Material preprocessing method and system based on sensitive content detection
By preprocessing and identifying material data in a sensitive content detection system in real time, identifying and eliminating it in combination with a sensitive information identification model, and evaluating the data source in quality, the problem of low accuracy in sensitive information identification in the existing technology is solved, and the problem of low accuracy in recognition of sensitive information is realized More efficient and secure content processing.
Patent Information
- Application Number
- CN202411140195.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-20
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-08-20
AI Technical Summary
When the existing sensitive content detection methods process material data, it is difficult to accurately identify sensitive information, resulting in the deformation of sensitive words avoiding detection and affecting the recognition accuracy.
Adopt material preprocessing methods and systems based on sensitive content detection, and use real-time reception and preprocessing of material data, combine sensitive information recognition models (including image and text recognition models) to identify and eliminate, and perform quality evaluation and blocking of data sources. .
It effectively reduces the risk of sensitive information dissemination, improves the security and processing efficiency of content, ensures that the system receives high-quality data, and enhances reliability and flexibility.
Smart Images

Figure CN119046613A_ABST
Abstract
Description
Technical Field
[0001] The invention proposes a material preprocessing method and system based on sensitive content detection, belonging to the technical field of material data processing. Background Art
[0002] Social networks are popular among netizens for their convenient and flexible information release and fast and efficient network communication methods, but they also provide space for the spread of sensitive content. In order to create a clean and healthy network space and cultivate a positive, healthy and positive network culture, it is necessary to use advanced technical means to conduct necessary detection and filtering of social network content. Publishers of sensitive information often distort sensitive words to avoid detection. Traditional methods of sensitive content detection do not properly process material data, resulting in low accuracy in subsequent material sensitivity recognition. Summary of the invention
[0003] The present invention provides a material preprocessing method and system based on sensitive content detection to solve the technical problems in the above-mentioned prior art. The technical solutions adopted are as follows:
[0004] A material preprocessing method based on sensitive content detection, the material preprocessing method based on sensitive content detection comprising:
[0005] receiving material data from a data source in real time, and preprocessing the material data according to the type of the material data to obtain preprocessed material data;
[0006] Performing sensitive information identification on the material data through a sensitive information identification model, and removing the material data with sensitive information; wherein the sensitive information identification model includes a sensitive information identification model for image identification and a sensitive information identification model for text identification;
[0007] The material quality of the data source is evaluated according to the material data rejection status, and the data source whose material quality does not meet the evaluation requirements is subjected to material reception shielding processing.
[0008] Furthermore, receiving material data from a data source in real time, and preprocessing the material data according to the type of the material data to obtain the preprocessed material data, includes:
[0009] Receive material data from data source in real time;
[0010] Classifying the material data to obtain different types of data information, wherein the different types of data information include text data information, image data information and video data information;
[0011] Performing data processing on the text data information to obtain text material data after data processing corresponding to the text data information;
[0012] Performing data processing on the image data information to obtain image material data after data processing corresponding to the image data information;
[0013] Data processing is performed on the video data information to obtain frame image material data after data processing corresponding to the video data information.
[0014] Furthermore, data processing is performed on the text data information to obtain text material data after data processing corresponding to the text data information, including:
[0015] Performing data cleaning processing on the text data information to remove irrelevant information in the text data information and obtain cleaned text data information, wherein the irrelevant information includes HTML tags, special characters, and URL links, etc.;
[0016] Using a word segmentation tool to segment the cleaned text data information to obtain independent words and phrases corresponding to the cleaned text data information;
[0017] Performing stop word extraction and unified format conversion on the independent words and phrases corresponding to the cleaned text data information, obtaining the text data information after the format conversion, and using the text data information after the format conversion as the target text data information;
[0018] The target text data information is the text material data after data processing corresponding to the text data information.
[0019] Furthermore, data processing is performed on the image data information to obtain image material data after data processing corresponding to the image data information, including:
[0020] Performing noise reduction processing on the image data information to obtain the image data information after the noise reduction processing;
[0021] Performing grayscale adjustment on the image data information after the noise reduction processing to obtain the image data information after the grayscale adjustment;
[0022] Performing image segmentation on the grayscale-adjusted image data information to obtain image data blocks corresponding to different areas;
[0023] Performing feature extraction on the image data blocks to obtain key features corresponding to each image data block, wherein the key features include color features, shape features, texture features, etc.;
[0024] The key feature corresponding to each image data block is the image material data after data processing corresponding to the image data information.
[0025] Further, grayscale adjustment is performed on the image data information after the noise reduction process to obtain the image data information after grayscale adjustment, including:
[0026] Extracting the grayscale value corresponding to the target pixel block;
[0027] Extracting each pixel block contained in the image data information after the noise reduction process, and taking each pixel block as a target pixel block;
[0028] Extracting a pixel block connected to each pixel block, and taking the pixel block connected to each pixel block as a first selected pixel block;
[0029] Extract the grayscale value of the first selected pixel block, and use the grayscale value of the first selected pixel block to obtain a first grayscale coefficient; wherein the first grayscale coefficient is obtained by the following formula:
[0030]
[0031] Among them, h 01 Represents the first grayscale coefficient; H m represents the grayscale value corresponding to the target pixel block; n represents the number of pixel blocks connected to the target pixel block; H 01i represents the gray value corresponding to the i-th pixel block connected to the target pixel block; exp represents the exponential function symbol with e as the base;
[0032] Extracting pixel blocks connected to each first selected pixel block, excluding the target pixel block from the pixel blocks connected to the first selected pixel block, and taking the pixel blocks connected to each first selected pixel block excluding the target pixel block as second selected pixel blocks;
[0033] Extract the gray value of the second selected pixel block, and use the gray value of the second selected pixel block to obtain a second gray coefficient; wherein the second gray coefficient is obtained by the following formula:
[0034]
[0035] Among them, h 02 represents the second grayscale coefficient; m represents the number of second selected pixel blocks corresponding to each first selected pixel block; H 02i represents the gray value corresponding to the i-th second selected pixel block; K represents the first adjustment coefficient; wherein the first adjustment coefficient is obtained by the following formula:
[0036]
[0037] Wherein, m represents the number of second selected pixel blocks corresponding to each first selected pixel block; H 02i H represents the gray value corresponding to the i-th second selected pixel block; x Indicates the grayscale value of the first selected pixel block corresponding to the second selected pixel block;
[0038] Performing grayscale adjustment on the target pixel block by using the grayscale value corresponding to the target pixel block in combination with the first grayscale coefficient and the second grayscale coefficient;
[0039] All target pixel blocks are traversed, and grayscale adjustment is performed on each target pixel block to obtain image data information after grayscale adjustment.
[0040] Further, grayscale adjustment is performed on the target pixel block by using the grayscale value corresponding to the target pixel block in combination with the first grayscale coefficient and the second grayscale coefficient, including:
[0041] Extracting the grayscale value corresponding to the target pixel block;
[0042] extracting a first grayscale coefficient and a second grayscale coefficient;
[0043] The target grayscale value corresponding to the target pixel block is obtained by using the grayscale value corresponding to the target pixel block and the first grayscale coefficient and the second grayscale coefficient; wherein the target grayscale value is obtained by the following formula:
[0044]
[0045] Among them, H f Indicates the target gray value; h 01 Represents the first grayscale coefficient; h 02 Represents the second grayscale coefficient; H m Indicates the grayscale value corresponding to the target pixel block;
[0046] The grayscale value of the target pixel block is adjusted according to the target grayscale value to obtain the target pixel block after the grayscale value is adjusted.
[0047] Furthermore, performing image segmentation on the grayscale-adjusted image data information to obtain image data blocks corresponding to different regions includes:
[0048] Extracting image data information after grayscale adjustment;
[0049] According to the grayscale value contained in the grayscale-adjusted image data information, a first grayscale threshold and a second grayscale threshold are obtained; wherein the first grayscale threshold and the second grayscale threshold are obtained by the following formula:
[0050]
[0051] Among them, H y01 Represents the first grayscale threshold; H y02 Represents the second grayscale threshold; H 01 represents the initial first grayscale threshold; H 02 represents the initial second grayscale threshold; k represents the number of pixel blocks whose target grayscale value is lower than the grayscale value before grayscale adjustment; r represents the number of pixel blocks whose target grayscale value is higher than the grayscale value before grayscale adjustment; H mj represents the grayscale value of the pixel block whose jth target grayscale value is lower than the grayscale value before grayscale adjustment; H fj represents the target grayscale value of the pixel block whose jth target grayscale value is lower than the grayscale value before grayscale adjustment; H mt represents the grayscale value of the pixel block whose tth target grayscale value is higher than the grayscale value before grayscale adjustment; H ft represents the target grayscale value of the pixel block whose t-th target grayscale value is higher than the grayscale value before grayscale adjustment;
[0052] Comparing the grayscale value of each pixel block in the grayscale-adjusted image data information with the first grayscale threshold and the second grayscale threshold;
[0053] Pixel blocks with grayscale values lower than the first grayscale threshold form one or more first image areas;
[0054] Pixel blocks whose grayscale values are not lower than the first grayscale threshold but lower than the second grayscale threshold form one or more second image regions;
[0055] The pixel blocks whose grayscale values are not lower than the second grayscale threshold form one or more third image areas;
[0056] Image segmentation is performed according to the areas enclosed by the first image area, the second image area and the third image area to obtain image data blocks corresponding to different areas.
