Data annotation method and system based on artificial intelligence

Abstract

The invention is suitable for the technical field of data annotation, and provides a data annotation method and system based on artificial intelligence, and the method comprises the steps: analyzing whether the complexity level of a current scene of a target time slice is higher than a corresponding threshold value or not when the proportion of a triggering secondary annotation reminding frame in the target time slice exceeds a threshold value proportion, and if yes, judging whether the current scene complexity level of the target time slice is higher than the corresponding threshold value; and if yes, acquiring a historical video processing record and historical annotation deviation feedback data. According to the method, double statistical parameters of the trend curve slope mean value and the historical annotation deviation mean value are introduced, the confidence correction factor is constructed, a dynamic adjustment mechanism of confidence parameters is achieved, and the trend misjudgment deviation problem formed by accumulation of the system under similar scene complexity can be effectively recognized and corrected.

Description

Technical Field

[0001] The present invention belongs to the technical field of data annotation, and in particular, relates to a data annotation method and system based on artificial intelligence. Background Art

[0002] In existing data labeling technologies, artificial intelligence models usually generate confidence parameters for each frame during the analysis and processing of video, image and other data to measure the system's credibility of the label recognition results. To improve data labeling efficiency, existing systems generally set a fixed confidence threshold. When the confidence of a frame is lower than the threshold, it automatically triggers a "secondary labeling reminder" to mark the frame as a suspected low-confidence frame that requires further manual confirmation. This method can assist manual verification to a certain extent, avoid the direct acceptance of obviously erroneous labeling results, and thus improve the quality of labeling.

[0003] However, existing technologies generally use static threshold strategies for judgment, and fail to effectively model the rules of system confidence fluctuations in different video scenarios. Especially in scenes with complex structures or many interference factors, the labeling system is prone to concentrated false alarms, but the current system cannot dynamically identify whether there are systematic deviations based on historical processing trends or manual correction feedback. This lack of an adaptive correction mechanism can easily cause the system to repeatedly trigger redundant reminders in similar scenes, increasing the burden of manual verification, and as the false alarm trend continues to accumulate, the accuracy of confidence judgments will continue to decline, seriously affecting the overall efficiency and consistency of the labeling process. Summary of the invention

[0004] The purpose of the present invention is to provide a data annotation method and system based on artificial intelligence, aiming to solve the problems raised in the background technology.

[0005] The present invention is implemented as follows: a data labeling method based on artificial intelligence, the method comprising: When the proportion of frames triggering secondary annotation reminders in the target time segment exceeds the threshold ratio, analyze whether the current scene complexity level of the target time segment is higher than the corresponding threshold. If so, obtain the historical video processing records and historical annotation deviation feedback data; Analyze historical video processing records and select several historical comparison segment records that are within the same level range as the current scene complexity level and have matching time length; Analyze each historical comparison segment record in turn, calculate the corresponding secondary annotation reminder frame ratio, draw a trend curve in chronological order, and calculate the average slope of the trend curve; Analyze the historical annotation deviation feedback data, obtain the deviation value between the initial annotation value and the correction result corresponding to each historical comparison segment record, and calculate the average of all deviation values; Based on the mean of the slope and the mean of all deviation values, a confidence correction factor is generated by weighting and used to adjust the confidence parameters of all frames in the target time segment.

[0006] As a further limitation of the technical solution of the embodiment of the present invention, the steps of parsing each historical comparison segment record in turn, calculating the corresponding secondary annotation reminder frame ratio, drawing a trend curve in chronological order, and calculating the slope mean of the trend curve include: Analyze each historical comparison segment record in turn, calculate the ratio between the number of frames that trigger the secondary annotation reminder and the corresponding total number of frames in each historical comparison segment record, and obtain the respective secondary annotation reminder frame ratios; With the time axis as the horizontal axis and the secondary annotation reminder frame ratio as the vertical axis, draw the corresponding trend curve; The mean slope of the trend curve is calculated.

[0007] As a further limitation of the technical solution of the embodiment of the present invention, the steps of parsing the historical annotation deviation feedback data, obtaining the deviation value between the initial annotation value and the correction result corresponding to each historical comparison segment record, and calculating the mean of all the deviation values ​​include: Parse the historical annotation deviation feedback data, and extract the local annotation deviation feedback data corresponding to each historical comparison segment record; Analyze each local annotation deviation feedback data in turn, and extract the initial annotation value and the correction result from it, wherein the initial annotation value refers to all frames in each historical comparison segment record that trigger the secondary annotation reminder, and the correction result refers to the number of frames that are finally determined to require secondary annotation reminder after manual confirmation or feedback; The deviation between each group of initial annotation values ​​and the corresponding correction results is calculated, and the deviation values ​​corresponding to all historical comparison segment records are counted to obtain the mean of the deviation values.

[0008] As a further limitation of the technical solution of the embodiment of the present invention, the step of weighting and generating a confidence correction factor based on the slope mean and the mean of all deviation values, and adjusting the confidence parameters of all frames in the target time segment includes: The confidence correction factor generation formula is called, and the mean value of the slope of the trend curve and the mean value of all deviation values ​​are used as input parameters, and a weighted calculation is performed to generate the confidence correction factor; Adjusting the initial confidence parameters of all frames in the target time segment based on the confidence correction factor to obtain corrected confidence parameters; It is determined whether the corrected confidence parameter of each frame is lower than the preset confidence threshold. If so, the secondary annotation reminder of the frame is retained; if not, the original secondary annotation reminder is cancelled.

