A cloud computing-based security talent training resource sharing system and method

Abstract

The present application relates to the field of safety training, and more particularly to a cloud computing-based safety talent training resource sharing system and method, the method comprising: determining a test question acquisition method according to a historical training frequency corresponding to a generation request and an effective duration; when the test question acquisition method is based on calibration training test questions, determining calibration training test questions based on a label support proportion; determining a video feature difference degree according to a position change reference value of a first type of mark box in a test question video corresponding to the calibration training test questions; selecting preliminary training according to the video feature difference degree of the test questions corresponding to the calibration training test questions; when the test question acquisition method is based on rotation number, selecting training test questions according to the rotation number corresponding to the test questions; and the present application improves the selection accuracy of training test questions.

Description

Technical Field

[0001] This invention relates to the field of security training, and in particular to a cloud-based security talent training resource sharing system and method. Background Technology

[0002] Traditional construction safety training primarily relies on in-person lectures, distribution of printed manuals, or viewing of fixed safety education videos. These methods generally suffer from problems such as monotonous training formats, outdated content, and high resource consumption. Against this backdrop, online safety training platforms have emerged. These platforms integrate high-quality training resources and significantly improve training efficiency and coverage through standardized courses and online assessments. A common question format involves showing trainees a short video simulating or depicting a real construction site, requiring them to identify and point out violations, safety hazards, or management loopholes within the video. This approach simulates the real work environment more effectively than traditional text or image tests.

[0003] However, faced with a massive number of questions in the question bank, the selection of existing training questions is often a fixed and singular method, such as selecting training questions based on the number of times they have been selected in the past, which has a poor selection effect. In addition, existing technologies can select training questions based on the tags corresponding to the training questions, but the confirmation of tags is usually done manually, which is difficult and time-consuming. Therefore, how to improve the efficiency of safety training question selection is a problem that needs to be solved. Summary of the Invention

[0004] To address this, the present invention provides a cloud-based security talent training resource sharing system and method, which at least solves the problems in existing construction video safety training and examinations where, faced with a massive question bank, the selection method for training questions is singular and cannot be adjusted according to the characteristics of training videos with historical questions, resulting in poor adaptability of the selection results and difficulty in meeting user needs in terms of question selection effectiveness.

[0005] To achieve the above objectives, this invention provides a cloud computing-based method for sharing security talent training resources, comprising:

[0006] The method for obtaining test questions is determined based on the historical training sessions and effective duration corresponding to the generated request.

[0007] When acquiring test questions based on calibrated training test questions, the selection of calibrated training test questions is based on the proportion of tag support, the evaluation value, or the learning degree under the tag.

[0008] Based on the reference value of the positional change of the first type of selection box in the test video corresponding to the calibration training test, the video feature difference is determined as the difference in the number of influences of the second type of selection box or the difference in the surrounding area of ​​the first type of selection box.

[0009] The initial training questions are selected based on the difference in video features between the questions to be selected and the calibrated training questions, and the number of initial training questions determines whether to perform question compensation.

[0010] When the test questions are obtained based on the number of rotations, the test questions for the current training session are selected according to the number of rotations corresponding to the test questions to be selected.

[0011] Furthermore, if the number of historical training sessions corresponding to the generated request is greater than or equal to a threshold set for the number of historical training sessions, or the effective duration is greater than or equal to a threshold set for the effective duration, then the test questions will be obtained based on the calibrated training test questions.

[0012] Furthermore, the methods for confirming the training test questions include:

[0013] The percentage of tags supported by the historical answer information corresponding to the generated request is detected.

[0014] If the percentage of labels supported is less than the threshold set for the percentage of labels supported, then the training test questions will be historical test questions whose processing evaluation value is greater than the threshold set for the processing evaluation value.

[0015] If the percentage of tags supported is greater than or equal to the threshold set for the percentage of tags supported, then the training questions will be historical questions under the tags whose learning degree is less than the threshold set for the learning degree.

[0016] Furthermore, the evaluation value is determined based on the processing time and number of reviews for the historical test questions;

[0017] The processing evaluation value is positively correlated with the processing time and the number of reviews.

[0018] Furthermore, test item acquisition is based on the calibration training test items, including:

[0019] The candidate training questions whose video feature difference is greater than the set difference threshold are recorded as the initial training questions. The required number of initial training questions are selected as the training questions for the current training session in descending order of video feature difference.

[0020] If the number of initial training test questions is less than the required number, test question compensation is required.