[0057] Further, data processing is performed on the video data information to obtain frame image material data after data processing corresponding to the video data information, including:
[0058] Performing frame processing on the video data information to obtain a frame image data set corresponding to the video data information;
[0059] A frame interval quantity range is obtained for the frame image data; wherein the upper limit quantity value and the lower limit quantity value corresponding to the frame interval quantity range are obtained by the following formula:
[0060]
[0061] Among them, M up and Mdown represents the upper limit and lower limit of the frame interval range; x represents the number of unit times corresponding to the video data reception, and the unit time is 2s-5s; M c Indicates the preset frame number reference value; M p Indicates the average number of frames per unit time; M i M represents the number of frames of frame data contained in the received video data information corresponding to the i-th unit time; d0 and M u0 represents the upper limit value and the lower limit value of the preset initial frame interval quantity range; s represents the second adjustment coefficient, and the second adjustment coefficient is obtained by the following formula:
[0062]
[0063] Where, s represents the second adjustment coefficient; M z represents the median value of the number of frames contained in the received video data corresponding to n unit time; M max and M min Indicates the maximum and minimum number of frames contained in the received video data corresponding to a unit time;
[0064] Randomly setting a frame interval number that meets the frame interval number range according to the frame interval number range, and extracting frame images from the frame image data set according to the frame interval number to obtain target frame image data;
[0065] Performing noise reduction processing on the target frame image data to obtain the target frame image data after the noise reduction processing;
[0066] Performing grayscale adjustment on the target frame image data after the noise reduction processing to obtain the target frame image data after the grayscale adjustment;
[0067] Performing image segmentation on the grayscale-adjusted target frame image data to obtain image data blocks corresponding to the target frame image data corresponding to different regions;
[0068] Performing feature extraction on the image data blocks corresponding to the target frame image data to obtain key features contained in the image data blocks corresponding to each target frame image data, wherein the key features include color features, shape features, texture features, etc.;
[0069] The key feature corresponding to the image data block corresponding to each target frame image data is the image material data after data processing corresponding to the image data information.
[0070] Furthermore, the material quality of the data source is evaluated according to the material data rejection status, and the data source whose material quality does not meet the evaluation requirements is subjected to material reception shielding processing, including:
[0071] Extract the rejection status information of the material data of each data source;
[0072] The material data quality evaluation parameter corresponding to each data source is obtained according to the rejection status information of the material data; wherein the material data quality evaluation parameter is obtained by the following formula:
[0073]
[0074] Wherein, Q represents the material data quality evaluation parameter; p represents the number of sending unit time experienced by the data source to send the material data, and the sending unit time is 1s; P i represents the proportion of sensitive information data to the total data sent in the i-th sending unit time; P c represents a preset data ratio threshold; z represents the number of transmission unit times during which the data ratio of sensitive information data in p unit times to the total data transmitted does not exceed the preset data ratio threshold; y represents the number of transmission unit times during which the data ratio of sensitive information data in p unit times to the total data transmitted exceeds the preset data ratio threshold; P j represents the data ratio of the jth sensitive information data to the total data sent, which does not exceed the preset data ratio threshold, and corresponds to the data ratio of the sensitive information data to the total data sent per sending unit time; P g Indicates the data ratio of the g-th sensitive information data to the total data sent, corresponding to the sending unit time when the data ratio of the g-th sensitive information data to the total data sent exceeds the preset data ratio threshold;
[0075] Comparing the material data quality evaluation parameter with a preset parameter threshold;
[0076] When the material data quality evaluation parameter is lower than a preset parameter threshold, it is determined that the material quality of the data source does not meet the evaluation requirements;
[0077] Data sources whose material quality does not meet the evaluation requirements are blocked from receiving materials.
[0078] A material preprocessing system based on sensitive content detection, the material preprocessing system based on sensitive content detection comprising:
[0079] A material data receiving module, used to receive material data from a data source in real time, and pre-process the material data according to the type of the material data to obtain pre-processed material data;
[0080] A sensitive information identification module, used to identify sensitive information of the material data through a sensitive information identification model, and remove the material data with sensitive information; wherein the sensitive information identification model includes a sensitive information identification model for image identification and a sensitive information identification model for text identification;
[0081] The material quality assessment module is used to assess the material quality of the data source according to the material data rejection status, and to perform material reception shielding processing on the data source whose material quality does not meet the assessment requirements.
[0082] Beneficial effects of the present invention:
[0083] The material preprocessing method and system based on sensitive content detection proposed in the present invention effectively reduce the risk of sensitive information propagation in the system and improve the security of the content by receiving and preprocessing material data in real time, and combining the sensitive information recognition model for accurate recognition and elimination. The data source quality assessment and shielding mechanism can ensure that the system receives material data from high-quality data sources, reduce the risk of problems caused by receiving low-quality data, and enhance the reliability of the system. The corresponding preprocessing methods and sensitive information recognition models are used for different types of material data, so that the system can process a large amount of material data more quickly and accurately, and improve the processing efficiency. The technical solution is based on modular design, and can flexibly adjust and optimize the functions and performance of each module according to actual needs, and has strong scalability. At the same time, with the continuous development of technology, the system can continuously introduce new preprocessing methods and sensitive information recognition models to adapt to the ever-changing network environment and user needs. BRIEF DESCRIPTION OF THE DRAWINGS
[0084] Figure 1 is a flow chart of the method of the present invention;
[0085] Figure 2 The system block diagram of the system of the present invention. DETAILED DESCRIPTION
[0086] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0087] The embodiment of the present invention proposes a material preprocessing method based on sensitive content detection, such as Figure 1 As shown, the material preprocessing method based on sensitive content detection includes:
[0088] S1. receiving material data from a data source in real time, and preprocessing the material data according to the type of the material data to obtain preprocessed material data;
[0089] S2. Performing sensitive information identification on the material data through a sensitive information identification model, and removing the material data with sensitive information; wherein the sensitive information identification model includes a sensitive information identification model for image identification and a sensitive information identification model for text identification;
[0090] S3. Evaluate the material quality of the data source according to the material data rejection status, and perform material reception shielding processing on the data source whose material quality does not meet the evaluation requirements.
[0091] The working principle of the above technical solution is as follows: the system receives material data from the data source in real time. The data may include images, videos, texts and other forms. According to the type of material data, the system performs corresponding preprocessing on the received data. For images, it may include operations such as scaling, grayscale, and denoising; for text, it may include word segmentation, stop word removal, part-of-speech tagging and other processing. The above preprocessing steps are intended to improve the efficiency and accuracy of subsequent sensitive information identification.
[0092] The pre-processed material data is fed into the sensitive information recognition model for further analysis. The model includes two sensitive information recognition models for images and texts. The image recognition model may use deep learning technology, such as convolutional neural network (CNN), to detect sensitive content in images, such as pornography, violence, etc.
[0093] The text recognition model may use natural language processing technology, such as word segmentation and part-of-speech tagging, combined with machine learning or deep learning algorithms to identify sensitive words or sensitive topics in the text. Once sensitive information is detected, the system will remove the material data with sensitive information to ensure that the material processed later is safe and compliant.
[0094] The system evaluates the quality of data sources based on the exclusion of material data. Specifically, the proportion of sensitive information materials that have been excluded from the materials sent by each data source can be counted as an evaluation indicator of data source quality. For data sources whose quality does not meet the evaluation requirements, the system will take shielding measures and suspend or terminate the receipt of material data from the data source. This helps reduce the risk of sensitive information entering the system and improve the overall security and compliance of the system.