[0009] As a further limitation of the technical solution of the embodiment of the present invention, the confidence correction factor generation formula is: ,in refers to the confidence correction factor, Refers to the mean slope of the trend curve. Refers to the adjustment weight corresponding to the mean of the slope, refers to the absolute value of the mean of all deviation values, Refers to the adjustment weight corresponding to this absolute value; In the confidence correction factor generation formula: ,in Refers to the total number of local annotation deviation feedback data, where Refers to the The correction result of the local annotation deviation feedback data, Refers to the The initial annotation value of the local annotation deviation feedback data, Refers to the The deviation value between the initial annotation value corresponding to the local annotation deviation feedback data and the corresponding correction result.

[0010] A data annotation system based on artificial intelligence, the system comprises: a data acquisition module, a data screening module, a slope calculation module, a deviation value calculation module and a correction factor generation module, wherein: The data acquisition module is used to analyze whether the current scene complexity level of the target time segment is higher than the corresponding threshold when the proportion of frames triggering secondary annotation reminders in the target time segment exceeds the threshold ratio. If so, the historical video processing records and historical annotation deviation feedback data are obtained; A data screening module is used to parse historical video processing records and screen out a number of historical comparison segment records that are within the same level range as the current scene complexity level and have a matching time length; The slope calculation module is used to parse each historical comparison segment record in turn, calculate the corresponding secondary annotation reminder frame ratio, draw a trend curve in chronological order, and calculate the slope mean of the trend curve; The deviation value calculation module is used to analyze the historical annotation deviation feedback data, obtain the deviation value between the initial annotation value and the correction result corresponding to each historical comparison segment record, and calculate the average value of all deviation values; The correction factor generation module is used to generate a confidence correction factor based on the mean of the slope and the mean of all deviation values ​​by weighting, and is used to adjust the confidence parameters of all frames in the target time segment.

[0011] As a further limitation of the technical solution of the embodiment of the present invention, the slope calculation module specifically includes: A ratio calculation unit is used to parse each historical comparison segment record in turn, calculate the ratio between the number of frames triggering the secondary marking reminder and the corresponding total number of frames in each historical comparison segment record, and obtain the respective secondary marking reminder frame ratio; A curve drawing unit, used to draw a corresponding trend curve with the time axis as the horizontal axis and the secondary marking reminder frame ratio as the vertical axis; The slope calculation unit is used to calculate the average slope of the trend curve.

[0012] As a further limitation of the technical solution of the embodiment of the present invention, the deviation value calculation module specifically includes: A local feedback data acquisition unit, used to parse the historical annotation deviation feedback data and extract the local annotation deviation feedback data corresponding to each historical comparison segment record; A local feedback data parsing unit is used to parse each local annotation deviation feedback data in turn, and extract the initial annotation value and the correction result therefrom, wherein the initial annotation value refers to all frames in each historical comparison segment record that trigger the secondary annotation reminder, and the correction result refers to the number of frames that are finally determined to require the secondary annotation reminder after manual confirmation or feedback; The deviation value mean calculation unit is used to calculate the deviation value between each group of initial marking values ​​and the corresponding correction result, and to perform statistics on the deviation values ​​corresponding to all historical comparison segment records to obtain the mean value of the deviation value.

[0013] As a further limitation of the technical solution of the embodiment of the present invention, the correction factor generation module specifically includes: A confidence correction factor generation unit is used to call a confidence correction factor generation formula, and use the average value of the slope of the trend curve and the average value of all deviation values ​​as input parameters to perform weighted calculation to generate a confidence correction factor; A confidence parameter correction unit, used to adjust the initial confidence parameters of all frames in the target time segment based on the confidence correction factor to obtain a corrected confidence parameter; The reminder retention determination unit is used to determine whether the revised confidence parameter of each frame is lower than a preset confidence threshold. If so, the secondary marking reminder of the frame is retained; if not, the original secondary marking reminder is cancelled.

