Image-based job duplication checking system and method

Abstract

The present invention provides an image-based homework duplicate checking system and method. The specific system includes a module for inputting a file to be checked for duplicates, inputting a handwritten homework image or an electronic document, wherein the image or the electronic document is the file to be checked for duplicates; an image preprocessing module for performing image classification and image preprocessing on the file input by the file to be checked for duplicates module; a similarity calculation module for performing image feature extraction and similarity calculation on the image processed by the image preprocessing module; and a display and storage module for displaying the similarity result calculated by the similarity calculation module and storing the similarity result. The duplicate checking analysis of different images is achieved by preprocessing different images in different ways and using corresponding similarity calculation methods. The system and method of the present invention are more widely used for duplicate checking. No matter it is based on duplicate checking detection of handwritten homework pictures or image duplicate checking detection in electronic documents, manual comparison is not required, thereby reducing manual workload.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular, to an image-based job duplication checking system and method. Background Art

[0002] At present, the existing methods of checking for duplicate work can generally be divided into two types: the traditional method of extracting image features and then comparing image features to calculate similarity; the deep learning method of extracting image depth features and then comparing image depth features to calculate similarity. When using the traditional method to extract image features, the methods used include calculating image fingerprint features and calculating image scale-invariant features. However, the calculation of image fingerprint features has poor robustness in terms of image brightness and scale. Therefore, there are often problems such as inaccurate detection results or poor clarity or low restoration of the results obtained by the unified recognition method due to different types of detection.

[0003] At the same time, there is currently a lack of a means of checking for duplicate work images on the market. Manually correcting handwritten homework images is time-consuming and labor-intensive, and often because the handwritten homework images have undergone secondary processing, it leads to problems such as difficulty in checking for duplicate work and poor accuracy, which adds a lot of difficulty and workload to the user's homework checking for duplicate work, resulting in low efficiency, poor effect, and inaccurate results. Summary of the invention

[0004] According to the above-mentioned technical problems that the traditional method of manual evaluation is labor-intensive and inefficient, and the existing computer recognition system has poor robustness, a system and method for checking duplicate work based on images is provided. The present invention mainly focuses on the different characteristics of handwritten work images and images in electronic documents, uses appropriate traditional algorithms to extract image features, and calculates image similarity, thereby reducing the workload of manual evaluation and improving work efficiency.

[0005] Specifically, the present invention includes: an image-based homework duplication checking system, comprising:

[0006] Input the module of files to be checked for duplicates, input the handwritten homework image or electronic document; the image or electronic document is the file to be checked;

[0007] The image preprocessing module performs image classification and preprocessing on the files input by the input duplicate file checking module. First, the type of the file to be detected is determined by reading the suffix of the input file to be detected. The types of the files to be detected include: handwritten homework images and electronic documents. Among them, the images contained in the electronic documents are further divided into plain text images and graphic images. All the electronic documents are unified into PDF format, and the electronic documents are image extracted and classified. For different categories of images, the image of the file to be detected is preprocessed using an image preprocessing method corresponding to the type of the image to be detected.

[0008] A similarity calculation module, according to the type of the file to be detected, uses a feature extraction method and a similarity calculation method corresponding to the type of the file to be detected to perform similarity calculation on the image preprocessed by the image preprocessing module;

[0009] as well as,

[0010] The display and storage module displays and stores the similarity calculation results obtained by the similarity calculation module according to the types of the files to be detected.

[0011] At the same time, the present invention also includes a method for checking for duplicate work based on an image, comprising the following steps:

[0012] Step S1: input a file; the file is a handwritten work image or an electronic document to be detected;

[0013] Step S2: classify and preprocess the input file to be tested; if the input file is determined to be an electronic document, extract the image in the document; perform SVM classification on the image in the input electronic document; save all images in three folders of corresponding handwriting images, plain text images and graphic images; after determining the type of the image to be detected, perform image preprocessing; if the image to be detected is a handwriting image, crop the handwriting image through the image preprocessing module and scale down the cropped image and then perform grayscale processing; if the image to be detected is a plain text image, convert the image into text through the OCR algorithm and save the text as a TXT file; if the image to be detected is a graphic image, perform grayscale processing after unified adjustment;

[0014] Step S3: performing corresponding similarity calculations on the document images to be detected according to the document images to be detected classified in step S2;

[0015] Step S4: Based on the document image to be detected classified in step S2, the corresponding similarity results of the document image to be detected are displayed and stored.

[0016] Compared with the prior art, the present invention has the following advantages:

[0017] 1. The system of the present invention is applicable to a wider range of objects for duplicate checking. Whether it is duplicate checking based on handwritten homework pictures or duplicate checking of images in electronic documents, manual comparison is not required, thus reducing manual workload.