[0095] The effect of the above technical solution is: by real-time reception and preprocessing of material data, and combining with sensitive information identification models for accurate identification and elimination, the risk of sensitive information spreading in the system is effectively reduced, and the security of the content is improved. The data source quality assessment and shielding mechanism can ensure that the system receives material data from high-quality data sources, reduce the risk of problems caused by receiving low-quality data, and enhance the reliability of the system. The corresponding preprocessing methods and sensitive information identification models are used for different types of material data, so that the system can process a large amount of material data more quickly and accurately, and improve processing efficiency. The technical solution is based on modular design, and can flexibly adjust and optimize the functions and performance of each module according to actual needs, and has strong scalability. At the same time, with the continuous development of technology, the system can continuously introduce new preprocessing methods and sensitive information identification models to adapt to the ever-changing network environment and user needs.
[0096] In one embodiment of the present invention, material data is received from a data source in real time, and the material data is preprocessed according to the type of the material data to obtain the preprocessed material data, including:
[0097] S101, receiving material data from a data source in real time;
[0098] S102, classifying the material data to obtain different types of data information, wherein the different types of data information include text data information, image data information and video data information;
[0099] S103, performing data processing on the text data information to obtain text material data after data processing corresponding to the text data information;
[0100] S104, performing data processing on the image data information to obtain processed image material data corresponding to the image data information;
[0101] S105 . Perform data processing on the video data information to obtain frame image material data after data processing corresponding to the video data information.
[0102] The working principle of the above technical solution is: the system or software receives material data from a specified data source (such as a database, API interface, file system, etc.) in real time. Real-time reception ensures the timeliness and freshness of the data, so that subsequent processing can be based on the latest data. The received material data is classified into different types at this step, including text data information, image data information, and video data information. Data classification provides a basis for subsequent specialized processing of different types of data.
[0103] For text data information, the system performs specific data processing operations, such as denoising, word segmentation, part-of-speech tagging, named entity recognition, sentiment analysis, etc. The above processing helps to extract key information from the text or change its format to meet the needs of subsequent applications.
[0104] For image data information, the system may perform image enhancement, denoising, feature extraction, target detection, image segmentation, etc. The above processing can improve image quality, extract key features, or convert the image into a form more suitable for subsequent analysis or application.
[0105] For video data information, processing usually involves video frame extraction, key frame selection, video transcoding, video feature extraction, etc. Video data is usually converted into frame image material data for frame-by-frame analysis or processing.
[0106] The effect of the above technical solution is: through data cleaning, denoising and other processing, the accuracy and reliability of the data are improved, providing a better foundation for subsequent data analysis or application. Data classification and specialized processing enable different types of data to be used more effectively. For example, text data can be used for natural language processing tasks, image data can be used for image recognition or analysis, and video data can be used for video analysis and processing. Real-time reception and processing of data ensures the timeliness and rapid response of data. In addition, specialized processing for different types of data can improve processing efficiency and reduce unnecessary consumption of computing resources. The technical solution is capable of processing multiple types of data, including text, images and videos, and can therefore support a wide range of application scenarios, such as natural language processing, image recognition, video analysis, and the like. The various steps in the technical solution can be used independently or in combination, and can be flexibly configured according to specific needs. In addition, with the emergence of new data processing technologies, the solution can also be easily expanded to support new data types or processing requirements.
[0107] In one embodiment of the present invention, data processing is performed on text data information to obtain text material data after data processing corresponding to the text data information, including:
[0108] S1031, performing data cleaning processing on the text data information to remove irrelevant information in the text data information and obtain cleaned text data information, wherein the irrelevant information includes HTML tags, special characters and URL links, etc.;
[0109] S1032, using a word segmentation tool to segment the cleaned text data information to obtain independent words and phrases corresponding to the cleaned text data information;
[0110] S1033. Remove stop words from and perform unified format conversion on the independent words and phrases corresponding to the text data information after cleaning, obtain the text data information after format conversion, and use the text data information after format conversion as the target text data information;
[0111] Among them, the target text data information is the text material data after data processing corresponding to the text data information.
[0112] The working principle of the above technical solution is as follows: The system will clean the received text data information, aiming to remove the irrelevant information therein, such as HTML tags, special characters, and URL links. The above irrelevant information may cause interference or is unnecessary in subsequent text processing. Therefore, cleaning can improve the quality of text data.
[0113] The text data information after cleaning will be sent to a word segmentation tool for text segmentation. The word segmentation tool will segment the continuous text into independent words and phrases. This step is a basic step in natural language processing and provides a basis for subsequent tasks such as part-of-speech tagging and named entity recognition.
[0114] Next, the system will remove stop words from the segmented words and phrases. Stop words usually refer to those words with high occurrence frequencies but little contribution to the meaning of the text, such as "de" (of), "shi" (is), etc. Removing the above words can reduce the redundancy of text data.
[0115] At the same time, the system will also perform unified format conversion on the words and phrases, such as converting all words to lowercase, removing punctuation marks, etc. The purpose of this step is to unify the format of text data and facilitate subsequent analysis and processing.
[0116] The text data obtained after the above steps of processing, that is, the text data after cleaning, segmentation, stop word removal, and format conversion, is called the target text data information. This is the text material data after data processing corresponding to the text data information and can be used for subsequent natural language processing tasks or applications.
[0117] The effect of the above technical solution is: through cleaning and format conversion, irrelevant information and redundant parts in the text are removed, making the text data cleaner and more standardized, and improving the quality of the text data. After word segmentation and stop word removal, the text data is divided into independent words and phrases, and words that do not contribute much to the meaning of the text are removed, making subsequent natural language processing tasks more efficient and accurate. The processed text data can be used as input for natural language processing tasks such as text classification, sentiment analysis, and keyword extraction, supporting a variety of application scenarios. Operations such as cleaning, word segmentation, stop word removal, and format conversion are common steps in text processing. The above operations can improve processing efficiency through algorithm optimization and parallel computing, making text data processing faster. The technical solution can be flexibly configured according to specific needs, such as adding new cleaning rules, replacing word segmentation tools, etc. At the same time, with the continuous development of natural language processing technology, the solution can also be easily expanded to support more text processing tasks and application scenarios.
[0118] In one embodiment of the present invention, data processing is performed on image data information to obtain processed image material data corresponding to the image data information, including:
[0119] S1041, performing noise reduction processing on the image data information to obtain image data information after noise reduction processing;
[0120] S1042, performing grayscale adjustment on the image data information after the noise reduction processing to obtain grayscale adjusted image data information;
[0121] S1043, performing image segmentation on the grayscale-adjusted image data information to obtain image data blocks corresponding to different areas;
[0122] S1044, performing feature extraction on the image data blocks to obtain key features corresponding to each image data block, wherein the key features include color features, shape features, texture features, etc.;
[0123] The key feature corresponding to each image data block is the image material data after data processing corresponding to the image data information.
[0124] The working principle of the above technical solution is to remove the noise in the image, that is, the unwanted interference information in the image, such as random pixels, color fluctuations, etc. Noise reduction is usually achieved through digital filters or algorithms, such as Gaussian filtering, median filtering, etc., to improve the clarity and quality of the image.
[0125] Grayscale adjustment is to adjust the brightness or contrast of an image to make it visually clearer or meet specific requirements. This is usually achieved by stretching the image histogram, applying gamma correction, etc. to improve the visual effect of the image.
[0126] Image segmentation is the process of dividing an image into multiple regions or objects. This is usually done based on the differences in certain features in the image, such as color, brightness, texture, etc. The result of image segmentation is that the image is divided into multiple regions with similar characteristics, each corresponding to an image data block.
[0127] For each image data block, the system performs a feature extraction operation to obtain the key features of the region. The above key features usually include color features (such as color histogram, color moment, etc.), shape features (such as edges, contours, area, etc.) and texture features (such as gray-level co-occurrence matrix, wavelet transform, etc.). The above features can describe the visual characteristics of the image data block and provide a basis for subsequent applications (such as image recognition, target detection, etc.).
[0128] The effect of the above technical solution is: through noise reduction processing and grayscale adjustment, the clarity and visual effect of the image are improved, providing a better foundation for subsequent image analysis and processing. Image segmentation and feature extraction operations can extract key areas and features in the image, which are crucial for subsequent image recognition, target detection and other tasks. The processed image material data can be used in a variety of application scenarios, such as face recognition, license plate recognition, medical image analysis, industrial automation, etc. Operations such as noise reduction, grayscale adjustment, image segmentation and feature extraction can usually improve processing efficiency through algorithm optimization and parallel computing, making image data processing faster and more efficient. The technical solution can be flexibly configured according to specific needs, such as adding new noise reduction methods, adjusting grayscale adjustment parameters, improving image segmentation algorithms, etc. At the same time, with the continuous development of image processing technology, the solution can also be easily expanded to support more image processing and analysis tasks.