[0014] As a further limitation of the technical solution of the embodiment of the present invention, the confidence correction factor generation formula is: ,in refers to the confidence correction factor, Refers to the mean slope of the trend curve. Refers to the adjustment weight corresponding to the mean of the slope, refers to the absolute value of the mean of all deviation values, Refers to the adjustment weight corresponding to this absolute value; In the confidence correction factor generation formula: ,in Refers to the total number of local annotation deviation feedback data, where Refers to the The correction result of the local annotation deviation feedback data, Refers to the The initial annotation value of the local annotation deviation feedback data, Refers to the The deviation value between the initial annotation value corresponding to the local annotation deviation feedback data and the corresponding correction result.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention constructs a confidence correction factor by introducing dual statistical parameters of the mean value of the trend curve slope and the mean value of the historical annotation deviation, thus realizing a dynamic adjustment mechanism for the confidence parameter, which can effectively identify and correct the trend misjudgment offset problem accumulated by the system under similar scene complexity. Compared with the prior art that only judges based on static confidence thresholds, the present invention is based on the fusion modeling of historical processing records and manual feedback data, which improves the system's perception ability and judgment accuracy of the cause of confidence fluctuations, significantly reduces redundant reminders, reduces the burden of manual review, and enhances the stability and adaptability of the system under complex video structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flowchart of a method provided by an embodiment of the present invention; Figure 2 A flow chart of drawing a trend curve and calculating the mean value of its slope in the method provided in an embodiment of the present invention; Figure 3 A flow chart of calculating the deviation value between the initial marked value and the corresponding correction result in the method provided in an embodiment of the present invention; Figure 4 A flow chart of generating a confidence correction factor in the method provided in an embodiment of the present invention; Figure 5 An application architecture diagram of a system provided by an embodiment of the present invention; Figure 6 A structural block diagram of a slope calculation module in a system provided by an embodiment of the present invention; Figure 7 A structural block diagram of a deviation value calculation module in a system provided by an embodiment of the present invention; Figure 8 This is a structural block diagram of a correction factor generation module in a system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] Figure 1 A flow chart of a method provided by an embodiment of the present invention is shown.

[0019] Specifically, a data annotation method based on artificial intelligence includes the following steps: Step S100, when the proportion of frames triggering secondary annotation reminders in the target time segment exceeds a threshold ratio, analyze whether the current scene complexity level of the target time segment is higher than the corresponding threshold. If so, obtain historical video processing records and historical annotation deviation feedback data.

[0020] In the embodiment of the present invention, the target time segment refers to a local continuous frame segment demarcated by the system according to the triggering situation of the secondary annotation reminder when the current video data is annotated and analyzed, and is used to form a sub-range to be further identified and processed. The time segment usually has a fixed frame length or is dynamically demarcated according to the annotation behavior, and its boundary frame is determined based on the dense frame area that triggers the secondary annotation reminder.

[0021] The secondary annotation reminder frame is a suspected low-confidence frame determined by the system by judging whether the confidence parameter corresponding to each frame is lower than the preset confidence threshold. The secondary annotation reminder frame automatically marks such frames as suspected annotation frames that require further manual confirmation during the preliminary annotation stage. The reminder frame usually comes from the system's initial perception of typical low-confidence scenes such as blurred image target boundaries, interrupted action continuity, occlusion overlap, or uncertain semantic recognition. It serves as an important reminder point that requires manual review again in the annotation process.

[0022] The confidence parameter refers to the credibility value generated by the labeling system based on the results of the target recognition, semantic judgment or behavior analysis in each frame of video data, which is used to measure the certainty of the system's judgment result on the label of the current frame. This parameter is usually expressed in numerical form. The higher the value, the more confident the system is in the labeling result of the frame, and the lower the value, the higher the uncertainty of the system.

[0023] The preset confidence threshold refers to a judgment boundary value set within the system to determine whether the confidence parameter of the current frame is reliable. When the confidence parameter of a frame is lower than the threshold, the system believes that the annotation result of the frame is not sufficiently credible and needs to trigger a secondary annotation reminder to include it in the manual review scope; when the confidence parameter is higher than the threshold, the system believes that the annotation result has reasonable accuracy and can directly retain the original label without further processing. The threshold can be adjusted according to the task type, model performance or scenario requirements.

[0024] The threshold ratio is used to measure whether the number of secondary annotated reminder frames in the target time segment has reached the abnormal density determined by the system, and is used to determine whether to enter the trend correction process. The threshold ratio can be set based on the system's annotation experience data. For example, in historical tasks, when the proportion of reminder frames in a certain time segment exceeds 15%, it often corresponds to a model misjudgment concentration area. Therefore, the threshold ratio can be set to 15% or dynamically adjusted according to specific application scenarios.

[0025] The scene complexity level refers to the scene structure complexity grading result comprehensively evaluated by the system based on multiple dimensions such as the spatial structure of the image within the target time segment, target density, difference between previous and next frames, occlusion, etc. This level is usually expressed in multiple levels, such as from level 1 (simple) to level 10 (complex), and can be automatically generated by the image feature extraction module.

[0026] The reason why "whether the current scene complexity level is higher than the corresponding threshold" is set as the judgment condition for triggering further analysis in the present invention is that when the scene structure of the video clip is complex, the probability of misjudgment of the object boundary, action intention or semantic judgment by the annotation system increases significantly, and even a short-term annotation reminder concentration in a simple scene may not represent a system deviation, so it is necessary to use the complexity level as a threshold for whether to perform trend correction analysis. The corresponding threshold can be set based on historical annotation performance. For example, when the scene complexity level is at level three or above, the model confidence fluctuates significantly, so level three is used as the starting judgment threshold.

[0027] The historical video processing records are derived from the data processing logs recorded during the system's previous annotation tasks, including original annotation-related data such as feature recognition of video clips, confidence output, label generation and modification, and trigger reminder records. The historical annotation deviation feedback data is derived from the proofreading operation records of the manual annotation stage, including the difference information between the initial annotation and the final correction result in each paragraph, which is an important data basis for the system to build deviation identification and correction trends.

