Method and system for detecting duplicate reports

Abstract

The application discloses a kind of detection report duplicate detection method and system, comprising: according to document selection instruction, the detection report document to be checked is retrieved.The hash value of detection report document is calculated, and the hash value is compared with each document identifier in the registry as the unique document identifier of detection report, to determine the current duplicate detection state of detection report.If the duplicate detection of detection report is not completed, the detection data is extracted from the detection report document by selecting the appropriate data extraction method.The extracted detection data is stored according to the preset data structure, and the corresponding data index is established by using multi-path balanced tree.Detection data is quickly retrieved by data index during duplicate detection, and the similarity algorithm adapted to detection data is used to evaluate the similarity between detection report and other detection reports, so as to obtain the duplicate detection result of detection report.

Description

Technical Field

[0001] This invention relates to the field of data processing methods or equipment applicable to specific functions, and specifically to a method and system for checking duplicate test reports. Background Technology

[0002] In road and highway inspection projects, the preparation and review of inspection reports are core processes. Inspection reports contain a large number of specialized inspection images, inspection curve data, and related structured information, and it is explicitly required that all images and curve data must be unique. Therefore, the review of inspection reports involves processing a large volume of documents. If the same document is repeatedly inspected multiple times, it will waste system resources and reduce review efficiency.

[0003] Currently, the review of image and curve data in road and highway inspection reports mainly relies on manual screening. Inspectors must compare each type of inspection data in the report one by one, which is not only extremely inefficient but also prone to missed or false positives, failing to meet the needs of rapid review of large batches of inspection reports. Furthermore, the lack of document duplication checks leads to resource waste caused by multiple checks of the same document, further hindering rapid review. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a method and system for detecting duplicate test reports, which can meet the need for rapid review of test reports. The specific technical solution is as follows: Firstly, a method for detecting plagiarism in test reports is provided. In a first feasible implementation of this method, the method includes: Obtain document selection instructions and retrieve the corresponding detection report based on the document selection instructions; Calculate the hash value of the detection report, and compare the hash value as a document identifier with the document identifier in the registry to determine whether the detection report has completed the plagiarism detection. In response to the fact that the detection report has not completed the plagiarism detection, the detection data is extracted from the detection report by retrieving the appropriate data extraction method according to the document format of the detection report; The detection data is stored according to a preset data structure, and a data index is established using a multi-way balanced tree structure; The corresponding detection data is retrieved through the data index, and the similarity between the detection data and other detection reports is calculated using a similarity algorithm adapted to the detection data, thereby obtaining the plagiarism detection result.

[0005] In conjunction with the first possible implementation of the first aspect, in the second possible implementation of the first aspect, calculating the hash value of the detection report includes: The hash value of the detection report is calculated using the SHA-256 algorithm.

[0006] In conjunction with the first possible implementation of the first aspect, in the third possible implementation of the first aspect, extracting test data from the test report includes: The detection data is preprocessed using a preprocessing method adapted to the document format of the detection data.

[0007] In conjunction with the first possible implementation of the first aspect, in the fourth possible implementation of the first aspect, extracting test data from the test report includes: encrypting the test data.

[0008] In conjunction with the first possible implementation of the first aspect, in the fifth possible implementation of the first aspect, extracting test data from the test report includes: For the first type of document format, the detection images and detection curves are selected based on the document association and data structure; For the second type of document format, a page image conversion method is used to convert the document into an image, and an image segmentation algorithm is used to locate the image region and the curve region; An image extraction algorithm is used to extract the detection image from the image region, and a curve recognition algorithm is used to extract the detection curve from the curve region; For the third type of document format, based on the document association and data structure, detection images and detection curves are selected, and detection data are extracted from the detection images and detection curves respectively using appropriate extraction algorithms.

[0009] In conjunction with the first feasible method of the first aspect, in the sixth feasible method of the first aspect, the similarity between the test data and other test reports is calculated, including: When the detection data is a detection image, feature extraction is performed on the detection image to obtain the image feature vector; Based on the image feature vector, the cosine similarity algorithm is used to calculate the similarity between the detection data and other detection reports; When the detection data is a detection curve, feature points are extracted from the detection curve to obtain a feature point sequence; Based on the feature point sequence, the cumulative distance between the detection data and other detection reports is calculated using a dynamic time rule algorithm.

[0010] In conjunction with the first implementable method of the first aspect, the seventh implementable method of the first aspect includes: The image perceptual value of the detected image is calculated using a perceptual hash algorithm, and the image perceptual value is calibrated using an affine transformation method to obtain the image feature vector; And / or, the Douglas-Puk algorithm is used to thin the feature point sequence.