[0129] In one embodiment of the present invention, grayscale adjustment is performed on the image data information after the noise reduction process to obtain the image data information after the grayscale adjustment, including:
[0130] Step A1, extracting the grayscale value corresponding to the target pixel block;
[0131] Step A2, extracting each pixel block contained in the image data information after the noise reduction process, and taking each pixel block as a target pixel block;
[0132] Step A3, extracting pixel blocks connected to each pixel block, and taking the pixel blocks connected to each pixel block as the first selected pixel block;
[0133] Step A4: extract the gray value of the first selected pixel block, and use the gray value of the first selected pixel block to obtain a first gray coefficient; wherein the first gray coefficient is obtained by the following formula:
[0134]
[0135] Among them, h 01 Represents the first grayscale coefficient; H m represents the grayscale value corresponding to the target pixel block; n represents the number of pixel blocks connected to the target pixel block; H 01i represents the gray value corresponding to the i-th pixel block connected to the target pixel block; exp represents the exponential function symbol with e as the base;
[0136] Step A5, extracting pixel blocks connected to each first selected pixel block, excluding the target pixel block from the pixel blocks connected to the first selected pixel blocks, and taking the pixel blocks connected to each first selected pixel block excluding the target pixel block as the second selected pixel blocks;
[0137] Step A6: extract the gray value of the second selected pixel block, and use the gray value of the second selected pixel block to obtain a second gray coefficient; wherein the second gray coefficient is obtained by the following formula:
[0138]
[0139] Among them, h 02 represents the second grayscale coefficient; m represents the number of second selected pixel blocks corresponding to each first selected pixel block; H 02i represents the gray value corresponding to the i-th second selected pixel block; K represents the first adjustment coefficient; wherein the first adjustment coefficient is obtained by the following formula:
[0140]
[0141] Wherein, m represents the number of second selected pixel blocks corresponding to each first selected pixel block; H 02i H represents the gray value corresponding to the i-th second selected pixel block; x Indicates the grayscale value of the first selected pixel block corresponding to the second selected pixel block;
[0142] Step A7, adjusting the grayscale of the target pixel block by using the grayscale value corresponding to the target pixel block in combination with the first grayscale coefficient and the second grayscale coefficient;
[0143] Step A8: traverse all target pixel blocks, and perform grayscale adjustment on each target pixel block to obtain image data information after grayscale adjustment.
[0144] The working principle of the above technical solution is as follows: first, the grayscale value of each pixel block (or single pixel, if a single pixel is processed) is extracted from the image after noise reduction processing.
[0145] Next, each pixel block or pixel in the image is taken as a target pixel block and prepared for grayscale adjustment.
[0146] For each target pixel block, the pixel blocks connected to it (i.e., neighboring pixel blocks) are extracted, and a certain relationship between the grayscale values of the neighboring pixel blocks and the grayscale value of the target pixel block is calculated (through an exponential function). This relationship is used to calculate the first grayscale coefficient, which reflects the grayscale relationship between the target pixel block and its direct neighbors.
[0147] Then, the pixel blocks connected to each neighbor pixel block (i.e., the second-level neighbors) are further considered, but the target pixel block itself is excluded. The grayscale values of the above second-level neighbors are used to calculate the second grayscale coefficient, which reflects the grayscale relationship between the target pixel block and its indirect neighbors. This process also involves a first adjustment coefficient, which takes into account the relationship between the grayscale values of the second-level neighbors and the direct neighbors.
[0148] The target pixel block is gray-scaled by using the original gray-scale value of the target pixel block in combination with the first gray-scale coefficient and the second gray-scale coefficient. The above process takes into account the gray-scale relationship between the target pixel block and its direct and indirect neighbors, thereby obtaining an adjusted gray-scale value.
[0149] Finally, all pixel blocks or pixels in the image are traversed, and the above grayscale adjustment process is repeated for each pixel block or pixel, thereby obtaining the grayscale adjustment result of the entire image.
[0150] The effect of the above technical solution is: by considering the grayscale relationship between the pixel block and its neighbors, the technology can adjust the grayscale value of the image more finely, so that the image has better visual effects while maintaining details. Since the grayscale adjustment is based on the grayscale values of the pixel block (or pixel) and its neighbors, the technology has local adaptability and can make different adjustments according to the grayscale characteristics of different regions. Parameters such as the first adjustment coefficient involved in the technology can be adjusted according to specific needs to adapt to different image processing needs. The grayscale adjustment technology can be further expanded, for example, by introducing more neighbor pixel block levels, using more complex grayscale relationship calculation methods, etc., to improve the accuracy and effect of grayscale adjustment. Grayscale adjustment is one of the basic operations in image processing. The technology can be applied to various scenarios that need to improve the grayscale distribution of the image, such as image enhancement, feature extraction, target detection, etc.
[0151] At the same time, by calculating the grayscale coefficient twice, the grayscale value of the target pixel block can be adjusted more accurately, so that the grayscale of the final image is more in line with expectations. This gradual adjustment method helps to improve the meticulousness and accuracy of image processing. Using the grayscale information of adjacent pixel blocks for adjustment can ensure the grayscale consistency of the image in the local area and reduce the unnatural feeling caused by grayscale mutation. This method helps to improve the overall quality and visual effect of the image. By combining the first grayscale coefficient and the second grayscale coefficient, the effects of the two adjustments can be effectively utilized to further improve the grayscale adjustment accuracy and consistency of the image. This superposition effect can optimize the quality of the final image. At the same time, the ability of dynamic adjustment is provided, so that the method can be flexibly adjusted according to the characteristics of the specific image, thereby adapting to the processing requirements of different types of images.
[0152] By calculating the above formula, the grayscale value of each pixel block can be accurately obtained. The calculations of average value, difference value, etc. involved in the formula ensure the accuracy of the result. The adjustment coefficient is introduced into the second grayscale adjustment formula. By dynamically adjusting the result of the first grayscale adjustment, the second grayscale adjustment can be more flexibly adapted to the characteristics of different images, thereby achieving a better adjustment effect. At the same time, the above formula performs weighted averaging and difference calculations on the grayscale information of adjacent pixel blocks, so that the adjustment result not only considers the grayscale of the current pixel block, but also integrates the grayscale information of the surrounding pixel blocks. This method can improve the overall consistency and naturalness of the image and reduce the unnatural feeling caused by grayscale mutation.
[0153] At the same time, by calculating the grayscale coefficient twice, each adjustment gradually refines the calculation of the grayscale value, making the final result more accurate and natural. The first adjustment provides a good foundation for the second adjustment, and the layer-by-layer fine adjustment method can gradually optimize the image quality. On the other hand, during the calculation process, by averaging and weighting the grayscale values, the random noise in the image can be effectively suppressed, and the smoothness and quality of the image can be improved. This provides an effective solution to the problem of noise interference in image processing. The exponential function and average value calculation involved in the formula ensure the stability and robustness of the adjustment result. Even when processing different types of images, the formula can still maintain a good adjustment effect.
[0154] In general, the above formula can achieve high-precision grayscale adjustment through precise calculation and gradual adjustment, improve the overall quality and consistency of the image, and has strong dynamic adaptability and noise suppression effect.
[0155] In one embodiment of the present invention, grayscale adjustment is performed on the target pixel block by using the grayscale value corresponding to the target pixel block in combination with the first grayscale coefficient and the second grayscale coefficient, including:
[0156] Step A701, extracting the grayscale value corresponding to the target pixel block;
[0157] Step A702, extracting the first grayscale coefficient and the second grayscale coefficient;
[0158] Step A703: Obtain a target grayscale value corresponding to the target pixel block using the grayscale value corresponding to the target pixel block and the first grayscale coefficient and the second grayscale coefficient; wherein the target grayscale value is obtained by the following formula:
[0159]
[0160] Among them, H f Indicates the target gray value; h 01 Represents the first grayscale coefficient; h 02 Represents the second grayscale coefficient; H m Indicates the grayscale value corresponding to the target pixel block;
[0161] Step A704: adjust the grayscale value of the target pixel block according to the target grayscale value to obtain the target pixel block after the grayscale value is adjusted.
[0162] The working principle of the above technical solution is: first, the original grayscale value (Hm) corresponding to the target pixel block is extracted from the image data.
[0163] Then, based on the results calculated in the previous step, the first grayscale coefficient (h01) and the second grayscale coefficient (h02) are extracted.
[0164] The target grayscale value (Hf) is calculated by combining the original grayscale value (Hm), the first grayscale coefficient (h01) and the second grayscale coefficient (h02) of the target pixel block using a specific formula. The above formula takes into account the grayscale relationship between the target pixel block and its direct and indirect neighbors, and uses these two coefficients to weight and adjust the original grayscale value.