[0028] Among them, the historical video processing record includes at least: video clip number, frame index range, initial annotation label generated by the system, confidence value corresponding to each frame, whether to trigger the secondary annotation reminder mark, scene structure feature summary information, and corresponding video metadata, etc.

[0029] The historical annotation deviation feedback data at least includes: the system original annotation frame set in each annotation section, the manual final correction confirmation frame set, the label difference result between the two, the manual feedback timestamp, the correction confirmation status mark, etc.

[0030] Furthermore, the artificial intelligence-based data labeling method further includes the following steps: Step S200, parsing the historical video processing records, and selecting a number of historical comparison segment records that are within the same level range as the current scene complexity level and have a matching time length.

[0031] In an embodiment of the present invention, screening out several historical reference segment records that are within the same level range as the current scene complexity level and have matching time lengths means that in the acquired historical video processing records, based on the structural attributes of the current target time segment, historical segments with similar background complexity and frame number characteristics are screened out as reference objects to ensure the comparability and reference value of subsequent trend analysis and deviation judgment.

[0032] Specifically, the system first obtains the scene complexity level of the current target time segment and sets a level tolerance interval, such as a range of ±1 level centered on the current level, as the level range allowed for matching. The system then traverses the recorded scene complexity levels of each video segment in the historical video processing record and selects the segments whose complexity levels fall within the tolerance range.

[0033] On this basis, the system further matches the time length of these candidate segments. The time length matching condition can be set as the frame difference does not exceed the preset percentage threshold (for example, ±10%) of the current segment frame number, to ensure that the selected historical comparison segment records have similar behavior rhythm and annotation span as the current segment in terms of duration. Finally, the historical segments that meet the above two conditions are determined as the historical comparison segment record set required for the current analysis.

[0034] We select historical clips with similar complexity levels to the current scene because in the annotation system, scene complexity is often closely related to annotation confidence fluctuations and model judgment accuracy. Video segments of different complexity levels have significantly different target overlap levels, background dynamic factors, and visual structure characteristics. If the comparative analysis is not within the same level range, it is easy to cause bias judgment errors.

[0035] At the same time, limiting the time length matching can ensure that the compared historical segments and the current segments have a basis for structural alignment in terms of frame density, action continuity, label trigger rhythm, etc., so that the subsequent calculation of the mean slope and deviation value of the trend curve is more objective and stable, avoiding statistical distortion caused by differences in segment dimensions. Through these two screening criteria, the representativeness and referenceability of the historical control segment records are ensured, providing accurate data support for the subsequent generation of confidence correction factors in the present invention.

[0036] Furthermore, the artificial intelligence-based data labeling method further includes the following steps: Step S300 , parse each historical comparison segment record in turn, calculate the corresponding secondary annotation reminder frame ratio, draw a trend curve in chronological order, and calculate the average slope of the trend curve.

[0037] Specifically, Figure 2 A flow chart for drawing a trend curve and calculating the mean of its slope is shown.

[0038] The following steps are used to parse each historical comparison segment record in turn, calculate the corresponding secondary annotation reminder frame ratio, draw a trend curve in chronological order, and calculate the slope mean of the trend curve: Step S301, parsing each historical comparison segment record in turn, calculating the ratio between the number of frames triggering secondary marking reminders and the corresponding total number of frames in each historical comparison segment record, and obtaining the respective secondary marking reminder frame ratios; Step S302, drawing a corresponding trend curve with the time axis as the horizontal axis and the secondary marking reminder frame ratio as the vertical axis; Step S303, calculating the mean slope of the trend curve.

[0039] In the embodiment of the present invention, the "analyzing each historical comparison segment record, calculating the ratio between the number of frames that trigger the secondary annotation reminder and the corresponding total number of frames in each historical comparison segment record, and obtaining the respective secondary annotation reminder frame ratio" means that for the historical comparison segment records obtained by the above screening, the system sequentially reads the total number of frames contained in each segment record, as well as the frame index set marked by the system as triggering the secondary annotation reminder in the segment. The system calculates the ratio of the number of frames that trigger the secondary annotation reminder in each segment to the total number of frames in the segment, and obtains a normalized ratio value, which is used to reflect the density of the segment being identified as low-confidence frames during the historical annotation process.

[0040] Specifically, the system reads the frame index range of each segment in the historical video processing record and the corresponding secondary annotation reminder mark table, counts the frames with reminder marks, and calculates the total number of frames based on the start and end frame numbers of the segment. The percentage of secondary annotation reminder frames for the segment can be obtained by dividing the number of reminder frames by the total number of frames. This percentage value is used to express the relative density of system uncertainty marks generated by the historical segment during the annotation process, and can reflect the confidence fluctuation of the segment in model recognition.

[0041] The drawing of the trend curve means that after the system obtains the proportion of secondary annotation reminder frames of multiple historical comparison segment records, it sets the time axis as the horizontal axis and the corresponding proportion value as the vertical axis according to the order of appearance of these historical segments on the timeline, and connects these points in the two-dimensional coordinate system to form a continuous trend line. The time sequence can be the timestamp sequence generated by the historical segment records, or the order of their records in the system history processing log, with the purpose of establishing the change trend of the proportion indicator through the time dimension.

