Image material automatic detection method, terminal device and storage medium

Abstract

The application belongs to the field of communication, and particularly relates to an image material automatic detection method, a terminal device and a storage medium. In the method, each platform auditing rule is atomized and disassembled into a unified standard minimum judgment unit, effectively solving the technical problems of different platform rule expression differences and non-uniform quantitative threshold values. By generating the intersection of the multi-platform rule atoms as a unified detection standard, the auditing requirements of all target platforms can be met at one time, and the inefficient process of separately detecting and repeatedly adjusting and optimizing the material for each platform in the prior art is abandoned.

Description

Technical Field

[0001] This application belongs to the field of communications, and in particular relates to an automatic detection method for image materials, a terminal device, and a storage medium. Background Technology

[0002] With the rapid development of the digital marketing industry, the popularization of AI-generated technology has led to an exponential increase in the output of advertising video materials. In order to expand the reach of their advertising campaigns, advertisers generally adopt the strategy of "one material for multiple platforms".

[0003] Before publishing video footage on a platform, users typically need to inspect the footage to ensure it complies with the platform's rules, thus avoiding removal due to violations after publication. However, current technology usually inspects footage separately for each platform. Since the rules of different platforms are not consistent, passing the inspection of one platform does not guarantee that the footage will pass the inspection of other platforms. Therefore, it is necessary to inspect the footage according to the rules of each target platform and make corresponding adjustments and optimizations before the footage can be published, which is very inefficient. Summary of the Invention

[0004] In view of this, embodiments of this application provide an automatic image material detection method, terminal device, and storage medium, which can solve the above-mentioned technical problems.

[0005] A first aspect of this application provides an automatic image material detection method, the method comprising: S1: Atomize all the review rules of each platform to obtain several rule atoms; S2: Determine the mapping parameters for each rule atom, and then integrate each rule atom and its corresponding mapping parameters into a set of rule atoms for the platform; S3: Whenever a user uploads video footage, determine the target platform to which the video footage needs to be uploaded; S4: Retrieve the set of rule atoms corresponding to each target platform; S5: Determine the intersection of the sets of all rule atoms to obtain the target set; S6: Use the rules in the target set to detect image materials so that the images that pass the detection can simultaneously meet the review rules of all target platforms.

[0006] A second aspect of this application provides a terminal device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the automatic image material detection method.

[0007] A third aspect of this application provides a terminal-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the automatic image material detection method.

[0008] The beneficial effects of this application embodiment compared with the prior art are as follows: The method provided by the present invention includes atomizing and decomposing all review rules of each platform to obtain several rule atoms; determining the mapping parameters of each rule atom, and then integrating each rule atom and the corresponding mapping parameters into a set of rule atoms for that platform; whenever a user-uploaded image material is received, determining the target platform to which the image material needs to be uploaded; retrieving the rule atom sets corresponding to each target platform; determining the intersection of each set of rule atoms to obtain a target set; using each rule atom in the target set to detect the image material, so that the detected image can simultaneously meet the review rules of all target platforms; In this application, by atomizing and decomposing the review rules of each platform into the smallest judgment unit of a unified standard, the technical problem of differences in rule expression and inconsistent quantification thresholds between different platforms is effectively solved. By generating the intersection of rule atoms of multiple platforms as a unified detection standard, it is possible to simultaneously meet the review requirements of all target platforms with a single detection, abandoning the inefficient process of requiring separate detection and repeated adjustment and optimization of materials for each platform in the prior art. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram illustrating the implementation process of the automatic image material detection method provided in the embodiments of this application; Figure 2 This is a schematic diagram illustrating the implementation process of the automatic image material detection method provided in the embodiments of this application; Figure 3 This is a schematic diagram of the terminal device provided in the embodiments of this application. Detailed Implementation

[0011] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0012] To illustrate the technical solution described in this application, specific embodiments are provided below.

[0013] Figure 1 This application illustrates an automatic image material detection method according to Embodiment 1, the method comprising: S1: Atomize all the review rules of each platform to obtain several rule atoms; S2: Determine the mapping parameters for each rule atom, and then integrate each rule atom and its corresponding mapping parameters into a set of rule atoms for the platform; S3: Whenever a user uploads video footage, determine the target platform to which the video footage needs to be uploaded; S4: Retrieve the set of rule atoms corresponding to each target platform; S5: Determine the intersection of the sets of all rule atoms to obtain the target set; S6: Use the rules in the target set to detect image materials so that the images that pass the detection can simultaneously meet the review rules of all target platforms.

