage GAN adversarial loss, which enables the model to achieve better results; Optimization for practical application scenarios,Different from many traditional SR methods, Real-ESRGAN pays more attention to the super-encoding rate of real-world images, including images with complex textures and different noise levels.

3. The method for identifying the authenticity of a seal based on a multi-dimensional deep convolution kernel detection scan according to claim 1, characterized in that: The OCR text recognition mainly includes the following steps: Preprocessing usually includes background noise removal, grayscale processing, binarization, and character segmentation. Background noise removal is to retain only text information as much as possible. Grayscale and binarization are to convert color or grayscale images into black and white images that are easier to process. Character segmentation is for subsequent recognition of individual characters. Feature extraction, obtaining effective feature values ​​is the key to improving recognition rate. The purpose of this step is to find the features that distinguish each character, such as character edges, character stroke intersections, etc. Character recognition, after completing feature extraction, the last step is character recognition, this step usually uses some machine learning methods, such as support vector machine (SVM), neural network, etc.

4. The method for identifying the authenticity of a seal based on a multi-dimensional deep convolution kernel detection scan according to claim 1, characterized in that: The NLP natural language processing mainly includes the following steps: Semantic analysis, which involves understanding the meaning of words in a specific context, usually by extracting roots and suffixes, or mapping words to predetermined concepts; Part-of-speech tagging, the process of assigning the correct phonetic label to each word in a text; Syntactic analysis, which involves determining the syntactic structure of a sentence and generating a tree or graph that represents the sentence's syntactic structure; Semantic comprehension, which requires understanding word meanings and the output of syntactic analysis to understand the exact meaning of the sentence; Context recognition and ambiguity resolution: Computers need to understand the specific meaning of words in context, which is very difficult because many words have different meanings in different contexts. Sentiment analysis, which uses natural language processing to determine the emotional tone of text and whether a review is positive or negative; Sentence generation, constructing new sentences through natural language processing, generating responses or descriptions related to the input.

5. The method for identifying the authenticity of a seal based on a multi-dimensional deep convolution kernel detection scan according to claim 1, characterized in that: The kmeans algorithm classification mainly includes the following steps: Initialization: randomly select K points as the initial cluster centers. There are also other optimized initialization methods, such as K-means++. Clustering: For each sample, calculate its distance to each cluster center and classify it into the corresponding category based on the principle of the closest distance. Update: For each category, calculate the mean of all samples in the class and update the cluster center of the category to the newly obtained mean. Iterate and repeat the clustering and updating until the cluster center can no longer be updated, or the update amplitude is less than a given threshold, or the preset maximum number of iterations is reached.

6. The method for identifying the authenticity of a seal based on a multi-dimensional deep convolution kernel detection scan according to claim 1, characterized in that: Scrapy is used when crawling the enterprise information of the construction industry. Scrapy is an open source web data crawling framework written in Python. It provides users with a complete set of solutions for web crawling, data extraction, and data storage.

7. The method for identifying the authenticity of a seal based on a multi-dimensional deep convolution kernel detection scan according to claim 6, characterized in that: The Scrapy crawler framework includes a Scrapy engine, a Scheduler, Downloader, Spiders, Item Pipeline, Downloader Middlewares and Spider middleware.

Images