[0011] In conjunction with the first possible implementation of the first aspect, in the eighth possible implementation of the first aspect, thread pool optimization technology and task priority scheduling algorithm are used to process the detection data in parallel.

[0012] In conjunction with the first possible implementation of the first aspect, the ninth possible implementation of the first aspect employs an LRU caching mechanism to cache frequently accessed detection data.

[0013] Secondly, a plagiarism detection system for testing reports is provided, including: The document parsing module is configured to obtain document selection instructions and retrieve the corresponding detection report based on the document selection instructions; The consistency detection module is configured to calculate the hash value of the detection report and compare the hash value as a document identifier with the document identifier in the registry to determine whether the detection report has completed the deduplication detection. The data extraction module is configured to, in response to the fact that the detection report has not completed the plagiarism detection, retrieve the appropriate data extraction method from the detection report according to the document format of the detection report to extract the detection data; The data management module is configured to store the detection data according to a preset data structure and to establish a data index using a multi-way balanced tree structure. The intelligent plagiarism detection module is configured to retrieve the corresponding detection data through the data index, and use a similarity algorithm adapted to the detection data to calculate the similarity between the detection data and other detection reports to obtain the plagiarism detection result.

[0014] Beneficial Effects: The detection report plagiarism detection method and system of this invention can use the calculated hash value as a unique document identifier for the detection report. By comparing it with the document identifier recorded in the registry, the detection status of the currently retrieved detection report can be determined. If the detection report has already completed plagiarism detection, there is no need to perform plagiarism detection on the report again, avoiding resource waste and meeting the needs of rapid review. For incomplete detection reports, the detection data can be extracted from the detection report according to the document format of the detection report using an appropriate data extraction method, to meet the plagiarism detection needs of detection reports with different structures. The detection data is stored according to a preset data structure, and a data index is established using a multi-way balanced tree structure, so that the detection data can be quickly extracted for plagiarism comparison directly through the data index of the multi-way balanced tree structure during subsequent plagiarism comparison, improving data storage and extraction efficiency to meet the needs of rapid review. Finally, a similarity algorithm adapted to the detection data can be used to calculate the similarity between the detection data and other detection reports to obtain the plagiarism detection results, adapting to the plagiarism review of detection reports with different formats. Attached Figure Description

[0015] To more clearly illustrate the specific embodiments of the present invention, the accompanying drawings used in the specific embodiments will be briefly described below. In all the drawings, the elements or parts are not necessarily drawn to scale.

[0016] Figure 1 This is a flowchart of a method for checking duplicate test reports provided in an embodiment of the present invention; Figure 2 This is a system block diagram of a detection report plagiarism detection system provided in an embodiment of the present invention. Detailed Implementation

[0017] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0018] like Figure 1 The flowchart shown illustrates the plagiarism detection method for test reports. This method includes: Step 1: Obtain the document selection instruction and retrieve the corresponding detection report based on the document selection instruction; Step 2: Calculate the hash value of the detection report, and compare the hash value as a document identifier with the document identifier in the registry to determine whether the detection report has completed the plagiarism detection. Step 3: In response to the fact that the detection report has not completed the plagiarism detection, retrieve the detection data from the detection report using the appropriate data extraction method according to the document format of the detection report; Step 4: Store the detection data according to a preset data structure, and establish a data index using a multi-way balanced tree structure; Step 5: Retrieve the corresponding detection data through the data index, and use a similarity algorithm adapted to the detection data to calculate the similarity between the detection data and other detection reports to obtain the plagiarism detection results.

[0019] Specifically, firstly, the testing personnel can input document selection commands through the interactive interface. The system can then automatically retrieve the test report document to be checked from the database storing test report documents. Next, the system automatically calculates the hash value of the test report document and uses this hash value as a unique document identifier. This hash value is compared with various document identifiers in the stored registry to determine the current plagiarism check status of the test report. The plagiarism check status is either "completed" or "incomplete." If the plagiarism check status is "completed," there is no need to further review the retrieved test report document for plagiarism checking.

[0020] Conversely, if the plagiarism check is incomplete, since different highway inspection reports have different document formats, to ensure compatibility with plagiarism checks of multiple formats, a suitable data extraction method can be selected based on the document format to extract relevant inspection data from the inspection report document. The extracted inspection data can then be stored according to a preset data structure, and a corresponding data index can be built using a multi-way balanced tree. This allows for rapid retrieval and location of the inspection data during subsequent plagiarism checks, improving data query and plagiarism check efficiency to meet the needs of rapid review.