[0018] 2. The system of the present invention comprehensively evaluates and determines the similarity between the plain text image and the graphic image in the electronic document through rules.

[0019] 3. At the same time, the system of the present invention builds multiple databases, including a handwritten homework image database and feature database, and a document image database set feature database. It can be used for subsequent homework duplication checking or resource management. It can check the homework of students horizontally for duplication, and it can also check the duplication vertically by comparing the images in the database. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0021] Figure 1 This is a structural diagram of the image-based job duplication checking system of the present invention.

[0022] Figure 2 This is a flow chart of the image-based job duplication checking method of the present invention. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme 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 only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] like Figure 1-2 As shown, the present invention provides an image-based job duplication checking system and method. Specifically, firstly, in combination with the attached Figure 1 A corresponding explanation is given to the image-based homework duplication checking system described in the present invention.

[0026] As a preferred embodiment of the present application, in the present application, an image-based job duplication checking system includes:

[0027] Input the file to be checked for duplicates module to input the file to be tested; the files to be tested in this application can be divided into handwritten homework images and electronic documents. The handwritten homework images to be tested here can be images taken by a mobile phone, or they can be images after secondary processing of the taken images, such as after cropping, brightening, etc. It can be understood that any method that can obtain a clearer handwritten homework image is acceptable; the electronic document to be tested is the Word or PDF format of the electronic version of the homework, which will contain two types of images in Word or PDF, namely, plain text images and graphic images.

[0028] The image preprocessing module performs image classification and preprocessing on the files input by the input duplicate file check module. First, the type of the file to be detected is determined by reading the suffix of the input file to be detected. The types of the file to be detected include: handwritten homework images and electronic documents, wherein the images contained in the electronic documents can be divided into plain text images and graphic images; all electronic documents are unified into PDF format, and image extraction and image classification are performed; all images are divided into three image types: handwritten homework images, plain text images and graphic images, and the image to be detected is preprocessed using the image preprocessing method corresponding to the type of the image to be detected for different categories of images.

[0029] Specifically, the system has different preprocessing methods for different input images.

[0030] When the file to be checked is a handwritten homework image, since the source of the handwritten homework image (including the image before cropping) is mostly taken with a mobile phone, according to the current smart phone parameters, the image size is generally greater than 600, so a too small image size is not considered. After determining that the input file to be checked for duplicates is a handwritten homework image, the image size is reduced to a certain size and then grayed out.

[0031] When the file to be detected is an electronic document, the electronic document is first unified into PDF format, and then the image in the document is extracted and saved in a folder; by dividing the training image into n blocks, as a preferred implementation, in the embodiment of the present application, the division method is that the block size is set to 16*16, and the grayscale histogram and gradient vector are extracted for each block, and input into the SVM classifier for binary classification training of plain text images and graphic images; the trained SVM classifier is used to classify the image extracted from the electronic document; if the extracted image is a plain text image, the image is converted into text through the OCR algorithm and the text is saved as a TXT file; if the extracted image is a graphic image, the graphic image is unified in size and then grayscale processed.

[0032] The similarity calculation module performs similarity calculation on the image preprocessed by the image preprocessing module according to the type of the file to be detected and using a feature extraction method and a similarity calculation method corresponding to the type of the file to be detected.

[0033] Specifically, different similarity calculations are performed for the image types that are distinguished after analysis by the image preprocessing module. When the file to be detected is a handwritten homework image, the similarity calculation module extracts multiple key points and feature descriptors of the handwritten homework image through the SURF algorithm; and calculates the closest distance and the second closest distance between the key points through the kd tree algorithm in FLANN;

[0034] Calculation point R i and departure point R i The nearest point S j The distance between the points R i R i The next closest point S p The distance relationship between two points is considered to be a match when the following formula is satisfied:

[0035]

[0036] After completing the key point matching of the two images, the RANSAC algorithm is used to remove the wrong matching point pairs to obtain the final matching results, that is, the final matching key point pairs of the two images; the images are screened according to the matching results, the number of matching key point pairs is determined, and the two images that meet the threshold conditions are judged as similar images, wherein the threshold value can be set based on experimental experience. Here, as a preferred implementation, in the embodiment of the present application, based on the experimental experience of the image, the threshold value is set to 12, and similar image pairs are obtained, and the obtained similar image pairs are clustered to cluster images with high similarity.