[0014] In this embodiment, as Figure 2 As shown, this method can be executed on a terminal device, which can be an independent physical server or terminal, or a server cluster consisting of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud servers, cloud databases, cloud storage, and CDN. The terminal device communicates with various user terminals, can receive image materials sent by user terminals, and then detect the image materials. The terminal device also connects to the platform terminals of various platforms, and then automatically uploads the detected image materials to the respective platforms.

[0015] This application effectively solves the technical challenges of differing rule expressions and inconsistent quantification thresholds across different platforms by atomically breaking down the review rules of various platforms into the smallest, unified judgment units. By generating the intersection of rule atoms from multiple platforms as a unified detection standard, a single detection can simultaneously meet the review requirements of all target platforms. This eliminates the inefficient process of requiring separate detection and repeated adjustments to materials for each platform, as is common in existing technologies. Furthermore, it fundamentally avoids the risk of material removal and account bans due to missed detections, and significantly reduces material waste caused by over-detection, thereby greatly improving the overall efficiency and economic benefits of advertising.

[0016] As a preferred embodiment, all review rules for each platform are atomically decomposed to obtain several rule atoms, including: S1: Retrieve several set rule category names; S2: For each platform, obtain the platform's review rules document; S3: Select a rule category name as the comparison name; S4: Compare the comparison items with the text of the rule document one by one to extract the rule items corresponding to the comparison items from the text of the rule document, i.e., rule atoms; S5: Take another rule category name as the comparison name, and execute steps S4 to S5 until the rule atom corresponding to each rule category name is determined.

[0017] In this embodiment, the rule category name is a predefined, standardized classification system covering all types of advertising violations. It serves as a unified benchmark template for cross-platform rule extraction, ensuring that rules from different platforms can be compared and integrated on the same dimension. The comparison name is a specific rule category name currently used for semantic matching with platform rule documents, serving as search keywords in the rule extraction process. The rule atom is an indivisible, smallest executable judgment unit corresponding to a single violation feature. Each rule atom contains only one violation judgment condition and a corresponding quantification threshold. The semantic similarity matching method is a text similarity calculation method based on a pre-trained language model, which can identify content with different expressions but the same semantics, solving the problem of differences in rule expressions across different platforms.

[0018] In this embodiment, for example, for mainstream advertising platforms such as Toutiao, Kuaishou, and Tencent Ads, the rules for four core violation categories—pornography, gambling, sensitive terms, and absolute terms—are atomically decomposed. The implementation environment is a cloud server equipped with an NVIDIA A10 graphics card, and the semantic parsing is performed using a BERT-base-chinese pre-trained model, with a rule extraction accuracy target of ≥95%.

[0019] Step S1: Retrieve the names of the set rule categories The system retrieves the four primary rule categories and their corresponding 16 secondary subcategories from a pre-defined global rule category library. This category library was developed by advertising compliance experts based on the review rules of 12 mainstream platforms over the past five years, and includes 12 primary categories, 64 secondary categories, and 256 tertiary categories, comprehensively covering all common advertising violation scenarios. The specific categories retrieved in this instance include: pornography violations (including four subcategories: nudity, vulgar actions, and sexual innuendo), gambling violations, sensitive violations (including four subcategories: sensitive figures, sensitive events, and sensitive symbols), and violations involving absolute terms (including three subcategories: superlative terms and exaggerated advertising terms).

[0020] Step S2: Obtain the platform's review rules document The system automatically captures and parses the "Toutiao Advertising Material Review Standard V2026.05", "Kuaishou Promotion Content Review Standard V2026.05", and "Tencent Advertising Review Standard V2026.05" via the official open platform API interface. First, it converts unstructured documents such as PDF and Word documents into UTF-8 encoded plain text format. Then, it uses a document structure parsing algorithm to identify hierarchical structures such as titles, chapters, and paragraphs, assigning a unique paragraph ID to each paragraph and recording its position information in the original document. For tables contained in the documents, the system automatically converts them into structured data, extracting information such as violation types, judgment conditions, and penalty levels. For images and flowcharts, the system uses OCR technology to recognize the text information, ensuring that no rule content is missed.

[0021] Step S3: Select a rule category name as the comparison name. The system selects "pornography-related violations - nudity" as the first comparison item according to a preset priority order. The system automatically loads the semantic feature vector corresponding to this comparison item, which is generated based on the BERT-base-chinese pre-trained model and has a dimension of 768. Simultaneously, the system loads a thesaurus and related terminology database corresponding to this comparison item, containing over 200 related words such as "nudity," "exposure," "wardrobe malfunction," and "exposed nipples," to improve the accuracy of semantic matching.