[0021] Furthermore, test item compensation includes:

[0022] Historical test questions with error rates exceeding the set error rate threshold under the detection label are randomly selected as the test questions for the current training session, with a compensation quantity among them.

[0023] Furthermore, the method for confirming the video feature difference is as follows:

[0024] The position change reference value corresponding to the first type of bounding box is detected. If the position change reference value is greater than the set threshold of the position change reference value, the video feature difference is the difference in the number of influences of the second type of bounding box.

[0025] If the position change reference value is less than or equal to the position change reference value set threshold, then the video feature difference is the difference in the area surrounding the selection box of a class.

[0026] Furthermore, if the number of historical training sessions corresponding to the generated request is less than the set threshold for the number of historical training sessions and the effective duration is less than the set threshold for the effective duration, then the test questions will be obtained based on the number of rounds.

[0027] Furthermore, question acquisition is based on the number of rounds, including:

[0028] The required number of test questions are selected as the test questions for the current training session, based on the descending order of the number of rotations.

[0029] The present invention also provides a system for applying the cloud computing-based security talent training resource sharing method, comprising:

[0030] The request analysis unit is used to determine the test question acquisition method based on the historical training number and effective duration corresponding to the uploaded generation request, whether it is based on the calibrated training test questions or the number of rotations.

[0031] The historical analysis unit, which is connected to the request analysis unit, is used to determine the selection of calibration training questions based on the label support ratio, the processing evaluation value, or the learning degree of the label.

[0032] The video analysis unit, which is connected to the historical analysis unit, is used to determine the video feature difference degree as the difference in the number of influences of the second type of label boxes or the difference in the surrounding area of ​​the first type of label boxes based on the reference value of the position change of the first type of label box in the test video corresponding to the calibration training test questions.

[0033] The test question search unit, which is connected to the request analysis unit and the video analysis unit respectively, is used to select initial training test questions based on the difference in video features between the test questions to be selected and the calibration training test questions, and to determine whether to perform test question compensation based on the number of initial training test questions, or to select the training test questions for the current session based on the number of rotations corresponding to the test questions to be selected.

[0034] Compared with the prior art, the beneficial effect of the present invention is that it uses the number of historical training sessions and the effective duration as the criteria for determining the test question acquisition method. Based on the number of historical training sessions and the effective duration, it represents the historical answer accumulation of the test takers. When the number of historical training sessions is greater than or equal to a set threshold or the effective duration is greater than or equal to a set threshold, it is determined that the test takers' historical answer situation has reference value, and test questions are acquired based on the calibrated training test questions accordingly. This avoids the problem that the test question acquisition results are difficult to meet the training needs due to the single test question acquisition based on the number of rotations, thereby improving the test question acquisition efficiency of the present invention.

[0035] Furthermore, in the method for confirming calibration training questions in this invention, the completeness of the tags corresponding to the current historical questions is represented by the tag support ratio. This avoids the problem of tag issues such as missing tags corresponding to the question videos causing subsequent question acquisition failures. Different calibration training questions are selected for confirmation based on the tag support ratio, making the calibration training questions more effective and thereby improving the accuracy of subsequent question acquisition based on the calibration training questions.

[0036] Furthermore, the present invention uses a method to confirm the video feature differences based on the positional change reference values ​​corresponding to a class of selection boxes, which makes the determination of video differences more consistent with actual video features and avoids the problem of poor test question acquisition accuracy caused by the single matching method in the prior art. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the cloud computing-based security talent training resource sharing method of the present invention;

[0038] Figure 2 This is a flowchart illustrating how the test questions are obtained based on the number of historical training sessions and the effective duration of the invention.

[0039] Figure 3 This is a flowchart illustrating the process of selecting calibration training questions based on the evaluation value or the learning degree of the tag under the label, according to the determination of the tag support ratio in this invention.

[0040] Figure 4 This is a flowchart illustrating the process of determining whether to perform test question compensation based on the number of initial selection training test questions in this invention.

[0041] Figure 5 This is a schematic diagram of the cloud-based security talent training resource sharing system of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0044] Please see Figures 1 to 4 As shown, this invention provides a cloud computing-based method for sharing security talent training resources, including:

[0045] The method for obtaining test questions is determined based on the historical training sessions and effective duration corresponding to the generated request.

[0046] When acquiring test questions based on calibrated training test questions, the selection of calibrated training test questions is based on the proportion of tag support, the evaluation value, or the learning degree under the tag.