[0165] The calculated target grayscale value (Hf) is used to update the grayscale value of the target pixel block. This usually means setting the grayscale value of the target pixel block to the newly calculated target grayscale value, thereby completing the grayscale adjustment.
[0166] The effect of the above technical solution is: by considering the grayscale relationship between the target pixel block and its neighbors, the technology can enhance the local contrast of the image. If the grayscale value difference between the target pixel block and its neighbors is large, the adjusted grayscale value may be more obvious, thereby improving the visual quality of the image. Since the technology is calculated based on the grayscale values of the pixel block (or pixel) and its neighbors, it can retain the detail information of the image to a certain extent. This is particularly useful for application scenarios that need to maintain tiny details or textures in the image. The calculation method of the first grayscale coefficient and the second grayscale coefficient can be adjusted according to specific needs to adapt to different image processing needs. In addition, since the technology is calculated based on the grayscale value of the pixel block (or pixel), it can be flexibly applied to images of different resolutions and sizes. Through grayscale adjustment, the visual effect of the image can be significantly improved. For example, in some cases, the brightness and contrast of the image may be more uniform, making the image look clearer and easier to observe. Grayscale adjustment is a basic operation in image processing, and the technology can be applied to a variety of scenarios, such as image enhancement, image restoration, target detection, etc. By adjusting the grayscale value, the visual effect of the image can be improved, providing a better basis for subsequent processing and analysis.
[0167] At the same time, by introducing the first grayscale coefficient (h01) and the second grayscale coefficient (h02), the technical solution allows the grayscale value of the target pixel block to be flexibly adjusted. These two coefficients can be set according to specific application scenarios and requirements, so as to achieve fine control of the grayscale distribution of the image. This flexibility enables the technical solution to adapt to a variety of image processing requirements, such as enhancing image contrast, improving image brightness, etc. By adjusting the grayscale coefficient, the grayscale dynamic range of the image can be effectively expanded or compressed. The technical solution is a grayscale adjustment for the target pixel block, which means that it can perform fine processing on the local area of the image. This local processing capability is very useful for enhancing or weakening specific areas in the image (such as dark areas, bright areas or edge areas), which helps to improve the overall quality of the image. At the same time, the basic framework of the technical solution (i.e., extracting grayscale values, grayscale coefficients, and applying formulas to calculate target grayscale values) provides the possibility for further expansion and customization. For example, more grayscale coefficients can be introduced to implement more complex grayscale adjustment strategies, or the method can be combined with other image processing technologies to achieve more advanced image processing effects.
[0168] In one embodiment of the present invention, image segmentation is performed on the grayscale-adjusted image data information to obtain image data blocks corresponding to different regions, including:
[0169] Step B1, extracting image data information after grayscale adjustment;
[0170] Step B2: obtaining a first grayscale threshold and a second grayscale threshold according to the grayscale value contained in the grayscale-adjusted image data information; wherein the first grayscale threshold and the second grayscale threshold are obtained by the following formula:
[0171]
[0172] Among them, H y01 Represents the first grayscale threshold; H y02 Represents the second grayscale threshold; H 01 represents the initial first grayscale threshold; H 02 represents the initial second grayscale threshold; k represents the number of pixel blocks whose target grayscale value is lower than the grayscale value before grayscale adjustment; r represents the number of pixel blocks whose target grayscale value is higher than the grayscale value before grayscale adjustment; H mj represents the grayscale value of the pixel block whose jth target grayscale value is lower than the grayscale value before grayscale adjustment; H fj represents the target grayscale value of the pixel block whose jth target grayscale value is lower than the grayscale value before grayscale adjustment; H mt represents the grayscale value of the pixel block whose tth target grayscale value is higher than the grayscale value before grayscale adjustment; H ft represents the target grayscale value of the pixel block whose t-th target grayscale value is higher than the grayscale value before grayscale adjustment;
[0173] Step B3, comparing the grayscale value of each pixel block in the grayscale-adjusted image data information with the first grayscale threshold and the second grayscale threshold;
[0174] Step B4, forming one or more first image areas from pixel blocks whose grayscale values are lower than the first grayscale threshold;
[0175] Step B5, forming one or more second image regions with pixel blocks whose grayscale values are not lower than the first grayscale threshold but lower than the second grayscale threshold;
[0176] Step B6, forming one or more third image areas with pixel blocks whose grayscale values are not lower than the second grayscale threshold;
[0177] Step B7: performing image segmentation according to the areas enclosed by the first image area, the second image area and the third image area to obtain image data blocks corresponding to different areas.
[0178] The working principle of the above technical solution is as follows: first, image data information is extracted from the grayscale-adjusted image, and the above information includes the grayscale value of each pixel block.
[0179] According to the grayscale value distribution of the image data information after grayscale adjustment, two grayscale thresholds are calculated: the first grayscale threshold (Hy01) and the second grayscale threshold (Hy02). The calculation of these two thresholds takes into account the change of the target grayscale value relative to the grayscale value before grayscale adjustment, as well as the distribution of the above changes. The initial thresholds (H01 and H02) can be set according to the characteristics of the image or experience, and then adjusted by iterative or adaptive methods.
[0180] The grayscale value of each pixel block in the grayscale-adjusted image is compared with two thresholds to determine the grayscale range to which each pixel block belongs.
[0181] According to the result of the gray value comparison, pixel blocks with gray values lower than the first gray threshold are combined into one or more first image regions.
[0182] Pixel blocks with grayscale values between the first grayscale threshold and the second grayscale threshold are combined into one or more second image regions.
[0183] Pixel blocks with grayscale values higher than the second grayscale threshold are combined into one or more third image regions.
[0184] Based on the image regions with different grayscale ranges formed above, the image is segmented to obtain image data blocks corresponding to different regions. The image data blocks represent regions with different grayscale characteristics in the image.
[0185] The effect of the above technical solution is: the image segmentation method based on grayscale threshold usually has high computational efficiency and can quickly divide the image into different regions. Since a clear grayscale threshold is used for segmentation, the segmentation result has strong interpretability and is convenient for subsequent analysis and processing. The calculation of the grayscale threshold takes into account the change in the grayscale value of the pixel block before and after the grayscale adjustment, so the segmentation method can adapt to images with different grayscale distribution characteristics. The initial grayscale threshold can be adjusted according to the specific application scenario to meet different segmentation requirements. In addition, the segmentation process can be further refined as needed, for example, by introducing more grayscale thresholds to form more image regions. This image segmentation technology can be applied to a variety of image processing and analysis tasks, such as target detection, feature extraction, image classification, etc. By segmenting image regions with different grayscale characteristics, more useful information can be provided for subsequent tasks.
[0186] On the other hand, by dynamically calculating the first grayscale threshold (Hy01) and the second grayscale threshold (Hy02), the technical solution can adaptively adjust the threshold according to the image data information after grayscale adjustment. This adaptability enables the threshold to more accurately reflect the actual changes in the image content, thereby improving the accuracy and robustness of image segmentation. When calculating the threshold, the technical solution not only considers the initial grayscale threshold (H01 and H02), but also considers the changes in the grayscale values of the pixel blocks before and after grayscale adjustment (such as Hmj and Hfj, Hmt and Hft). This consideration makes the determination of the threshold more consistent with the image characteristics after grayscale adjustment, which helps to more accurately segment different areas in the image.
[0187] By setting two grayscale thresholds, the technical solution can subdivide the image into three different areas (first image area, second image area and third image area), thereby achieving more refined image segmentation. This fine segmentation is helpful for subsequent image analysis and processing tasks, such as target detection, feature extraction, etc. Since the determination of the threshold is based on the overall image data information after grayscale adjustment, the technical solution has a certain robustness to noise and local outliers in the image. This means that even if there are some undesirable grayscale value changes in the image, the technical solution can still segment different areas in the image more accurately.
[0188] At the same time, the basic framework of this technical solution (i.e., grayscale adjustment, threshold calculation, pixel block comparison, and region division) provides the possibility for further expansion and optimization. For example, more complex threshold calculation methods can be introduced, and other image features can be combined for segmentation to meet different image processing requirements.