[0022] Step S4: Compare the text of the comparison items with the text of the rule document one by one. The system uses cosine similarity to compare the semantic feature vector of the comparison item with the feature vector of each text segment in the rule documents of the three platforms one by one, setting a similarity threshold of 0.85. When the similarity between a text segment and the comparison item exceeds the threshold, the system automatically extracts the violation judgment conditions and quantification parameters from that text segment, generating the corresponding rule atom. For example: extracting "The proportion of skin color pixels in exposed parts of the human body exceeds 15% is judged as pornography" from the Toutiao rule document generates rule atom V001; extracting "The proportion of skin color pixels in exposed parts of the human body exceeds 12% is judged as pornography" from the Kuaishou rule document generates rule atom V002; extracting "The proportion of skin color pixels in exposed parts of the human body exceeds 18% is judged as pornography" from the Guangdiantong rule document generates rule atom V003. Each rule atom contains the following fields: unique global ID, violation type, subcategory, judgment condition, quantification threshold, platform identifier, source document, paragraph ID, creation time, and confidence level. Among them, the confidence level field indicates the accuracy of the rule atom extraction, which is determined by the semantic similarity score and the results of manual verification.

[0023] Step S5: Repeat until all category names have been processed. The system sequentially selects the remaining 15 secondary subcategories, including "Pornography-related violations - Vulgar actions," "Gambling-related violations - Betting keywords," "Sensitive violations - Sensitive identifiers," and "Absolute terminology violations - Superlative terms," ​​as new comparison categories, and repeats step S4. For example, it extracts "Containing keywords such as 'betting' and 'winning money' is considered gambling" (rule atoms T001, T002, T003) and "Using absolute terms such as 'best,' 'first,' and 'top' is considered a violation" (rule atoms T004, T005, T006) from the three platform documents. Ultimately, this embodiment extracts a total of 218 rule atoms from the rule documents of the three platforms, including 72 pornography-related, 65 gambling-related, 53 sensitive, and 28 absolute terminology-related rules. The system automatically performs preliminary deduplication on the extracted rule atoms, merging rule atoms with identical semantics to obtain 189 unique rule atoms. Subsequently, professional compliance personnel manually sampled and verified all rule atoms, achieving an accuracy rate of 96.2%, which met the preset requirements. This laid a solid foundation for the subsequent construction of a platform-specific set of rule atoms and a unified cross-platform testing standard.

[0024] As a preferred embodiment, determining the mapping parameters for each rule atom, and then integrating each rule atom and its corresponding mapping parameters into a set of rule atoms for the platform includes: For each rule atom, context analysis is performed in the document in which it was initially located to determine the corresponding rule parameters. The rule parameters are then mapped to the rule atom, forming a mapping binding between the rule parameters and the rule atom. Each rule atom and its bound mapping parameters are integrated into a set to obtain a set of rule atoms.

[0025] In this embodiment, mapping parameters are a set of interface parameters that convert abstract rule atoms into executable instructions for the detection engine, including core information such as detection engine type, model version, judgment threshold, and execution conditions; context analysis is a technical process that extracts implicit information such as the scope of application, exceptions, and quantification standards of rules by parsing the context of rule atoms in the original document; and the rule atom set is a structured storage unit for all effective rule atoms and their bound mapping parameters on a single platform, which is the basic data structure for cross-platform rule comparison.

[0026] In this embodiment, building upon the 189 unique rule atoms extracted from the three major platforms—Toutiao, Kuaishou, and Tencent Ads—in the previous stage, mapping parameter binding and the construction of platform-specific rule atom sets are completed. The implementation environment adopts the TensorFlow 2.15 deep learning framework, integrating the YOLOv8 object detection engine, the PaddleOCR text recognition engine, and the Baidu Wenxin Yiyan semantic analysis engine.

[0027] Mapping parameter determination and binding process: The system traverses each rule atom, performs context window analysis (window size is 3 paragraphs of text before and after) in its original source document, extracts all parameter information related to rule execution, and establishes a one-to-one mapping relationship: Visual rule atom mapping parameter binding: Taking rule atom V001 (Toutiao: Skin color pixels in exposed human body parts exceeding 15% are judged as pornographic) as an example: The system analyzes the context in the Toutiao rule document and extracts two implicit conditions: "Only applicable to images where the main body of the person accounts for more than 50%" and "Skin color of the face and hands is not included in the statistics." The generated mapping parameters include: detection engine type = image semantic segmentation engine, model version = UNet_v2.3, target category = human skin, statistical region = human body region excluding face and hands, judgment threshold = 15%, and precondition = main body of the person > 50%. These parameters are uniquely mapped and bound to rule atom V001.