[0042] The mean slope of the trend curve is calculated to identify whether the annotation system in a historically similar context shows a trend of persistent misjudgment. When the mean slope is greater than zero, it means that in a context of similar complexity, the proportion of secondary annotation reminder frames of the system has an upward trend, which may mean that the system shows increasing uncertainty in this type of clip; when the mean slope is less than zero, it means that the reminder ratio is gradually decreasing, indicating that the system processing stability is enhanced; if the mean slope is close to zero, it means that the misjudgment volatility of this type of video segment in history is small, and the system judgment is relatively stable.

[0043] Therefore, the formation of the trend misjudgment deviation direction is essentially due to the fact that the annotation system fails to dynamically adjust its confidence output strategy when processing video segments with similar structural complexity levels, resulting in the accumulation of systematic deviations in the historical processing process. Specifically, in this type of complexity background, the error between the original judgment result of the model and the manual correction result is not fed back to the confidence adjustment mechanism in a timely manner, which in turn causes the reminder density to gradually shift higher or lower over time, seriously affecting the system's confidence judgment accuracy and reminder control capabilities for the current video clip. The present invention introduces the trend curve slope mean parameter and combines historical correction deviation statistics to construct a confidence correction mechanism with time trend recognition capabilities. It can achieve adaptive correction for the above-mentioned problems and improve the stability and robustness of the annotation system in complex structure scenarios.

[0044] Furthermore, the artificial intelligence-based data labeling method further includes the following steps: Step S400 , parsing the historical annotation deviation feedback data, obtaining the deviation value between the initial annotation value and the correction result corresponding to each historical comparison segment record, and calculating the average of all the deviation values.

[0045] Specifically, Figure 3 A flow chart for calculating the deviation value between the initial annotation value and the corresponding correction result is shown.

[0046] The process of parsing the historical annotation deviation feedback data, obtaining the deviation value between the initial annotation value and the correction result corresponding to each historical comparison segment record, and calculating the mean of all deviation values ​​specifically includes the following steps: Step S401, parsing the historical annotation deviation feedback data, and extracting the local annotation deviation feedback data corresponding to each historical comparison segment record; Step S402, analyzing each local annotation deviation feedback data in turn, and extracting the initial annotation value and the correction result therefrom, wherein the initial annotation value refers to all frames in each historical comparison segment record that trigger a secondary annotation reminder, and the correction result refers to the number of frames that are finally determined to require a secondary annotation reminder after manual confirmation or feedback; Step S403, calculating the deviation value between each group of initial marking values ​​and the corresponding correction result, and performing statistics on the deviation values ​​corresponding to all historical comparison segment records to obtain the average value of the deviation values.

[0047] In an embodiment of the present invention, the local annotation deviation feedback data corresponding to the historical comparison segment record is extracted by association through the index relationship established when the historical data is archived. While the system performs the annotation task and records the historical video processing record, it will store the corresponding manual feedback data as supplementary information. Each historical comparison segment record contains unique identification information of the time segment, such as the paragraph number, frame start and end range or timestamp. Based on this type of identification information, the system can locate the corresponding local feedback data from the historical annotation deviation feedback data to ensure the accuracy of the one-to-one correspondence.

[0048] When analyzing each set of local annotation deviation feedback data, the system extracts two key indicators: one is the number of frames that the system believes need to trigger a secondary annotation reminder during the initial judgment stage, as the initial annotation value; the other is the number of frames that are finally confirmed to need a secondary annotation reminder after manual review, as the correction result. After the system completes the traversal of all historical comparison segment records, it calculates the difference between the initial annotation value and the correction result in each set of data one by one, and quantifies these difference results into deviation values.

[0049] Next, the system summarizes and counts the deviation values ​​corresponding to all historical comparison segment records, and finally calculates the average value of these deviation values. This average value represents the average error degree between the system's automatic judgment and manual correction under similar scene complexity, and is an important reference for measuring whether the system's confidence output is stable.

[0050] By calculating the mean of the deviation value, we can quantify the system's misjudgment range in similar backgrounds and reflect the deviation trend of the annotation system's judgment on uncertain areas. If the mean is large, it means that the system has obvious confidence judgment deviation in such scenes and needs to be compensated by a correction mechanism; if the mean is small, it means that the system is consistent with the manual results and the current strategy can be maintained without adjustment. Therefore, as a component of the subsequent generation of confidence correction factors, the mean can improve the system's responsiveness to historical misjudgment trends and achieve more accurate annotation strategy optimization.

[0051] Furthermore, the artificial intelligence-based data labeling method further includes the following steps: Step S500, based on the slope mean and the mean of all deviation values, a confidence correction factor is generated by weighting, and is used to adjust the confidence parameters of all frames in the target time segment.

[0052] Specifically, Figure 4 A flow chart for generating a confidence correction factor is shown.

[0053] The confidence correction factor is generated by weighting based on the slope mean and the mean of all deviation values, and is used to adjust the confidence parameters of all frames in the target time segment. Specifically, the following steps are included: Step S501, calling a confidence correction factor generation formula, taking the average slope of the trend curve and the average of all deviation values ​​as input parameters, and performing weighted calculation to generate a confidence correction factor; Step S502, adjusting the initial confidence parameters of all frames in the target time segment based on the confidence correction factor to obtain corrected confidence parameters; Step S503, determining whether the corrected confidence parameter of each frame is lower than a preset confidence threshold, if so, retaining the secondary marking reminder of the frame, if not, canceling the original secondary marking reminder.