[0189] In one embodiment of the present invention, data processing is performed on video data information to obtain frame image material data after data processing corresponding to the video data information, including:
[0190] Step C1, performing frame processing on the video data information to obtain a frame image data set corresponding to the video data information;
[0191] Step C2, obtaining a frame interval quantity range for the frame image data; wherein the upper limit quantity value and the lower limit quantity value corresponding to the frame interval quantity range are obtained by the following formula:
[0192]
[0193] Among them, M up and M down represents the upper limit and lower limit of the frame interval range; x represents the number of unit times corresponding to the video data reception, and the unit time is 2s-5s; M c Indicates the preset frame number reference value; Mp Indicates the average number of frames per unit time; M i M represents the number of frames of frame data contained in the received video data information corresponding to the i-th unit time; d0 and M u0 represents the upper limit value and the lower limit value of the preset initial frame interval quantity range; s represents the second adjustment coefficient, and the second adjustment coefficient is obtained by the following formula:
[0194]
[0195] Where, s represents the second adjustment coefficient; M z represents the median value of the number of frames contained in the received video data corresponding to n unit time; M max and M min Indicates the maximum and minimum number of frames contained in the received video data corresponding to a unit time;
[0196] Step C3, randomly setting a frame interval number that meets the frame interval number range according to the frame interval number range, and extracting frame images from the frame image data set according to the frame interval number to obtain target frame image data;
[0197] Step C4, performing noise reduction processing on the target frame image data to obtain the target frame image data after noise reduction processing;
[0198] Step C5, performing grayscale adjustment on the target frame image data after the noise reduction processing to obtain the target frame image data after grayscale adjustment;
[0199] Step C6, performing image segmentation on the grayscale-adjusted target frame image data to obtain image data blocks corresponding to the target frame image data corresponding to different regions;
[0200] Step C7, performing feature extraction on the image data blocks corresponding to the target frame image data, and obtaining key features contained in the image data blocks corresponding to each target frame image data, wherein the key features include color features, shape features, texture features, etc.;
[0201] The key feature corresponding to the image data block corresponding to each target frame image data is the image material data after data processing corresponding to the image data information.
[0202] The working principle of the above technical solution is: first, frame processing is performed on the video data information, that is, the continuous video data stream is split into independent frame images to form a frame image data set.
[0203] Next, a reasonable range of frame intervals is calculated based on the number of frames received per unit time of the video data. The upper and lower limits of the above range are obtained through a series of complex calculations, taking into account factors such as the average number of frames of the video, the median, maximum, minimum, and initial range of frame intervals, and are dynamically adjusted using the second adjustment coefficient s.
[0204] After determining the range of the number of frame intervals, the number of frame intervals that meets the above range is randomly selected, and a corresponding number of frame images are extracted from the frame image data set to form target frame image data.
[0205] The extracted target frame image data is subjected to noise reduction processing to reduce noise interference in the image and improve image quality.
[0206] The grayscale of the target frame image data after the noise reduction process is adjusted, which may include enhancing the contrast and brightness of the image, so as to further improve the visualization effect of the image.
[0207] The grayscale-adjusted target frame image data is segmented and divided into different regions, each of which corresponds to an image data block.
[0208] Feature extraction is performed on each image data block to obtain key features including color features, shape features, texture features, etc. The above key features will be used as image material data after image processing for subsequent analysis or application.
[0209] The effect of the above technical solution is: through frame processing and frame image extraction, key frame images can be quickly extracted from video data, thereby improving the efficiency of data processing. Through steps such as noise reduction processing, grayscale adjustment and image segmentation, the image quality is improved and the accuracy of feature extraction is improved. The determination of the range of frame interval quantity takes into account a variety of factors, and a dynamic adjustment method is used, so that the technical solution can adapt to video data of different quality and content. By extracting a variety of features including color, shape and texture, rich image material data is obtained, providing more possibilities for subsequent analysis or application. The technical solution can be expanded and optimized according to specific needs, such as adding new feature extraction methods, adjusting the method for determining the range of frame interval quantity, etc., to meet different application needs.
[0210] On the other hand, by dynamically calculating the range of frame intervals (Mup and Mdown), the technical solution can flexibly adjust the number of frame intervals according to the actual situation of the video data information (such as the number of frames received per unit time, the maximum and minimum number of frames, etc.), thereby achieving effective extraction of video frames. This flexibility enables the technical solution to adapt to the video data processing needs of different frame rates and different scenes. By setting a reasonable range of frame intervals and randomly extracting frame images within this range, the technical solution can effectively reduce the number of frame images that need to be processed while ensuring data representativeness, thereby optimizing data processing efficiency. This is particularly important for large-scale video data processing tasks.
[0211] The denoising and grayscale adjustment steps help improve the quality of the target frame image data. Noise reduction can reduce noise interference in the image and make the image clearer; grayscale adjustment can adjust the image's brightness, contrast and other parameters as needed to make the image more in line with the requirements of subsequent processing or analysis. The image segmentation step divides the target frame image data into different regions and extracts the key features of each region. This fine image segmentation helps with subsequent feature extraction and image analysis tasks and can more accurately capture useful information in the image. The technical solution not only extracts basic image features such as color features and shape features, but also considers higher-level features such as texture features. This comprehensive feature extraction can provide rich information support for subsequent image recognition, classification and other tasks. The basic framework of the technical solution (i.e., frame processing, frame interval number range calculation, frame image extraction, denoising, grayscale adjustment, image segmentation and feature extraction) provides the possibility for further expansion and customization. For example, more complex denoising algorithms, grayscale adjustment strategies or feature extraction methods can be introduced to adapt to different application scenarios and needs. By considering multiple factors such as the average number of frames per unit time, the maximum and minimum number of frames, etc. to calculate the range of the number of frame intervals, this technical solution has a certain degree of robustness against abnormal values or noise in the video data. This helps to ensure that even when the video data quality is not high, a more reliable data processing result can be obtained.
[0212] In one embodiment of the present invention, material quality evaluation is performed on the data source according to the material data rejection status, and material reception shielding processing is performed on the data source whose material quality does not meet the evaluation requirements, including:
[0213] S301, extracting the rejection status information of the material data of each data source;
[0214] S302: Obtain a material data quality evaluation parameter corresponding to each data source according to the rejection status information of the material data; wherein the material data quality evaluation parameter is obtained by the following formula:
[0215]
[0216] Wherein, Q represents the material data quality evaluation parameter; p represents the number of sending unit time experienced by the data source to send the material data, and the sending unit time is 1s; P i represents the proportion of sensitive information data to the total data sent in the i-th sending unit time; P c represents a preset data ratio threshold; z represents the number of transmission unit times during which the data ratio of sensitive information data in p unit times to the total data transmitted does not exceed the preset data ratio threshold; y represents the number of transmission unit times during which the data ratio of sensitive information data in p unit times to the total data transmitted exceeds the preset data ratio threshold; P j represents the data ratio of the jth sensitive information data to the total data sent, which does not exceed the preset data ratio threshold, and corresponds to the data ratio of the sensitive information data to the total data sent per sending unit time; P g Indicates the data ratio of the g-th sensitive information data to the total data sent, corresponding to the sending unit time when the data ratio of the g-th sensitive information data to the total data sent exceeds the preset data ratio threshold;
[0217] S303, comparing the material data quality evaluation parameter with a preset parameter threshold;
[0218] S304: When the material data quality evaluation parameter is lower than a preset parameter threshold, it is determined that the material quality of the data source does not meet the evaluation requirement;
[0219] S305: Perform material reception shielding processing on data sources whose material quality does not meet the evaluation requirements.
[0220] The working principle of the above technical solution is: extract the rejection status information of the material data from the data source. The rejection status information usually reflects whether the material data contains sensitive information or content that does not meet the requirements during transmission or processing, and the rejection status of the above content.
[0221] Based on the elimination status information, the material data quality evaluation parameter (Q) corresponding to each data source is calculated. The calculation of this parameter takes into account the proportion of sensitive information data in the sending unit time (e.g. 1 second) experienced by the data source to send the material data, and compares it with the preset data proportion threshold (Pc). Each item in the formula reflects the different impacts of the proportion of sensitive information data in different time periods.
[0222] The calculated material data quality evaluation parameter (Q) is compared with a preset parameter threshold, which represents the lower limit of acceptable data quality.
[0223] If the material data quality evaluation parameter (Q) is lower than the preset parameter threshold, it is determined that the material quality of the data source does not meet the evaluation requirements. This means that the material data sent by the data source contains too much sensitive information or content that does not meet the requirements and needs to be processed.
[0224] For data sources whose material quality does not meet the evaluation requirements, material reception shielding is performed. This usually means that the system or platform will no longer receive material data sent by the data source, or perform special processing on the above data (such as filtering, marking, etc.) to ensure the quality of data used later.