[0021] Finally, the detection data from each detection report can be quickly retrieved using a data index. The types of detection data contained in each report also differ, such as detection images and detection curves. During plagiarism detection, a similarity algorithm adapted to the detection data can be used to calculate the similarity between the detection data of the report to be checked and the detection data of other reports, thereby obtaining the plagiarism detection results.

[0022] In this embodiment, after the system obtains the test report document, the tester can select the root storage directory of the test report document through the system's interactive interface, which is usually the test report archive directory. The system can automatically construct a hierarchical storage structure of "root directory - report partition - data category folder", allocate an independent storage partition for each subsequently selected test report document, and set up an access isolation mechanism between different partitions.

[0023] Simultaneously, the system creates database files in the encrypted partition of the root storage directory and initializes data tables, such as data index tables, data attribute tables, data association tables, and data log tables. The system focuses on configuring the multi-way balanced tree and data encryption parameters of the data index tables to adapt to the structured storage, encryption protection, and traceability requirements of the detection data. Furthermore, users can be divided into different roles such as administrators, auditors, and operators, and assigned differentiated data access permissions to achieve hierarchical protection of detection data and prevent the leakage of heterogeneous data.

[0024] In this embodiment, optionally, calculating the hash value of the detection report includes: The hash value of the detection report is calculated using the SHA-256 algorithm.

[0025] Specifically, the SHA-256 algorithm is used to calculate the hash value of each document, which is then injected into the registry as a unique document identifier. Simultaneously, information such as the detection status, detection time, detection results, and storage path of the detection report can be injected into the registry, enabling full lifecycle tracking and intelligent management of the detection report documents.

[0026] In this way, when the system detects that the same document has already been detected, it can automatically skip the entire process and reuse historical results. Furthermore, when subsequent detection data extraction or similarity algorithms are upgraded, the logic of the implemented algorithm can be read from the hash value of each document. If the updated algorithm adds detection content or changes some logic, only an "incremental update" is triggered instead of a "full update." This means that only the data already detected is subjected to the added detection or logic changes, rather than a complete re-detection, avoiding redundant calculations and extending the overall lifespan of the system. This improves the overall system operating efficiency and resource utilization.

[0027] In this embodiment, optionally, extracting test data from the test report includes: For the first type of document format, the detection images and detection curves are selected based on the document association and data structure; For the second type of document format, a page image conversion method is used to convert the document into an image, and an image segmentation algorithm is used to locate the image region and the curve region; An image extraction algorithm is used to extract the detection image from the image region, and a curve recognition algorithm is used to extract the detection curve from the curve region; For the third type of document format, based on the document association and data structure, detection images and detection curves are selected, and detection data are extracted from the detection images and detection curves respectively using appropriate extraction algorithms.

[0028] Specifically, the document formats for the testing reports include three types: Type 1, Type 2, and Type 3. Type 1 is an embedded image format, Type 2 is a floating image format, and Type 3 is an embedded external curve format. Differentiated adaptive extraction algorithms can be used for testing report documents in different formats.

[0029] For the first document format, the image binary data and curve raw data can be filtered out by parsing the relationship and data structure of the detection report document and combining it with a document format parsing engine. Then, by parsing the encoding format of the image binary data, the binary data of various detection images can be obtained. After format recognition of the binary data, the detection images are extracted using existing image extraction algorithms.

[0030] For the second document format, each page of the document can be converted into an image of a preset resolution using a page image conversion algorithm. Combined with image segmentation technology, image and curve regions can be located, and detection-specific image and curve data can be extracted to complete the normalization process and feature extraction.

[0031] For third-party document formats, existing technologies often directly use common image hashing algorithms (such as perceptual hashing, mean hashing, etc.) to process the images in the document for plagiarism detection reports. The design principle of these algorithms is to compress global or local features such as pixel distribution and color depth of the entire image into a hash value of fixed length, and then determine the image similarity by using the Hamming distance of the hash values.

[0032] However, the detection curve in the detection report differs fundamentally from that of a regular image in terms of information structure: The information carrier of ordinary images (such as detection photos and schematic diagrams) is the color depth of pixels and their spatial distribution relationship. The information is large and distributed throughout the entire image. Hash calculation can cover most of the related information.