[0028] Text-based rule atomic mapping parameter binding: Taking rule atomic T001 (Kuaishou: keywords such as "gambling" and "betting" are considered gambling-related) as an example: Context analysis extracts two conditions: "containing homophonic variations is also considered a violation" and "appearing in subtitles and audio-to-text is also applicable." The generated mapping parameters include: detection engine type = keyword matching engine + homophonic conversion engine, keyword library version = gambling-related keyword library_v2026.05, matching mode = fuzzy matching + homophonic matching, applicable scope = image text / video subtitles / audio-to-text, judgment threshold = appearing once is a violation.

[0029] Audio rule atom mapping parameter binding: Taking rule atom A001 (Guangdiantong: containing violent sound effects is judged as a violation) as an example: Context analysis extracts the quantization condition of "volume exceeding 60 decibels and duration exceeding 2 seconds". The generated mapping parameters include: detection engine type = audio classification engine, model version = AudioCLIP_v1.2, target category = violent sound effects, volume threshold = 60 decibels, duration threshold = 2 seconds.

[0030] Rule Atom Set Integration Process: The system integrates all rule atoms bound to mapping parameters by platform dimension: An independent rule atom set is created for each platform, named in the format "Platform Name_Rule Atom Set_Version Number". Each set contains the following fields: Platform Identifier, Version Number, Effective Date, Expiration Date, Total Number of Rule Atoms, and Rule Atom List. Each entry in the rule atom list contains: Rule Atom ID, Violation Type, Judgment Condition, Mapping Parameters, Confidence Level, and Source Document. The system automatically performs integrity checks on the sets to ensure that each rule atom is bound to complete mapping parameters without missing items.

[0031] Ultimately, this embodiment generated three rule atom sets: Toutiao_v2026.05, Kuaishou_v2026.05, and Guangdiantong_v2026.05, containing a total of 189 rule atoms and 1134 mapping parameter items. Verified by professional compliance personnel, the mapping parameter accuracy reached 97.5%, laying the foundation for subsequent cross-platform rule intersection calculations and unified detection.

[0032] As a preferred embodiment, the intersection of the sets of rules atoms is determined to obtain the target set, which includes: S51: Generate a spare set, wherein the initial state of the spare set is an empty set; S52: Select a rule category name as the target directory; S53: Screen out the regular atoms corresponding to the target list from each set of regular atoms; S54: For each screened regular atom, determine the parameter range of the corresponding mapping parameters; S55: Find the intersection of the parameter ranges of the mapping parameters of each regular atom to obtain the target parameter range; S56: Generate a regular atom that is the same as the sieved regular atom, i.e., the target atom; S57: Bind the target parameter range to the target atom mapping and include the target atom in the spare set; S58: Select another rule category name as the target list, and execute steps S51 to S58 until all the target atoms corresponding to all rule category names are included in the standby set, thus obtaining the target set.

[0033] In this embodiment, the mapping parameter range is the value range of the quantization judgment condition in the rule atom, which explicitly indicates that "when the feature value falls into this range, the corresponding platform will judge it as a violation"; the parameter range intersection is the common part of the quantization parameter range of the same violation feature for multiple platforms, representing the violation judgment standard that is consistently recognized by all target platforms; the target atom is a unified rule atom generated based on the rule intersection of multiple platforms, and its judgment standard is applicable to all target delivery platforms; the target set is a structured set of all target atoms, which is the core execution standard for unified detection of cross-platform image materials.

[0034] In this embodiment, following the three rule atomic sets generated in the previous stage—Toutiao_v2026.05, Kuaishou_v2026.05, and Guangdiantong_v2026.05—the user selects to simultaneously distribute the same batch of materials to these three platforms. The system performs multi-platform rule intersection calculations to generate a unified target detection set, ensuring that materials that pass the detection meet the core compliance requirements of all platforms simultaneously.

[0035] Step S51: Generate the initial spare set The system creates an empty spare set in memory to temporarily store the generated target atoms. This set is stored using a hash table structure, with the rule category name as the key and the target atom as the value, supporting fast query and insertion operations.

[0036] Step S52: Select the first target list The system selects "pornography-related violations - nudity" as the first target list according to the preset priority order of violation types, and loads the global category identifier corresponding to this category.

[0037] Step S53: Filter the rule atoms of the corresponding category The system filters out rule atoms belonging to the "pornographic violation - nudity" category from the rule atom sets of the three platforms, resulting in three rule atoms: V001 from Toutiao, V002 from Kuaishou, and V003 from Tencent Ads.

[0038] Step S54: Determine the mapping parameter range for each regular atom. The system extracts the mapping parameters bound to each rule atom and converts them into a calculable numerical range: Toutiao V001: Skin color pixel ratio of exposed human body parts > 15%; Kuaishou V002: Skin color pixel ratio of exposed human body parts > 12%; Guangdiantong V003: Skin color pixel ratio of exposed human body parts > 18%.