[0225] The effect of the above technical solution is: by performing quality assessment on the material data sent by the data source, it can be ensured that the quality of the data received by the system or platform meets certain standards, thereby improving the accuracy and reliability of subsequent processing. By calculating the proportion of sensitive information data and setting corresponding thresholds, the risk of sensitive information leakage can be effectively reduced. For data sources containing too much sensitive information, shielding processing can avoid the potential security threats posed by the above sensitive information to the system or platform. Shielding processing for data sources that do not meet the evaluation requirements can reduce the amount of invalid data processed by the system or platform, thereby improving the efficiency and accuracy of data processing. The preset parameter thresholds and data ratio thresholds can be adjusted according to actual needs to adapt to different application scenarios and data quality requirements. This makes the technical solution highly configurable and flexible.
[0226] On the other hand, by comprehensively considering the proportion of sensitive information data in the process of data source sending material data and whether these proportions exceed the preset threshold, the technical solution can more accurately evaluate the material data quality of each data source. This accuracy helps to accurately identify those data sources that frequently send too much sensitive information, thereby ensuring the reliability of subsequent data processing and application. The technical solution uses a dynamic time window (i.e., sending unit time) to evaluate the material data quality of the data source in different time periods. This dynamic adaptability enables the evaluation results to better reflect the actual performance of the data source, especially when the data source performance is unstable or the data quality fluctuates greatly. The formulas involved in the calculation process are relatively simple and are mainly based on proportion and counting operations, which makes the entire evaluation process have high computational efficiency. At the same time, through the quick comparison of the preset parameter threshold, it is possible to quickly determine which data sources do not meet the evaluation requirements, and then take shielding measures to improve processing efficiency. The parameters in the technical solution (such as the preset data ratio threshold, parameter threshold, etc.) can be adjusted according to actual needs to adapt to different evaluation standards and application scenarios. This flexibility enables the technical solution to be widely used in various material data quality evaluation and data source management scenarios. By comprehensively considering the proportion of sensitive information data within multiple sending units of time, this technical solution has a certain degree of robustness against individual outliers or noise. Even if the quality of the material data sent by the data source in a certain period of time is poor, if the overall performance still meets the evaluation criteria, then the data source will not be incorrectly blocked. The entire evaluation process is automated and intelligent, and the evaluation of material data quality and the shielding of data sources can be automatically completed without human intervention. This reduces the cost and error rate of manual operations and improves work efficiency and accuracy.
[0227] To sum up, this technical solution has the technical effects of accuracy, dynamic adaptability, efficiency, flexibility, robustness, automation and intelligence in material quality assessment and data source management, and can effectively improve the quality and efficiency of data processing.
[0228] An embodiment of the present invention provides a material preprocessing system based on sensitive content detection, such as Figure 2 As shown, the material preprocessing system based on sensitive content detection includes:
[0229] A material data receiving module, used to receive material data from a data source in real time, and pre-process the material data according to the type of the material data to obtain pre-processed material data;
[0230] A sensitive information identification module, used to identify sensitive information of the material data through a sensitive information identification model, and remove the material data with sensitive information; wherein the sensitive information identification model includes a sensitive information identification model for image identification and a sensitive information identification model for text identification;
[0231] The material quality assessment module is used to assess the material quality of the data source according to the material data rejection status, and to perform material reception shielding processing on the data source whose material quality does not meet the assessment requirements.
[0232] The working principle of the above technical solution is as follows: the system receives material data from the data source in real time. The data may include images, videos, texts and other forms. According to the type of material data, the system performs corresponding preprocessing on the received data. For images, it may include operations such as scaling, grayscale, and denoising; for text, it may include word segmentation, stop word removal, part-of-speech tagging and other processing. The above preprocessing steps are intended to improve the efficiency and accuracy of subsequent sensitive information identification.
[0233] The pre-processed material data is fed into the sensitive information recognition model for further analysis. The model includes two sensitive information recognition models for images and texts. The image recognition model may use deep learning technology, such as convolutional neural network (CNN), to detect sensitive content in images, such as pornography, violence, etc.
[0234] The text recognition model may use natural language processing technology, such as word segmentation and part-of-speech tagging, combined with machine learning or deep learning algorithms to identify sensitive words or sensitive topics in the text. Once sensitive information is detected, the system will remove the material data with sensitive information to ensure that the material processed later is safe and compliant.
[0235] The system evaluates the quality of data sources based on the exclusion of material data. Specifically, the proportion of sensitive information materials that have been excluded from the materials sent by each data source can be counted as an evaluation indicator of data source quality. For data sources whose quality does not meet the evaluation requirements, the system will take shielding measures and suspend or terminate the receipt of material data from the data source. This helps reduce the risk of sensitive information entering the system and improve the overall security and compliance of the system.
[0236] The effect of the above technical solution is: by real-time reception and preprocessing of material data, and combining with sensitive information identification models for accurate identification and elimination, the risk of sensitive information spreading in the system is effectively reduced, and the security of the content is improved. The data source quality assessment and shielding mechanism can ensure that the system receives material data from high-quality data sources, reduce the risk of problems caused by receiving low-quality data, and enhance the reliability of the system. The corresponding preprocessing methods and sensitive information identification models are used for different types of material data, so that the system can process a large amount of material data more quickly and accurately, and improve processing efficiency. The technical solution is based on modular design, and can flexibly adjust and optimize the functions and performance of each module according to actual needs, and has strong scalability. At the same time, with the continuous development of technology, the system can continuously introduce new preprocessing methods and sensitive information identification models to adapt to the ever-changing network environment and user needs.
[0237] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if the above-mentioned modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include the above-mentioned modifications and variations.
Claims
1. A material preprocessing method based on sensitive content detection, characterized in that: The material preprocessing method based on sensitive content detection includes: receiving material data from a data source in real time, and preprocessing the material data according to the type of the material data to obtain preprocessed material data; Performing sensitive information identification on the material data through a sensitive information identification model, and removing the material data with sensitive information; wherein the sensitive information identification model includes a sensitive information identification model for image identification and a sensitive information identification model for text identification; The material quality of the data source is evaluated according to the material data rejection status, and the data source whose material quality does not meet the evaluation requirements is subjected to material reception shielding processing.
2. The material preprocessing method based on sensitive content detection according to claim 1 is characterized in that: Receiving material data from a data source in real time, and preprocessing the material data according to the type of the material data to obtain the preprocessed material data, including: Receive material data from data sources in real time; Classifying the material data to obtain different types of data information, wherein the different types of data information include text data information, image data information and video data information; Performing data processing on the text data information to obtain text material data after data processing corresponding to the text data information; Performing data processing on the image data information to obtain image material data after data processing corresponding to the image data information; Data processing is performed on the video data information to obtain frame image material data after data processing corresponding to the video data information.
3. The material preprocessing method based on sensitive content detection according to claim 2 is characterized in that: Data processing is performed on the text data information to obtain text material data corresponding to the text data information after data processing, including: Performing data cleaning processing on the text data information to remove irrelevant information in the text data information and obtain cleaned text data information, wherein the irrelevant information includes HTML tags, special characters and URL links; Using a word segmentation tool to segment the cleaned text data information to obtain independent words and phrases corresponding to the cleaned text data information; Performing stop word extraction and unified format conversion on the independent words and phrases corresponding to the cleaned text data information, obtaining the text data information after the format conversion, and using the text data information after the format conversion as the target text data information; The target text data information is the text material data after data processing corresponding to the text data information.
4. The material preprocessing method based on sensitive content detection according to claim 2 is characterized in that: Performing data processing on the image data information to obtain image material data corresponding to the image data information after data processing includes: Performing noise reduction processing on the image data information to obtain the image data information after the noise reduction processing; Performing grayscale adjustment on the image data information after the noise reduction processing to obtain the image data information after the grayscale adjustment; Performing image segmentation on the grayscale-adjusted image data information to obtain image data blocks corresponding to different areas; Performing feature extraction on the image data blocks to obtain key features corresponding to each image data block, wherein the key features include color features, shape features, and texture features; The key feature corresponding to each image data block is the image material data after data processing corresponding to the image data information.