[0054] The confidence correction factor generation formula is: ,in refers to the confidence correction factor, Refers to the mean slope of the trend curve. Refers to the adjustment weight corresponding to the mean of the slope, refers to the absolute value of the mean of all deviation values, Refers to the adjustment weight corresponding to this absolute value; In the confidence correction factor generation formula: ,in Refers to the total number of local annotation deviation feedback data, where Refers to the The correction result of the local annotation deviation feedback data, Refers to the The initial annotation value of the local annotation deviation feedback data, Refers to the The deviation value between the initial annotation value corresponding to the local annotation deviation feedback data and the corresponding correction result.

[0055] In an embodiment of the present invention, the mean slope of the trend curve and the mean of all deviation values ​​are used as parameter inputs, and weighted calculation is performed to generate a confidence correction factor because these two types of parameters reflect the stability and offset strength of the historical reference segment records in the time dimension and error dimension, respectively. The mean slope of the trend curve is used to characterize the confidence fluctuation evolution trend of the system when processing similar complexity scenarios. If the trend continues to rise or fall, it often indicates that the system's reliability in labeling this type of video segment is in dynamic change; and the mean of all deviation values ​​reflects the overall degree of deviation between the system's judgment results and the final manual correction results, which is an important indicator for measuring system stability and consistency. Weighted fusion of these two parameters helps the system accurately identify potential false alarm or missed alarm risks when processing the current target time segment, and achieves directional adjustment and strength correction of the initial confidence judgment.

[0056] Through the above mechanism, the system can adaptively and dynamically fine-tune the confidence parameters of the current frame segment according to changes in historical trends and feedback data while keeping the overall recognition strategy unchanged, thereby avoiding repetitive misjudgments caused by a fixed threshold mechanism, effectively enhancing the system's adaptability to diverse video scenes, and improving the accuracy of recognition results and the flexibility of reminder control.

[0057] During implementation, the system can use a variety of methods to adjust the confidence parameters, such as using the weighted offset method to use the confidence correction factor as an adjustment increment to directly act on the initial confidence value; or using the confidence correction factor as an adjustment ratio to proportionally compress or amplify the initial confidence value; or by looking up the correction reference table established based on the historical model calibration results, mapping the correction factor with the confidence interval to obtain the corrected confidence parameter. Any of the above methods can achieve confidence optimization based on historical trends.

[0058] Subsequently, the system judges the corrected confidence parameters of each frame in the target time segment one by one. If the confidence value corresponding to a frame is still lower than the preset confidence threshold, the system retains its original secondary annotation reminder label; if the confidence value is higher than the threshold, the original reminder mark is automatically cancelled. Through this judgment mechanism, the system can make a more reasonable judgment on the effectiveness of the current reminder label based on the experience of historical deviations, thereby effectively reducing invalid or redundant manual review operations, reducing the burden of manual intervention, and improving the accuracy and efficiency of the overall annotation process.

[0059] Furthermore, Figure 5 The application architecture diagram of the system provided by the embodiment of the present invention is shown.

[0060] Among them, in another preferred embodiment provided by the present invention, a data annotation system based on artificial intelligence includes: The data acquisition module 100 is used to analyze whether the current scene complexity level of the target time segment is higher than the corresponding threshold when the proportion of frames triggering secondary annotation reminders in the target time segment exceeds a threshold ratio. If so, the historical video processing records and historical annotation deviation feedback data are obtained.

[0061] Furthermore, the artificial intelligence-based data annotation system also includes: The data screening module 200 is used to parse the historical video processing records and screen out a number of historical comparison segment records that are within the same level range as the current scene complexity level and have a matching time length.

[0062] Furthermore, the artificial intelligence-based data annotation system also includes: The slope calculation module 300 is used to analyze each historical comparison segment record in turn, calculate the corresponding secondary annotation reminder frame ratio, draw a trend curve in chronological order, and calculate the slope average of the trend curve.

[0063] Specifically, Figure 6 The structure block diagram of the slope calculation module 300 in the system provided by the embodiment of the present invention is shown.

[0064] In a preferred embodiment of the present invention, the slope calculation module 300 specifically includes: The ratio calculation unit 301 is used to analyze each historical comparison segment record in turn, calculate the ratio between the number of frames triggering the secondary marking reminder and the corresponding total number of frames in each historical comparison segment record, and obtain the respective secondary marking reminder frame ratio; The curve drawing unit 302 is used to draw a corresponding trend curve with the time axis as the horizontal axis and the secondary marking reminder frame ratio as the vertical axis; The slope calculation unit 303 is used to calculate the mean slope of the trend curve.

[0065] Furthermore, the artificial intelligence-based data annotation system also includes: The deviation value calculation module 400 is used to analyze the historical annotation deviation feedback data, obtain the deviation value between the initial annotation value and the correction result corresponding to each historical comparison segment record, and calculate the average value of all deviation values.

[0066] Specifically, Figure 7The structure block diagram of the deviation value calculation module 400 in the system provided by the embodiment of the present invention is shown.