[0039] Step S55: Calculate the intersection of parameter ranges The system performs an intersection operation on the three parameter ranges. Since the common part of the three ranges is "the proportion of skin color pixels in exposed human body parts > 18%", the target parameter range is determined to be > 18%. This range represents the standard that all three platforms agree is a violation; as long as the material features fall within this range, it will inevitably be judged as a violation by all platforms.

[0040] Step S56: Generate the corresponding target atoms The system generates a target atom with the same semantics as the original rule atom, named TV001, whose violation type, sub-category, and judgment logic are consistent with the original rule atom.

[0041] Step S57: Bind target parameters and add them to the spare set The system establishes a unique mapping and binding between the calculated target parameter range >18% and the target atom TV001, and then adds TV001 to the standby set.

[0042] Step S58: Process all rule category names in a loop. The system sequentially selects the remaining 15 secondary subcategories, including "Pornography-related Violations - Vulgar Actions," "Gambling-related Violations," "Sensitive Violations - Sensitive Symbols," and "Absolute Terms Violations - Superlative Terms," ​​as new target lists, and repeats steps S53 to S57. For example, for the "Gambling-related Violations" category, the rule atomic parameter range for all three platforms is "one occurrence constitutes a violation," therefore the intersection remains "one occurrence constitutes a violation," generating target atom TT001.

[0043] Ultimately, this embodiment calculated and generated a target set containing 126 target atoms from 189 original rule atoms. For rule atoms that exist only on certain platforms, the system marked them as "platform-specific rules," which were then used as supplementary detection items in subsequent testing and separately notified to the user in the detection report. Verified by professional compliance personnel, this target set ensures that the detected material achieves a 100% pass rate for core violations across all target platforms, providing a reliable standard for subsequent unified testing.

[0044] In a preferred embodiment, the method further includes the following step between step S5 and step S6: Identify the target object for each target atom, where the target object includes video frames, images, and text.

[0045] Image materials were detected using the atoms of each rule in the target set, including: Identify all measurement points affecting the material and determine the detection object for each measurement point, including video frames, images, and text; For each measuring point, determine the number of objects to be detected at that measuring point; If the number of objects to be detected is 1, the target atom that matches the corresponding target object will be retrieved for detection, and the qualification of the test point will be determined based on the obtained detection score. If the number of objects to be detected is greater than 1, then for each object to be detected, the corresponding target object and the target atoms that match the object to be detected are retrieved and detected to obtain a detection score that is the same as the number of objects to be detected. Then, based on the detection scores, it is determined whether the test point is qualified.

[0046] In this embodiment, the target object refers to the type of material content that each target atom can detect, which is the basis for matching the rule atom with the material content. In this embodiment, it includes three categories: video frames, images, and text. The test point refers to an independent content unit in the video material that needs to be checked for compliance. It is the smallest unit for detection. A complete piece of material can contain multiple test points. The detection object refers to the type of material content actually contained in a single test point. A test point can contain one or more detection objects. The detection score is a quantitative score given by the detection engine to the degree of violation of the material content. The value ranges from 0 to 100. The higher the score, the greater the possibility of violation. Usually, a score of 60 is set as the passing threshold.

[0047] In this embodiment, following the target set containing 126 target atoms generated in the previous stage, the user uploaded three types of typical image materials for detection: a pure image, a combined image and text image, and a 30-second short video. The system first completes the target object binding of all target atoms, and then performs comprehensive detection of multiple detection objects based on the measurement points.

[0048] Between steps S5 and S6: Determining the target object of the target atom. The system iterates through all 126 target atoms in the target set and binds a corresponding target object to each target atom according to its violation type and judgment logic: the target objects of all visual target atoms (such as TV001 human nudity detection and TV004 sensitive label detection) are video frames and images; the target objects of all text target atoms (such as TT001 gambling-related keyword detection and TT004 absolute term detection) are text; the target objects of all audio target atoms are audio (not elaborated in this embodiment). For example, the target object of target atom TV001 (human nudity part skin color pixel ratio > 18% is judged as violation) is bound to "video frame, image"; the target object of target atom TT001 is bound to "text".

[0049] Step S6: Comprehensive detection of multiple detection objects based on measurement points The system identifies measurement points and determines the detection objects for the three types of uploaded materials: Pure image material: contains only 1 measurement point, the detection object is the image, and the number is 1; Image and text combination material: contains only 1 measuring point, the detection objects are images and text, and the number is 2; 30-second short video footage: 30 video frames are extracted at 1 frame per second as 30 independent measurement points. The detection objects of each measurement point are the video frame and the corresponding subtitle text of that frame, with a total of 2.