5. The material preprocessing method based on sensitive content detection according to claim 4 is characterized in that: Performing grayscale adjustment on the image data information after the noise reduction process to obtain the image data information after the grayscale adjustment includes: Extracting the grayscale value corresponding to the target pixel block; Extracting each pixel block contained in the image data information after the noise reduction process, and taking each pixel block as a target pixel block; Extracting a pixel block connected to each pixel block, and taking the pixel block connected to each pixel block as a first selected pixel block; Extract the grayscale value of the first selected pixel block, and use the grayscale value of the first selected pixel block to obtain a first grayscale coefficient; wherein the first grayscale coefficient is obtained by the following formula: Among them, h 01 Represents the first grayscale coefficient; H m represents the grayscale value corresponding to the target pixel block; n represents the number of pixel blocks connected to the target pixel block; H 01i represents the gray value corresponding to the i-th pixel block connected to the target pixel block; exp represents the exponential function symbol with e as the base; Extracting pixel blocks connected to each first selected pixel block, excluding the target pixel block from the pixel blocks connected to the first selected pixel block, and taking the pixel blocks connected to each first selected pixel block excluding the target pixel block as second selected pixel blocks; Extract the gray value of the second selected pixel block, and use the gray value of the second selected pixel block to obtain a second gray coefficient; wherein the second gray coefficient is obtained by the following formula: Among them, h 02 represents the second grayscale coefficient; m represents the number of second selected pixel blocks corresponding to each first selected pixel block; H 02i represents the gray value corresponding to the i-th second selected pixel block; K represents the first adjustment coefficient; wherein the first adjustment coefficient is obtained by the following formula: Wherein, m represents the number of second selected pixel blocks corresponding to each first selected pixel block; H 02i H represents the gray value corresponding to the i-th second selected pixel block; x Indicates the grayscale value of the first selected pixel block corresponding to the second selected pixel block; Performing grayscale adjustment on the target pixel block by using the grayscale value corresponding to the target pixel block in combination with the first grayscale coefficient and the second grayscale coefficient; All target pixel blocks are traversed, and grayscale adjustment is performed on each target pixel block to obtain image data information after grayscale adjustment.
6. The material preprocessing method based on sensitive content detection according to claim 5 is characterized in that: The grayscale value corresponding to the target pixel block is combined with the first grayscale coefficient and the second grayscale coefficient to adjust the grayscale of the target pixel block, including: Extracting the grayscale value corresponding to the target pixel block; extracting a first grayscale coefficient and a second grayscale coefficient; The target grayscale value corresponding to the target pixel block is obtained by using the grayscale value corresponding to the target pixel block and the first grayscale coefficient and the second grayscale coefficient; wherein the target grayscale value is obtained by the following formula: Among them, H f Indicates the target gray value; h 01 Represents the first grayscale coefficient; h 02 Represents the second grayscale coefficient; H m Indicates the grayscale value corresponding to the target pixel block; The grayscale value of the target pixel block is adjusted according to the target grayscale value to obtain the target pixel block after the grayscale value is adjusted.
7. The material preprocessing method based on sensitive content detection according to claim 4 is characterized in that: Performing image segmentation on the grayscale-adjusted image data information to obtain image data blocks corresponding to different regions includes: Extracting image data information after grayscale adjustment; According to the grayscale value contained in the grayscale-adjusted image data information, a first grayscale threshold and a second grayscale threshold are obtained; wherein the first grayscale threshold and the second grayscale threshold are obtained by the following formula: Among them, H y01 Represents the first grayscale threshold; H y02 Represents the second grayscale threshold; H 01 represents the initial first grayscale threshold; H 02 represents the initial second grayscale threshold; k represents the number of pixel blocks whose target grayscale value is lower than the grayscale value before grayscale adjustment; r represents the number of pixel blocks whose target grayscale value is higher than the grayscale value before grayscale adjustment; H mj represents the grayscale value of the pixel block whose jth target grayscale value is lower than the grayscale value before grayscale adjustment; H fj represents the target grayscale value of the pixel block whose jth target grayscale value is lower than the grayscale value before grayscale adjustment; H mt represents the grayscale value of the pixel block whose tth target grayscale value is higher than the grayscale value before grayscale adjustment; H ft represents the target grayscale value of the pixel block whose t-th target grayscale value is higher than the grayscale value before grayscale adjustment; Comparing the grayscale value of each pixel block in the grayscale-adjusted image data information with the first grayscale threshold and the second grayscale threshold; Pixel blocks with grayscale values lower than the first grayscale threshold form one or more first image areas; Pixel blocks whose grayscale values are not lower than the first grayscale threshold but lower than the second grayscale threshold form one or more second image regions; The pixel blocks whose grayscale values are not lower than the second grayscale threshold form one or more third image areas; Image segmentation is performed according to the areas enclosed by the first image area, the second image area and the third image area to obtain image data blocks corresponding to different areas.
8. The material preprocessing method based on sensitive content detection according to claim 2, characterized in that: Performing data processing on the video data information to obtain frame image material data corresponding to the video data information after data processing includes: Performing frame processing on the video data information to obtain a frame image data set corresponding to the video data information; A frame interval quantity range is obtained for the frame image data; wherein the upper limit quantity value and the lower limit quantity value corresponding to the frame interval quantity range are obtained by the following formula: Among them, M up and M down represents the upper limit and lower limit of the frame interval range; x represents the number of unit times corresponding to the video data reception, and the unit time is 2s-5s; M c Indicates the preset frame number reference value; M p Indicates the average number of frames per unit time; M i M represents the number of frames of frame data contained in the received video data information corresponding to the i-th unit time; d0 and M u0 represents the upper limit value and the lower limit value of the preset initial frame interval quantity range; s represents the second adjustment coefficient, and the second adjustment coefficient is obtained by the following formula: Where, s represents the second adjustment coefficient; M z represents the median value of the number of frames contained in the received video data corresponding to n unit time; M max and M min Indicates the maximum and minimum number of frames contained in the received video data corresponding to a unit time; Randomly setting a frame interval number that meets the frame interval number range according to the frame interval number range, and extracting frame images from the frame image data set according to the frame interval number to obtain target frame image data; Performing noise reduction processing on the target frame image data to obtain the target frame image data after the noise reduction processing; Performing grayscale adjustment on the target frame image data after the noise reduction processing to obtain the target frame image data after the grayscale adjustment; Performing image segmentation on the grayscale-adjusted target frame image data to obtain image data blocks corresponding to the target frame image data corresponding to different regions; Performing feature extraction on the image data blocks corresponding to the target frame image data to obtain key features contained in the image data blocks corresponding to each target frame image data, wherein the key features include color features, shape features and texture features; The key feature corresponding to the image data block corresponding to each target frame image data is the image material data after data processing corresponding to the image data information.
9. The material preprocessing method based on sensitive content detection according to claim 1, characterized in that: The data source is evaluated for material quality according to the material data rejection status, and a material receiving shielding process is performed on the data source whose material quality does not meet the evaluation requirements, including: Extract the rejection status information of the material data of each data source; The material data quality evaluation parameter corresponding to each data source is obtained according to the rejection status information of the material data; wherein the material data quality evaluation parameter is obtained by the following formula: Wherein, Q represents the material data quality evaluation parameter; p represents the number of sending unit time experienced by the data source to send the material data, and the sending unit time is 1s; P i represents the proportion of sensitive information data to the total data sent in the i-th sending unit time; P c represents a preset data ratio threshold; z represents the number of transmission unit times during which the data ratio of sensitive information data in p unit times to the total data transmitted does not exceed the preset data ratio threshold; y represents the number of transmission unit times during which the data ratio of sensitive information data in p unit times to the total data transmitted exceeds the preset data ratio threshold; P j represents the data ratio of the jth sensitive information data to the total data sent, which does not exceed the preset data ratio threshold, and corresponds to the data ratio of the sensitive information data to the total data sent per sending unit time; P g Indicates the data ratio of the g-th sensitive information data to the total data sent, corresponding to the sending unit time when the data ratio of the g-th sensitive information data to the total data sent exceeds the preset data ratio threshold; Comparing the material data quality evaluation parameter with a preset parameter threshold; When the material data quality evaluation parameter is lower than a preset parameter threshold, it is determined that the material quality of the data source does not meet the evaluation requirements; Data sources whose material quality does not meet the evaluation requirements are blocked from receiving materials.
10. A material preprocessing system based on sensitive content detection, characterized in that: The material preprocessing system based on sensitive content detection includes: A material data receiving module, used to receive material data from a data source in real time, and pre-process the material data according to the type of the material data to obtain pre-processed material data; A sensitive information identification module, used to identify sensitive information of the material data through a sensitive information identification model, and remove the material data with sensitive information; wherein the sensitive information identification model includes a sensitive information identification model for image identification and a sensitive information identification model for text identification; The material quality assessment module is used to assess the material quality of the data source according to the material data rejection status, and to perform material reception shielding processing on the data source whose material quality does not meet the assessment requirements.
Citation Information
Patent Citations
Sensitive information risk assessment method and system and storage medium
CN113849760A
Video later-stage particle noise removal method based on image processing
CN115908154A
Live broadcast information content auditing system and method based on data processing
CN116233479A
Fingerprint identification access control method and system
CN117576819A
Network information detection method and system based on big data
CN118300851A