[0033] The core information of detection curves (such as load-displacement curves and stress-strain curves) is not the pixels themselves, but the coordinate positions of the pixels and the geometric relationships between them (such as linear trends, nonlinear fitting patterns, and piecewise features). In curve images, the effective pixels constituting the curve account for a very small proportion of the entire image, with the vast majority of areas being blank background. When directly performing hash calculations on the entire curve image, the background pixels "dilute" the key information of the curve, causing the hash value to fail to accurately represent the morphological characteristics of the curve, resulting in insufficient reliability of the deduplication results.

[0034] To address the aforementioned issues, this invention incorporates a format adaptation algorithm to specifically optimize the extraction interface during the data extraction stage. Its core principle lies in employing differentiated feature extraction and representation strategies to address the semantic differences of different data types, rather than applying the same hash algorithm to all detection data.

[0035] Specifically, the "format adaptation algorithm optimization extraction interface" includes the following technical means: (1) Data type identification and branch routing The format adaptation algorithm first identifies the type of the extracted detection data, distinguishing whether the current data is a "detection image" or a "detection curve". For different types of data, it routes to different feature extraction interfaces: detection images go through the image feature extraction channel, and detection curves go through the curve feature extraction channel.

[0036] (2) Coordinate space mapping and feature point extraction of curve data For the detection curve, the format adaptation algorithm performs the following optimizations at the extraction interface level: Coordinate space mapping: Mapping a curve image from the pixel coordinate system to the data coordinate system to restore the actual numerical coordinates of each point on the curve; Feature point extraction: A curve feature point extraction algorithm is used to extract a sequence of key feature points that can characterize the overall trend of the curve; Morphological fitting and segmentation: Based on the actual shape of the curve, linear parts are fitted with straight lines, nonlinear parts are fitted with curves, and discontinuous parts are segmented to form a structured description of curve features.

[0037] (3) Differentiated similarity calculation strategy In the subsequent plagiarism detection and comparison stage, the format adaptation algorithm calls different similarity calculation interfaces based on the data type: The detection image uses image feature comparison algorithms such as cosine similarity; The detection curves are calculated using the Dynamic Time Warping (DTW) algorithm to determine the cumulative distance between feature point sequences in order to assess the morphological similarity of the curves.

[0038] By optimizing the extraction interface through the above-mentioned format adaptation algorithm, this invention achieves the following technical effects: Precise representation: Curve data can be accurately represented in the form of a sequence of feature points, avoiding information distortion caused by background pixel interference in traditional hash algorithms; Algorithm adaptation: Different types of data (images and curves) can call their respective adapted similarity algorithms, which significantly improves the accuracy and reliability of the plagiarism detection results; Interface scalability: After being optimized by the format adaptation algorithm, the extraction interface has good scalability and can be easily integrated with more data types (such as tabular data, waveform data, etc.) and their dedicated plagiarism detection algorithms in the future.

[0039] Ultimately, the system achieves full extraction of detection-specific images and curve data.

[0040] In this embodiment, optionally, extracting test data from the test report includes: The detection data is preprocessed using a preprocessing method adapted to the document format of the detection data.

[0041] Specifically, after extracting the detection data from the detection report document, preprocessing methods adapted to the detection data can be used to preprocess it. For example, for detection images, processing such as grayscale conversion, size normalization, and noise filtering can be performed to convert various detection images into unified, standardized images, facilitating subsequent processing of detection images with different structures. For detection curves, outlier removal, data interpolation, and normalization can be performed to convert the detection curves into structured curves with a unified structure.

[0042] In this embodiment, optionally, extracting the test data from the test report includes: encrypting the test data.

[0043] Specifically, to ensure information security during data interaction, existing asymmetric encryption algorithms, such as RSA-2048, can be used to encrypt the data transmission process. Furthermore, RBAC (Restricted Access Control) mechanisms can be employed to assign data access permissions, ensuring that different roles can only access data corresponding to their permissions.

[0044] In this embodiment, semantic parsing algorithms and data standardization algorithms can be used to store the extracted detection data in corresponding data tables according to a preset data structure for data management. In this embodiment, the data structure can include a data index table, a data attribute table, a data association table, and a data log table. In this embodiment, a multi-way balanced search tree structure can be used to build the data index, linking the data attribute table and the data association table to achieve structured storage, rapid extraction, and real-time updating of the detection data. Alternatively, existing symmetric encryption algorithms, such as AES-256 symmetric encryption, can be used to encrypt and store the processed detection data.