[0050] Single-object measurement point detection: Taking a test point with only an image as the target, the number of detection objects is 1, containing only the image: the system retrieves all target atoms (72 in total) whose target object is "image", and detects the image sequentially. Among them, the detection score of target atom TV001 is 42, the detection score of target atom TV004 is 18, and the detection scores of the remaining target atoms are all below 30. Since all detection scores do not exceed the passing threshold of 60 points, the test point is judged to be qualified.

[0051] Comprehensive testing of multiple testing objects and measuring points: Taking the measurement points of combined image and text materials as an example, the number of detection objects is 2, including both images and text: For image detection targets, all target atoms with "image" as the target object are retrieved for detection, and the highest detection score obtained is 35; For the text detection object, all target atoms whose target object is "text" are invoked for detection, wherein the detection score of the target atom TT004 (absolute term detection) is 78, which exceeds the qualification threshold; The system adopts "one-vote veto system" for comprehensive judgment. As long as the detection score of any detection object exceeds the threshold, the measuring point is determined as unqualified. Therefore, the measuring point of the image-text combined material is finally determined as unqualified.

[0052] Taking the 15th frame measuring point of the short-video material as an example, its detection objects are video frames and subtitle text: The highest score obtained from video frame detection is 28; The highest score obtained from subtitle text detection is 45; All detection scores do not exceed the threshold, so the measuring point is determined as qualified.

[0053] Finally, this embodiment completes the detection of 32 measuring points in total for three types of materials, among which 31 measuring points are determined as qualified and 1 measuring point is determined as unqualified. The system generates a detailed detection report, marking the position, violation type and specific reason of the unqualified measuring point, providing a basis for users to modify materials.

[0054] As a preferred embodiment, if an unqualified measuring point is identified, the method further comprises: For each unqualified measuring point, parameter adjustment is performed on each detection object of the unqualified measuring point based on the mapping parameter of the corresponding target atom, the unqualified measuring point is detected again, and then it is determined whether the unqualified measuring point has been converted into a qualified measuring point. This step is repeatedly executed until the unqualified measuring point is converted into a qualified measuring point.

[0055] In this embodiment, parameter adjustment refers to the technical process of automatically adjusting the feature value of the material content according to the mapping parameters of the target atom to make it deviate from the violation parameter range, and the adjustment process retains the original core advertising information of the material; iterative detection refers to the cyclic process of multiple parameter adjustments and detections on unqualified measuring points until the measuring point is qualified or the maximum number of adjustments is reached; safe adjustment interval refers to a preset parameter adjustment range, which ensures that the adjusted material can still maintain the original advertising expression effect and avoids excessive modification; adjustment priority refers to the adjustment sequence set according to the influence degree of parameter modification on the material effect, and the adjustment operation with the smallest influence on the advertising effect is preferentially performed.

[0056] In this embodiment, following the detection results of the previous stage, two unqualified measuring points are identified in total: one is the measuring point of the image-text combined material (the text detection object is unqualified), and the other is the measuring point of the 22nd frame of the short-video material (the video frame detection object is unqualified). The system automatically starts the parameter adjustment and iterative detection process, and can complete material correction without manual intervention.

[0057] Automatic adjustment and iterative detection process for non-conforming measuring points: Step 1: Extraction of information on non-conforming measurement points The system first extracts detailed information from two non-compliant measurement points: Test Point 1 (Text and Image Combination Material): The detection object is text, the violation target atom is TT004 (absolute term detection), the detection score is 78 points, the violation content is "the most effective weight loss product on the entire internet", and the mapping parameter threshold is 60 points. Test point 2 (frame 22 of the short video): The detection object is a video frame, the violation target atom is TV001 (human nudity detection), the detection score is 65 points, the violation feature is that skin color pixels account for 19%, and the mapping parameter threshold is >18%. Step 2: Adjustment and Iterative Detection of Measurement Point Parameters for Single Detection Object Taking measurement point 2 as an example, it only contains one detection object: video frame. The system loads the mapping parameters and safe adjustment range of the target atom TV001, determining the adjustable parameter to be "skin color pixel ratio" and the safe adjustment range to be 10% to 18%. Following the principle of minimizing adjustment, the system automatically performs edge cropping, reducing the percentage of skin tone pixels from 19% to 17.5%. The target atom TV001 was tested again, and a new detection score of 48 was obtained, which is below the passing threshold of 60. Once the measurement point has been determined to be a qualified measurement point, the adjusted video frame is saved, and the iteration process ends. Step 3: Adjustment and Iterative Detection of Measurement Point Parameters for Multiple Detection Objects Taking test point 1 as an example, it contains two detection objects: images and text. Only the text detection object fails: The system loads the mapping parameters and safety adjustment range of the target atom TT004, determines the adjustable parameters to be "textual semantics", and prioritizes the adjustment method of synonym substitution. First adjustment: Replace the offending keyword "most" with "very," generating new text "The most effective weight loss product on the entire internet." The target atom TT004 was tested again, and the result was a score of 32, which is below the passing threshold. At the same time, the image was re-examined, and its detection score was confirmed to be 35 points, maintaining a qualified status. Based on the comprehensive assessment, the measuring point has been converted into a qualified measuring point. The adjusted image and text combination materials are saved, and the iteration process ends. Step 4: Handling Abnormal Situations If a measurement point still fails to meet the qualification standard after 5 iterations of adjustment, the system will automatically terminate the automatic adjustment process, mark the measurement point as "requiring manual processing," and push it to the manual review platform. The platform will display all adjustment records and test results for that measurement point, providing a reference for manual modification.