[0037] When the file to be detected is a plain text image in an electronic document, the text job in the TXT file is divided into multiple sub-parts according to the rules, and the meaning group divided by comma is denoted as x. The text job can be expressed as Where X i represents the i-th job, m i The mth job of the i-th job i There are a number of idea groups, where i∈{1, 2, 3.., 1}, M i ∈{M 1 , M 2 , M 3 , ..., M I}, 1 means there are I jobs in total, M i It means that the i-th job has a total of M i Take the phrase x as the basic segmentation unit, take the minimum of 13 words as the bottom line, and keep the most complete semantics as the condition, and segment each input text job and save it as where Y i represents the result after the i-th job is split, Indicates the nth job of the i-th job i The molecular part of the cut, where N i ∈{N 1 , N 2 , N 3 , ..., N I},N i Indicates that the i-th job is divided into N i sub-parts; then the sub-parts are sim hashed; the sub-parts of each job are respectively tested for similarity with the sub-parts of other jobs, and the similarity scores of any two jobs are calculated; similar jobs are clustered according to the similarity between the jobs.

[0038] When the file to be detected is a graphic image in an electronic document, the key points and feature descriptors of the image are extracted by the ORB algorithm; then the matching key points are screened out by the HAMMING distance in the BF; the RANSAC algorithm is combined to remove the wrong matching points to obtain the final matching result; similar image pairs are screened out by the threshold, and the obtained similar image pairs are clustered to cluster images with high similarity together.

[0039] The display and storage module displays and stores the similarity calculation results obtained by the similarity calculation module according to the type of the file to be detected. For different image types, when it is a handwritten homework, the homework duplication checking system will visualize the clustering results of the handwritten homework image, and randomly select an image from the dissimilar images in the images to be checked for duplication and the similar images in each cluster; and store the image and image features of the dissimilar images in the images to be checked for duplication and the similar images in each cluster in the handwritten homework image library and feature library, and manually mark them.

[0040] If the job duplication checking system judges the similarity between the plain text image and the graphic job corresponding to the same ID, it will conduct a comprehensive similarity evaluation through rules, cluster similar jobs, obtain the clustering of similar jobs, and output the clustering results visually; at the same time, randomly select an image from the dissimilar images in the graphic images to be checked for duplication and the similar images in each cluster, and randomly select an image from the dissimilar images in the images to be checked for duplication and the similar images in each cluster, and store the image features and image library in the document graphic image library and feature library.

[0041] At the same time, if Figure 2 As shown, the present invention also includes an image-based job duplication checking method, comprising the following steps:

[0042] Step S1: input a file; the file is a handwritten work image or an electronic document to be detected;

[0043] Step S2: classify and preprocess the input file to be tested; if the input file is determined to be an electronic document, extract the image in the document; perform SVM classification on the image in the input electronic document; save all images in three folders of corresponding handwriting images, plain text images and graphic images; after determining the type of the image to be detected, perform image preprocessing; if the image to be detected is a handwriting image, crop the handwriting image through the image preprocessing module and scale down the cropped image and then perform grayscale processing; if the image to be detected is a plain text image, convert the image into text through the OCR algorithm and save the text as a TXT file; if the image to be detected is a graphic image, perform grayscale processing after unified adjustment;

[0044] Step S3: performing corresponding similarity calculations on the document images to be detected according to the document images to be detected classified in step S2;

[0045] Step S4: Based on the document image to be detected classified in step S2, the corresponding similarity results of the document image to be detected are displayed and stored.