[0067] In a preferred embodiment of the present invention, the deviation value calculation module 400 specifically includes: The local feedback data acquisition unit 401 is used to parse the historical annotation deviation feedback data and extract the local annotation deviation feedback data corresponding to each historical comparison segment record; The local feedback data parsing unit 402 is used to parse each local annotation deviation feedback data in turn, and extract the initial annotation value and the correction result therefrom, wherein the initial annotation value refers to all frames in each historical comparison segment record that trigger the secondary annotation reminder, and the correction result refers to the number of frames that are finally determined to require the secondary annotation reminder after manual confirmation or feedback; The deviation value mean value calculation unit 403 is used to calculate the deviation value between each group of initial marking values ​​and the corresponding correction result, and to perform statistics on the deviation values ​​corresponding to all historical comparison segment records to obtain the mean value of the deviation value.

[0068] Furthermore, the artificial intelligence-based data annotation system also includes: The correction factor generation module 500 is used to generate a confidence correction factor by weighting based on the slope mean and the mean of all deviation values, and to adjust the confidence parameters of all frames in the target time segment.

[0069] Specifically, Figure 8 It shows a structural block diagram of the correction factor generation module 500 in the system provided by the embodiment of the present invention.

[0070] In a preferred embodiment of the present invention, the correction factor generation module 500 specifically includes: The confidence correction factor generating unit 501 is used to call the confidence correction factor generating formula, and use the mean value of the slope of the trend curve and the mean value of all deviation values ​​as input parameters, and perform weighted calculation to generate the confidence correction factor; A confidence parameter correction unit 502, configured to adjust the initial confidence parameters of all frames in the target time segment based on the confidence correction factor to obtain a corrected confidence parameter; The reminder retention determination unit 503 is used to determine whether the revised confidence parameter of each frame is lower than a preset confidence threshold. If so, the secondary marking reminder of the frame is retained; if not, the original secondary marking reminder is cancelled.

[0071] The confidence correction factor generation formula is: ,in refers to the confidence correction factor, Refers to the mean slope of the trend curve. Refers to the adjustment weight corresponding to the mean of the slope, refers to the absolute value of the mean of all deviation values, Refers to the adjustment weight corresponding to this absolute value; In the confidence correction factor generation formula: ,in Refers to the total number of local annotation deviation feedback data, where Refers to the The correction result of the local annotation deviation feedback data, Refers to the The initial annotation value of the local annotation deviation feedback data, Refers to the The deviation value between the initial annotation value corresponding to the local annotation deviation feedback data and the corresponding correction result.

[0072] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0073] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0074] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0075] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0076] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A data annotation method based on artificial intelligence, characterized in that: The method comprises: When the proportion of frames triggering secondary annotation reminders in the target time segment exceeds the threshold ratio, analyze whether the current scene complexity level of the target time segment is higher than the corresponding threshold. If so, obtain the historical video processing records and historical annotation deviation feedback data; Analyze historical video processing records and select several historical comparison segment records that are within the same level range as the current scene complexity level and have matching time length; Analyze each historical comparison segment record in turn, calculate the corresponding secondary annotation reminder frame ratio, draw a trend curve in chronological order, and calculate the average slope of the trend curve; Analyze the historical annotation deviation feedback data, obtain the deviation value between the initial annotation value and the correction result corresponding to each historical comparison segment record, and calculate the average of all deviation values; Based on the mean of the slope and the mean of all deviation values, a confidence correction factor is generated by weighting and used to adjust the confidence parameters of all frames in the target time segment.

2. The artificial intelligence-based data annotation method according to claim 1, characterized in that: The steps of successively parsing each historical comparison segment record, calculating the corresponding secondary annotation reminder frame ratio, drawing a trend curve in chronological order, and calculating the slope mean of the trend curve include: Analyze each historical comparison segment record in turn, calculate the ratio between the number of frames that trigger the secondary annotation reminder and the corresponding total number of frames in each historical comparison segment record, and obtain the respective secondary annotation reminder frame ratios; With the time axis as the horizontal axis and the secondary annotation reminder frame ratio as the vertical axis, draw the corresponding trend curve; The mean slope of the trend curve is calculated.

3. The artificial intelligence-based data annotation method according to claim 2, characterized in that: The steps of parsing the historical annotation deviation feedback data, obtaining the deviation value between the initial annotation value and the correction result corresponding to each historical comparison segment record, and calculating the mean value of all deviation values ​​include: Parse the historical annotation deviation feedback data, and extract the local annotation deviation feedback data corresponding to each historical comparison segment record; Analyze each local annotation deviation feedback data in turn, and extract the initial annotation value and the correction result from it, wherein the initial annotation value refers to all frames in each historical comparison segment record that trigger the secondary annotation reminder, and the correction result refers to the number of frames that are finally determined to require secondary annotation reminder after manual confirmation or feedback; The deviation between each group of initial annotation values ​​and the corresponding correction results is calculated, and the deviation values ​​corresponding to all historical comparison segment records are counted to obtain the mean of the deviation values.