[0058] Ultimately, both non-compliant measurement points in this embodiment were converted into compliant measurement points through automatic parameter adjustment, achieving a 100% success rate. Compared to traditional manual modification methods, the processing time for a single measurement point was reduced from an average of 15 minutes to less than 10 seconds, significantly improving material processing efficiency while ensuring the advertising effectiveness of the materials.

[0059] Embodiment 2 of this application provides a terminal device, including a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, it causes the processor to perform the steps of the automatic image material detection method, including: S1: Atomize all the review rules of each platform to obtain several rule atoms; S2: Determine the mapping parameters for each rule atom, and then integrate each rule atom and its corresponding mapping parameters into a set of rule atoms for the platform; S3: Whenever a user uploads video footage, determine the target platform to which the video footage needs to be uploaded; S4: Retrieve the set of rule atoms corresponding to each target platform; S5: Determine the intersection of the sets of all rule atoms to obtain the target set; S6: Use the rules in the target set to detect image materials so that the images that pass the detection can simultaneously meet the review rules of all target platforms.

[0060] Embodiment 3 of this application provides a terminal-readable storage medium storing a computer program. When the computer program is executed by a processor, it causes the processor to perform the steps of the automatic image material detection method, including: S1: Atomize all the review rules of each platform to obtain several rule atoms; S2: Determine the mapping parameters for each rule atom, and then integrate each rule atom and its corresponding mapping parameters into a set of rule atoms for the platform; S3: Whenever a user uploads video footage, determine the target platform to which the video footage needs to be uploaded; S4: Retrieve the set of rule atoms corresponding to each target platform; S5: Determine the intersection of the sets of all rule atoms to obtain the target set; S6: Use the rules in the target set to detect image materials so that the images that pass the detection can simultaneously meet the review rules of all target platforms.

[0061] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0062] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0063] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0064] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0065] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance. It should also be understood that although the terms "first," "second," etc., are used in the text to describe various elements in some embodiments of this application, these elements should not be limited by these terms. These terms are merely used to distinguish one element from another. For example, a first table may be named a second table, and similarly, a second table may be named a first table, without departing from the scope of the various described embodiments. Both the first table and the second table are tables, but they are not the same table.

[0066] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0067] The automatic image material detection method provided in this application embodiment can be applied to terminal devices such as mobile phones, tablets, wearable devices, in-vehicle devices, augmented reality (AR) / virtual reality (VR) devices, laptops, ultra-mobile personal computers (UMPCs), netbooks, and personal digital assistants (PDAs). This application embodiment does not impose any restrictions on the specific type of terminal device.

[0068] For example, the terminal device may be a station (STAION, ST) in a WLAN, a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA) device, a handheld device with wireless communication capabilities, a computing device or other processing device connected to a wireless modem, an in-vehicle device, a vehicle networking terminal, a computer, a laptop computer, a handheld communication device, a handheld computing device, a satellite wireless device, a wireless modem card, a set-top box (STB), customer premises equipment (CPE), and / or other devices used for communication over a wireless system, as well as next-generation communication systems, such as mobile terminals in 5G networks or mobile terminals in future evolved Public Land Mobile Network (PLMN) networks.

[0069] As an example and not a limitation, when the terminal device is a wearable device, the term "wearable device" can also refer to any device that utilizes wearable technology to intelligently design and develop everyday wearables, such as glasses, gloves, watches, clothing, and shoes. Wearable devices are portable devices worn directly on the body or integrated into a user's clothing or accessories. Wearable devices are not merely hardware devices; they achieve powerful functions through software support, data interaction, and cloud interaction. Broadly defined, wearable smart devices include those with comprehensive functions, large sizes, and the ability to perform complete or partial functions without relying on a smartphone, such as smartwatches or smart glasses, as well as those focused on a specific application function that require interaction with other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0070] Figure 3 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. For example... Figure 3 As shown, the terminal device in this embodiment includes: at least one processor and a memory, wherein the memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the steps in the above-described embodiments of the automatic image material detection method, for example... Figure 1 Steps S1 to S6 are shown.