[0047] Based on the reference value of the positional change of the first type of selection box in the test video corresponding to the calibration training test, the video feature difference is determined as the difference in the number of influences of the second type of selection box or the difference in the surrounding area of ​​the first type of selection box.

[0048] The initial training questions are selected based on the difference in video features between the questions to be selected and the calibrated training questions, and the number of initial training questions determines whether to perform question compensation.

[0049] When the test questions are obtained based on the number of rotations, the test questions for the current training session are selected according to the number of rotations corresponding to the test questions to be selected.

[0050] This invention is applied to safety training for construction workers. Specifically, the training involves displaying test questions and corresponding videos (simulated or real construction site videos) to participants via a smart terminal. Participants are required to identify and point out any violations, safety hazards, or management loopholes in the videos. The request to generate questions is uploaded by the participant to the request analysis unit via a smart terminal, which may include, but is not limited to, a mobile phone or computer. The request to generate questions is the instruction to generate the test questions. Upon receiving the request, the request analysis unit retrieves the participant's personnel information and historical answer information from the shared cloud storage unit. The personnel information includes at least... This includes the number of training sessions and their effective duration for participants. Historical answer information includes the most recent answer records and the percentage of tags supported. Each answer record includes the questions, videos, and whether the participant's answers were correct. The questions corresponding to these answer records are recorded as historical questions. The number of answer records included in the historical answer information is set by the administrator. It is understood that the larger the number of answer records, the higher the accuracy of subsequent analysis. This will not be elaborated on here. It is worth noting that the number of answer records should be greater than or equal to the threshold set for the number of historical training sessions.

[0051] Specifically, if the number of historical training sessions corresponding to the generated request is greater than or equal to the historical training session threshold or the effective duration is greater than or equal to the effective duration threshold, then the test questions will be obtained based on the calibrated training test questions.

[0052] The historical training count refers to the number of times the respondent has undergone security training in the past process corresponding to the generated request. The effective duration is the sum of the time intervals between each adjacent security training session. It is easy to understand that setting thresholds for the historical training count and the effective duration can prevent the respondent from undergoing high-frequency security training in a short period of time, which could lead to the problem of malicious acquisition of test questions. A specific value is provided: the historical training count threshold is set to 10 times, and the effective duration threshold is set to 72 hours.

[0053] Specifically, the confirmation methods for the calibration training test questions include:

[0054] The percentage of tags supported by the historical answer information corresponding to the generated request is detected.

[0055] If the percentage of labels supported is less than the threshold set for the percentage of labels supported, then the training test questions will be historical test questions whose processing evaluation value is greater than the threshold set for the processing evaluation value.

[0056] If the percentage of tags supported is greater than or equal to the threshold set for the percentage of tags supported, then the training questions will be historical questions under the tags whose learning degree is less than the threshold set for the learning degree.

[0057] Tag support percentage = Number of first test questions / Number of first test questions + Number of second test questions. The number of first test questions is the number of historical test questions with tags in the answer records corresponding to the historical answer information, and the number of second test questions is the number of historical test questions without tags in the answer records corresponding to the historical answer information.

[0058] Setting a threshold for the tag support ratio is understandable because the tag support ratio reflects the referenceability of tags in historical test questions. This avoids the problem of poor accuracy in selecting calibration training questions when the number of tags is too small to meet the requirements for calibration training questions based on the learning degree of tags. Therefore, the higher the accuracy requirement of the calibration training questions obtained based on the learning degree of tags, the larger the threshold for setting the tag support ratio. One possible value is 80%, where the management personnel are the backend management personnel of the system described in this invention.

[0059] Specifically, the evaluation value is determined based on the processing time and number of reviews for the historical test questions;

[0060] The processing evaluation value is positively correlated with the processing time and the number of reviews.

[0061] In this embodiment of the invention, the processing evaluation value for a single historical question is calculated as: Processing Time / Preset Processing Time + Number of Reviews / Preset Number of Reviews. During security training, when a participant finishes processing a single question, they can click the "Complete" button to end the question. Furthermore, outside of security training hours, participants can review historical questions using the "Review" button. This is already known to those skilled in the art and will not be elaborated upon here. The processing time refers to the time elapsed from the last time the historical question was displayed on the smart terminal during the participant's past security training sessions until they clicked the "Complete" button. The number of reviews refers to the total number of times the participant has reviewed the historical question up to the current moment.

[0062] The preset processing time is the average of the processing time for each historical question, and the preset number of reviews is the average number of reviews for each historical question.