4. The artificial intelligence-based data annotation method according to claim 3, characterized in that: The steps of weighting and generating a confidence correction factor based on the slope mean and the mean of all deviation values, and using the factor to adjust the confidence parameters of all frames in the target time segment include: The confidence correction factor generation formula is called, and the mean value of the slope of the trend curve and the mean value of all deviation values ​​are used as input parameters, and a weighted calculation is performed to generate the confidence correction factor; Adjusting the initial confidence parameters of all frames in the target time segment based on the confidence correction factor to obtain corrected confidence parameters; It is determined whether the corrected confidence parameter of each frame is lower than the preset confidence threshold. If so, the secondary annotation reminder of the frame is retained; if not, the original secondary annotation reminder is cancelled.

5. The artificial intelligence-based data annotation method according to claim 4, characterized in that: The confidence correction factor generation formula is: ,in refers to the confidence correction factor, Refers to the mean slope of the trend curve. Refers to the adjustment weight corresponding to the mean of the slope, refers to the absolute value of the mean of all deviation values, Refers to the adjustment weight corresponding to this absolute value; In the confidence correction factor generation formula: ,in Refers to the total number of local annotation deviation feedback data, where Refers to the The correction result of the local annotation deviation feedback data, Refers to the The initial annotation value of the local annotation deviation feedback data, Refers to the The deviation value between the initial annotation value corresponding to the local annotation deviation feedback data and the corresponding correction result.

6. A data annotation system based on artificial intelligence, characterized in that: The system includes: a data acquisition module, a data screening module, a slope calculation module, a deviation value calculation module and a correction factor generation module, wherein: The data acquisition module is used to analyze whether the current scene complexity level of the target time segment is higher than the corresponding threshold when the proportion of frames triggering secondary annotation reminders in the target time segment exceeds the threshold ratio. If so, the historical video processing records and historical annotation deviation feedback data are obtained; A data screening module is used to parse historical video processing records and screen out a number of historical comparison segment records that are within the same level range as the current scene complexity level and have a matching time length; The slope calculation module is used to parse each historical comparison segment record in turn, calculate the corresponding secondary annotation reminder frame ratio, draw a trend curve in chronological order, and calculate the slope mean of the trend curve; The deviation value calculation module is used to analyze the historical annotation deviation feedback data, obtain the deviation value between the initial annotation value and the correction result corresponding to each historical comparison segment record, and calculate the average value of all deviation values; The correction factor generation module is used to generate a confidence correction factor based on the mean of the slope and the mean of all deviation values ​​by weighting, and is used to adjust the confidence parameters of all frames in the target time segment.

7. The artificial intelligence-based data annotation system according to claim 6, characterized in that: The slope calculation module specifically includes: A ratio calculation unit is used to parse each historical comparison segment record in turn, calculate the ratio between the number of frames triggering the secondary marking reminder and the corresponding total number of frames in each historical comparison segment record, and obtain the respective secondary marking reminder frame ratio; A curve drawing unit, used to draw a corresponding trend curve with the time axis as the horizontal axis and the secondary marking reminder frame ratio as the vertical axis; The slope calculation unit is used to calculate the average slope of the trend curve.

8. The artificial intelligence-based data annotation system according to claim 7, characterized in that: The deviation value calculation module specifically includes: A local feedback data acquisition unit, used to parse the historical annotation deviation feedback data and extract the local annotation deviation feedback data corresponding to each historical comparison segment record; A local feedback data parsing unit is used to parse each local annotation deviation feedback data in turn, and extract the initial annotation value and the correction result therefrom, wherein the initial annotation value refers to all frames in each historical comparison segment record that trigger the secondary annotation reminder, and the correction result refers to the number of frames that are finally determined to require the secondary annotation reminder after manual confirmation or feedback; The deviation value mean calculation unit is used to calculate the deviation value between each group of initial marking values ​​and the corresponding correction result, and to perform statistics on the deviation values ​​corresponding to all historical comparison segment records to obtain the mean value of the deviation value.

9. The artificial intelligence-based data annotation system according to claim 8, characterized in that: The correction factor generation module specifically includes: A confidence correction factor generation unit is used to call a confidence correction factor generation formula, and use the average value of the slope of the trend curve and the average value of all deviation values ​​as input parameters to perform weighted calculation to generate a confidence correction factor; A confidence parameter correction unit, used to adjust the initial confidence parameters of all frames in the target time segment based on the confidence correction factor to obtain a corrected confidence parameter; The reminder retention determination unit is used to determine whether the revised confidence parameter of each frame is lower than a preset confidence threshold. If so, the secondary marking reminder of the frame is retained; if not, the original secondary marking reminder is cancelled.

10. The artificial intelligence-based data annotation system according to claim 9, characterized in that: The confidence correction factor generation formula is: ,in refers to the confidence correction factor, Refers to the mean slope of the trend curve. Refers to the adjustment weight corresponding to the mean of the slope, refers to the absolute value of the mean of all deviation values, Refers to the adjustment weight corresponding to this absolute value; In the confidence correction factor generation formula: ,in Refers to the total number of local annotation deviation feedback data, where Refers to the The correction result of the local annotation deviation feedback data, Refers to the The initial annotation value of the local annotation deviation feedback data, Refers to the The deviation value between the initial annotation value corresponding to the local annotation deviation feedback data and the corresponding correction result.

Images