[0071] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that... Figure 3 This is merely an example of a terminal device and does not constitute a limitation on the terminal device. It may include more or fewer components than shown, or combine certain components, or different components. For example, the terminal device may also include input transmitting devices, network access devices, buses, etc.

[0072] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0073] In some embodiments, the memory may be an internal storage unit of the terminal device, such as a hard drive or RAM. The memory may also be an external storage device of the terminal device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory may include both internal and external storage units. The memory is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory can also be used to temporarily store data that has been sent or will be sent.

[0074] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0075] This application provides a computer program product that, when run on a mobile terminal device, enables the mobile terminal device to implement the steps described in the above-described method embodiments.

[0076] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0077] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0078] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0079] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0080] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An automatic detection method for image materials, characterized in that, The method includes: S1: Atomize all the review rules of each platform to obtain several rule atoms; S2: Determine the mapping parameters for each rule atom, and then integrate each rule atom and its corresponding mapping parameters into a set of rule atoms for the platform; S3: Whenever a user uploads video footage, determine the target platform to which the video footage needs to be uploaded; S4: Retrieve the set of rule atoms corresponding to each target platform; S5: Determine the intersection of the sets of all rule atoms to obtain the target set; S6: Use the rules in the target set to detect image materials so that the images that pass the detection can simultaneously meet the review rules of all target platforms.

2. The method according to claim 1, characterized in that, Atomically decompose all the review rules of each platform to obtain several rule atoms, including: S1: Retrieve several set rule category names; S2: For each platform, obtain the platform's review rules document; S3: Select a rule category name as the comparison name; S4: Compare the comparison items with the text of the rule document one by one to extract the rule items corresponding to the comparison items from the text of the rule document, i.e., rule atoms; S5: Take another rule category name as the comparison name, and execute steps S4 to S5 until the rule atom corresponding to each rule category name is determined.

3. The method according to claim 2, characterized in that, Determining the mapping parameters for each rule atom, and then integrating each rule atom and its corresponding mapping parameters into a set of rule atoms for the platform, includes: For each rule atom, context analysis is performed in the document in which it was initially located to determine the corresponding rule parameters. The rule parameters are then mapped to the rule atom, forming a mapping binding between the rule parameters and the rule atom. Each rule atom and its bound mapping parameters are integrated into a set to obtain a set of rule atoms.

4. The method according to claim 3, characterized in that, Determine the intersection of the sets of all rule atoms to obtain the target set, which includes: S51: Generate a spare set, wherein the initial state of the spare set is an empty set; S52: Select a rule category name as the target directory; S53: Screen out the regular atoms corresponding to the target list from each set of regular atoms; S54: For each screened regular atom, determine the parameter range of the corresponding mapping parameters; S55: Find the intersection of the parameter ranges of the mapping parameters of each regular atom to obtain the target parameter range; S56: Generate a regular atom that is the same as the sieved regular atom, i.e., the target atom; S57: Bind the target parameter range to the target atom mapping and include the target atom in the spare set; S58: Select another rule category name as the target list, and execute steps S51 to S58 until all the target atoms corresponding to all rule category names are included in the standby set, thus obtaining the target set.

5. The method according to claim 4, characterized in that, Between step S5 and step S6, the following is also included: Identify the target object for each target atom, where the target object includes video frames, images, and text.

6. The method according to claim 5, characterized in that, Image materials were detected using the atoms of each rule in the target set, including: Identify all measurement points affecting the material and determine the detection object for each measurement point, including video frames, images, and text; For each measuring point, determine the number of objects to be detected at that measuring point; If the number of objects to be detected is 1, the target atom that matches the corresponding target object will be retrieved for detection, and the qualification of the test point will be determined based on the obtained detection score. If the number of objects to be detected is greater than 1, then for each object to be detected, the corresponding target object and the target atoms that match the object to be detected are retrieved and detected to obtain a detection score that is the same as the number of objects to be detected. Then, based on the detection scores, it is determined whether the test point is qualified.

7. The method according to claim 6, characterized in that, If a measurement point is found to be non-compliant, the following are also included: For each non-compliant test point, the parameters of each detection object for that non-compliant test point are adjusted based on the mapping parameters of the corresponding target atom. The non-compliant test point is then detected again to determine whether it has been converted into a compliant test point. This step is repeated until the non-compliant test point is converted into a compliant test point.

8. A terminal device, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the automatic image material detection method according to any one of claims 1 to 7.

9. A terminal-readable storage medium, characterized in that, The terminal-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to perform the steps of the automatic image material detection method according to any one of claims 1 to 7.

Images