[0045] When analyzing the similarity between the test data and other test reports, a LRU caching mechanism can be used to cache frequently accessed test data. This further improves the response speed of data storage and retrieval, adapting to the review requirements of large batches of test reports. Furthermore, the test data in the database can be encrypted and backed up, with the backup data stored on a separate encrypted storage device. In the event of data loss or corruption, testers can quickly restore the data by selecting the corresponding backup file through the backup recovery unit.

[0046] In this embodiment, optionally, calculating the similarity between the detection data and other detection reports includes: When the detection data is a detection image, feature extraction is performed on the detection image to obtain the image feature vector; Based on the image feature vector, the cosine similarity algorithm is used to calculate the similarity between the detection data and other detection reports; When the detection data is a detection curve, feature points are extracted from the detection curve to obtain a feature point sequence; Based on the feature point sequence, the cumulative distance between the detection data and other detection reports is calculated using a dynamic time rule algorithm.

[0047] Specifically, when assessing the similarity between a plagiarism detection report and other reports, the detection data of both reports can be quickly retrieved from the database using an established data index. For the detection images in the data, existing image feature extraction algorithms can be used to extract image feature vectors from both the detection images corresponding to the plagiarism detection report and the comparison detection report. Then, based on the extracted image feature vectors, a cosine similarity algorithm is used to calculate the similarity between the detection images of the plagiarism detection report and the comparison detection report.

[0048] For the detection curves in the detection data, existing feature point extraction algorithms can be used to extract feature point sequences from the detection curves corresponding to the report to be checked for plagiarism and the report to be compared. Then, based on the feature point sequences, a dynamic time rule algorithm is used to calculate the cumulative distance between the detection data and other detection reports.

[0049] In this embodiment, optionally, a perceptual hash algorithm is used to calculate the image perceptual value of the detected image, and an affine transformation method is used to calibrate the image perceptual value to obtain the image feature vector; And / or, the Douglas-Puk algorithm is used to thin the feature point sequence.

[0050] Specifically, when extracting the image feature vector of the detection image, a local image feature extraction algorithm can be used to extract feature points from the detection image. Then, the image perceptual value of the feature points can be calculated using a perceptual hash algorithm, and the image perceptual value can be calibrated using an affine transformation method to obtain the image feature vector, thereby improving the accuracy of image deduplication.

[0051] When extracting the feature point sequence of the detection curve, the Douglas-Puk algorithm can be used to thin out the extracted feature point sequence, remove redundant data points in the feature point sequence, and retain only the core morphological features of the detection curve to improve the efficiency of deduplication.

[0052] In this embodiment, optionally, thread pool optimization technology and task priority scheduling algorithm are used to process the detection data in parallel.

[0053] Specifically, since there are a large number of detection reports involved in the comparison, thread pool optimization technology and task priority scheduling algorithm can be used to achieve concurrent processing of multiple data, prioritize the processing of urgent detection tasks, avoid data processing congestion, and improve the processing efficiency of large batches of detection data.

[0054] like Figure 2 The system block diagram shown is for a plagiarism detection system for test reports. The system includes: The document parsing module is configured to obtain document selection instructions and retrieve the corresponding detection report based on the document selection instructions; The consistency detection module is configured to calculate the hash value of the detection report and compare the hash value as a document identifier with the document identifier in the registry to determine whether the detection report has completed the deduplication detection. The data extraction module is configured to, in response to the fact that the detection report has not completed the plagiarism detection, retrieve the appropriate data extraction method from the detection report according to the document format of the detection report to extract the detection data; The data management module is configured to store the detection data according to a preset data structure and to establish a data index using a multi-way balanced tree structure. The intelligent plagiarism detection module is configured to retrieve the corresponding detection data through the data index, and use a similarity algorithm adapted to the detection data to calculate the similarity between the detection data and other detection reports to obtain the plagiarism detection result.

[0055] Specifically, the plagiarism detection system includes a document parsing module, a consistency detection module, a data extraction module, a data management module, and an intelligent plagiarism detection module. The document parsing module automatically retrieves the plagiarism detection report document from the database storing such reports based on document selection commands. The consistency detection module automatically calculates the hash value of the detection report document and uses this hash value as a unique document identifier. This hash value is then compared with various document identifiers in the stored registry to determine the current plagiarism detection status of the report. The plagiarism detection status is either "completed" or "incomplete." If the plagiarism detection status is "completed," there is no need to further review the retrieved detection report document for plagiarism detection.