[0046] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0047] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An image-based homework duplication checking system, characterized in that: include: Input the module for files to be checked for duplicates, input a handwritten homework image or an electronic document, the image or electronic document is the file to be checked for duplicates; The image preprocessing module performs image classification and preprocessing on the files input by the input duplicate file checking module. First, the type of the file to be detected is determined by reading the suffix of the input file to be detected. The types of the files to be detected include: handwritten homework images and electronic documents. Among them, the images contained in the electronic documents are further divided into plain text images and graphic images. All the electronic documents are unified into PDF format, and the electronic documents are image extracted and classified. For different categories of images, the image preprocessing method corresponding to the type of the image to be detected is used to preprocess the image to be detected. A similarity calculation module, according to the type of the file to be detected, uses a feature extraction method and a similarity calculation method corresponding to the type of the file to be detected to perform similarity calculation on the image preprocessed by the image preprocessing module; as well as, A display and storage module, which displays and stores the similarity calculation results obtained by the similarity calculation module according to the type of the file to be detected; When the image preprocessing module determines that the file to be detected is an electronic document, the electronic document is firstly unified into a PDF format, and then the image in the document is extracted and saved in a folder; then the image extracted from the electronic document is classified; by dividing the training image into n blocks, the grayscale histogram and gradient vector of each block are extracted, and the grayscale histogram and gradient vector of each block are input into the SVM classifier for binary classification training of plain text image and graphic image; the image extracted from the electronic document is classified using the trained SVM classifier, and the classified images are respectively saved in corresponding folders; according to the type of the file to be detected, the image of the file to be detected is preprocessed using an image preprocessing method corresponding to the type of the file to be detected, including: If the document to be detected is a handwritten image, the handwritten image is cropped and the cropped image is scaled down and then grayed; If the file to be detected is a plain text image of the electronic document, the image is converted into text through optical character recognition (OCR) algorithm and the text is saved as a TXT file; If the file to be detected is a graphic image of the electronic document, the graphic image is resized and then grayscaled; When the document to be detected is a handwritten homework image, the similarity calculation module extracts multiple key points and feature descriptors of the handwritten homework image through the SURF algorithm; and calculates the closest distance and the second closest distance between the key points through the kd tree algorithm in FLANN; calculates the point R i and departure point R i The nearest point S j The distance between the points R i R i The next closest point S p The distance relationship between two points is considered to be a match when the following formula is satisfied: After completing the key point matching of the two images, the RANSAC algorithm is used to remove the wrong matching point pairs to obtain the final matching result, that is, the final matching key point pairs of the two images; the images are screened according to the matching result, the number of matching key point pairs is determined, the two images that meet the threshold condition are judged as similar images, similar image pairs are obtained, and the obtained similar image pairs are clustered to cluster the images with high similarity together; When the file to be detected is a plain text image in an electronic document, the similarity calculation module divides the text work in the TXT file generated after the plain text image is preprocessed into multiple sub-parts according to the rules, and the meaning group divided by comma is denoted as x, and the text work is represented as Where X i represents the i-th job, m i The mth job of the i-th job i There are meaning groups, where i∈{1,2,3..,I}, M i ∈{M1,M2,M3,…,M I }, I means there are I jobs in total, M i It means that the i-th job has a total of M i a group of ideas; Taking the meaning group x as the basic segmentation unit, with a minimum of 13 words as the bottom line, and with the condition of retaining the most complete semantics, each input text assignment is segmented and saved as where Y i represents the result after the i-th job is split, Indicates the nth job of the i-th job i The molecular part of the cut, where N i ∈{N1,N2,N3,…,N I },N i Indicates that the i-th job is divided into N i sub-parts; then the sub-parts are sim hashed; the sub-parts of each job are respectively tested for similarity with the sub-parts of other jobs, and the similarity scores of any two jobs are calculated; similar jobs are clustered according to the similarity between the jobs; When the file to be detected is a graphic image in an electronic document, the similarity calculation module extracts the key points and feature descriptors of the graphic image through the ORB algorithm; then screens out the matching key points through the HAMMING distance in the BF; combines the RANSAC algorithm to remove erroneous matching points to obtain the final matching result; then screens out similar image pairs through the threshold, clusters the obtained similar image pairs, and groups images with high similarity together.

2. The image-based homework duplicate checking system according to claim 1, characterized in that: When the document image to be detected is a handwritten homework image, the display and storage module visualizes the clustering result of the handwritten homework image and randomly selects an image from the dissimilar images in the document image to be detected and the similar images in each cluster for visualization; and stores the image and image features of the dissimilar images in the image to be checked for duplicates and the similar images in each cluster into the handwritten homework image library and feature library, and performs manual marking.

3. The image-based homework duplication checking system according to claim 1, characterized in that: When the image of the document to be detected is a plain text image, the display and storage module performs a comprehensive similarity evaluation based on the similarity judgment result of the plain text image and the graphic job corresponding to the same ID through a custom rule of weighted summation, clusters similar jobs, obtains the clustering situation of similar jobs, and visually outputs the clustering result; At the same time, an image is randomly selected from the dissimilar images in the graphic images to be checked for duplicates and the similar images in each cluster, and the image and image features of the dissimilar images in the images to be checked for duplicates and the similar images in each cluster are stored in the document graphic image library and feature library.

4. An image-based job duplication checking method using the system described in any one of claims 1 to 3, characterized in that: The following steps are involved: S1: Input file; the file is a handwritten work image or electronic document to be detected; S2: Perform image classification and preprocessing on the input file to be tested; if the input file is determined to be an electronic document, it is necessary to extract the image in the document; perform SVM classification on the image in the input electronic document; save all images in three folders of corresponding handwriting images, plain text images and graphic images; after determining the type of the image to be detected, perform image preprocessing; if the image to be detected is a handwriting image, the handwriting image is cropped by the image preprocessing module and the cropped image is proportionally reduced and then grayed; if the image to be detected is a plain text image, the image is converted into text by the OCR algorithm and the text is saved as a TXT file; if the image to be detected is a graphic image, it is uniformly adjusted and then grayed; S3: performing corresponding similarity calculations on the document images to be detected according to the classification in step S2; S4: According to the document image to be detected classified in step S2, the corresponding similarity result of the document image to be detected is displayed and stored.

Images