[0056] Conversely, if the plagiarism check is incomplete, due to the different document formats of various highway inspection reports, the data extraction module can select an appropriate data extraction method to extract relevant inspection data from the inspection report document to ensure compatibility with multi-format inspection report plagiarism checks. The data management module can store the extracted inspection data according to a preset data structure and establish a corresponding data index using a multi-way balanced tree. This allows for rapid retrieval and location of inspection data during subsequent plagiarism checks, improving data query and plagiarism check efficiency to meet the needs of rapid review.

[0057] The intelligent plagiarism detection module can quickly retrieve detection data from various detection reports in the data management module through data indexing. The types of detection data contained in these reports also vary, such as detection images and detection curves. During plagiarism detection, a similarity algorithm adapted to the detection data can be used to calculate the similarity between the detection data of the report to be checked and the detection data of other detection reports, thereby obtaining the plagiarism detection results.

[0058] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions 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, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for checking the plagiarism of test reports, characterized in that, include: Obtain document selection instructions and retrieve the corresponding detection report based on the document selection instructions; Calculate the hash value of the detection report, and compare the hash value as a document identifier with the document identifier in the registry to determine whether the detection report has completed the plagiarism detection. In response to the fact that the detection report has not completed the plagiarism detection, the detection data is extracted from the detection report by retrieving the appropriate data extraction method according to the document format of the detection report; The detection data is stored according to a preset data structure, and a data index is established using a multi-way balanced tree structure; The corresponding detection data is retrieved through the data index, and the similarity between the detection data and other detection reports is calculated using a similarity algorithm adapted to the detection data, thereby obtaining the plagiarism detection result.

2. The method for checking duplicate test reports according to claim 1, characterized in that, Calculating the hash value of the detection report includes: The hash value of the detection report is calculated using the SHA-256 algorithm.

3. The method for checking duplicate test reports according to claim 1, characterized in that, Extract the test data from the test report, including: The detection data is preprocessed using a preprocessing method adapted to the document format of the detection data.

4. The method for checking duplicate test reports according to claim 1, characterized in that, Extracting test data from the test report includes encrypting the test data.

5. The method for checking duplicate test reports according to claim 1, characterized in that, Extract the test data from the test report, including: For the first type of document format, the detection images and detection curves are selected based on the document association and data structure; For the second type of document format, a page image conversion method is used to convert the document into an image, and an image segmentation algorithm is used to locate the image region and the curve region; An image extraction algorithm is used to extract the detection image from the image region, and a curve recognition algorithm is used to extract the detection curve from the curve region; For the third type of document format, based on the document association and data structure, detection images and detection curves are selected, and detection data are extracted from the detection images and detection curves respectively using appropriate extraction algorithms.

6. The method for checking duplicate test reports according to claim 1, characterized in that, Calculate the similarity between the test data and other test reports, including: When the detection data is a detection image, feature extraction is performed on the detection image to obtain the image feature vector; Based on the image feature vector, the cosine similarity algorithm is used to calculate the similarity between the detection data and other detection reports; When the detection data is a detection curve, feature points are extracted from the detection curve to obtain a feature point sequence; Based on the feature point sequence, the cumulative distance between the detection data and other detection reports is calculated using a dynamic time rule algorithm.

7. The method for checking duplicate test reports according to claim 6, characterized in that, include: The image perceptual value of the detected image is calculated using a perceptual hash algorithm, and the image perceptual value is calibrated using an affine transformation method to obtain the image feature vector; And / or, the Douglas-Puk algorithm is used to thin the feature point sequence.

8. The method for checking duplicate test reports according to claim 1, characterized in that, The detection data is processed in parallel using thread pool optimization techniques and task priority scheduling algorithms.

9. The method for checking duplicate test reports according to claim 1, characterized in that, The LRU caching mechanism is used to cache frequently accessed detection data.

10. A plagiarism detection system for testing reports, characterized in that, include: The document parsing module is configured to obtain document selection instructions and retrieve the corresponding detection report based on the document selection instructions; The consistency detection module is configured to calculate the hash value of the detection report and compare the hash value as a document identifier with the document identifier in the registry to determine whether the detection report has completed the deduplication detection. The data extraction module is configured to, in response to the fact that the detection report has not completed the plagiarism detection, retrieve the appropriate data extraction method from the detection report according to the document format of the detection report to extract the detection data; The data management module is configured to store the detection data according to a preset data structure and to establish a data index using a multi-way balanced tree structure. The intelligent plagiarism detection module is configured to retrieve the corresponding detection data through the data index, and use a similarity algorithm adapted to the detection data to calculate the similarity between the detection data and other detection reports to obtain the plagiarism detection result.

Images