[0063] For a single historical test question, the learning degree under its corresponding label is the number of other historical test questions with the same label as that historical test question. The learning degree threshold is determined by detecting the labels corresponding to each historical test question and recording the average number of historical test questions corresponding to each label as the learning degree threshold.

[0064] Specifically, test questions are acquired based on the calibration training test questions, including:

[0065] The candidate training questions whose video feature difference is greater than the set difference threshold are recorded as the initial training questions. The required number of initial training questions are selected as the training questions for the current training session in descending order of video feature difference.

[0066] If the number of initial training test questions is less than the required number, test question compensation is required.

[0067] The test questions for this training session are recorded as the test questions obtained corresponding to the generation request.

[0068] In this embodiment of the invention, the required quantity is 25. The required quantity is set by the management personnel. It can be understood that the greater the training intensity requirement, the greater the required quantity. This is something that those skilled in the art have already mastered and does not need to be elaborated.

[0069] Specifically, test question compensation includes:

[0070] Historical test questions with error rates exceeding the set error rate threshold under the detection label are randomly selected as the test questions for the current training session, with a compensation quantity among them.

[0071] For a single historical question, the error rate under its corresponding label is the number of other historical questions that were answered incorrectly in the historical process and have the same label as that historical question. The method for determining the error rate threshold is to detect the labels corresponding to each historical question and record the average number of historical questions answered incorrectly in the historical process corresponding to each label as the error rate threshold.

[0072] Compensation quantity = Demand quantity - Number of initial training test questions.

[0073] Specifically, the method for confirming video feature differences is as follows:

[0074] The position change reference value corresponding to the first type of bounding box is detected. If the position change reference value is greater than the set threshold of the position change reference value, the video feature difference is the difference in the number of influences of the second type of bounding box.

[0075] If the position change reference value is less than or equal to the position change reference value set threshold, then the video feature difference is the difference in the area surrounding the selection box of a class.

[0076] In this embodiment of the invention, a pre-trained YOLOv8 model is used to identify construction workers and construction equipment. One type of bounding box is the identification box for construction workers, and the other type of bounding box is the identification box for construction equipment. The identification boxes are rectangular. Construction equipment includes, but is not limited to, cranes, dynamic compaction machines, and transport vehicles. This is content that is easy for those skilled in the art to understand and will not be described in detail here.

[0077] The system detects the calibration training questions on which the current test questions are based. It then uniformly extracts video frames from the corresponding test video to obtain several calibration video frames. Each calibration video frame is labeled with a first-class bounding box. The sub-position reference value for each calibration video frame is calculated. The absolute value of the difference between the maximum and minimum sub-position reference values ​​is recorded as the position change reference value for the first-class bounding box of that calibration video frame. For a single calibration video frame, the sub-position reference value is the area of ​​the smallest circle within that frame that can encompass all first-class bounding boxes. The number of calibration video frames is 10. It is understood that the higher the accuracy requirement of the sub-position reference values ​​for management personnel, the larger the number of calibration video frames will be.

[0078] By setting a threshold for the position change reference value, it is known that the larger the position change reference value, the greater the range of overall positional movement of the construction personnel within the test video. Therefore, it is not possible to use the characteristics of a type of checkbox as the selection of video feature difference. Instead, the distribution characteristics of the construction equipment are chosen as the selection of video feature difference. The higher the accuracy requirement of the management personnel for the difference in the surrounding area of ​​a type of checkbox as the video feature difference, the smaller the threshold value for the position change reference value should be. This invention provides a value where the threshold value for the position change reference value is 30% of the video frame area. This invention avoids the problem of low accuracy caused by solely judging the video feature difference based on the characteristics of the construction personnel.

[0079] The difference in area around a type of selection box is defined as |S1-S2|. For the test questions to be selected, video frames are extracted uniformly to obtain several candidate video frames. The number of candidate video frames is the same as the number of calibration video frames. The average area of ​​the smallest circle that can cover all type-1 selection boxes in each candidate video frame is recorded as S1. The average area of ​​the smallest circle that can cover all type-1 selection boxes in the calibration video frame is recorded as S2.

[0080] The difference in the number of Class II selection boxes is calculated as |N1-N2|. For a single candidate video frame, the number of Class II selection boxes whose distance from the Class I selection box is less than a preset distance is detected and recorded as the number of sub-influences. The average number of sub-influences corresponding to each candidate video frame is recorded as N1. The method for confirming the number of sub-influences N2 corresponding to the selected video frame is that the candidate video frames are the same, which will not be elaborated further.

[0081] The preset distance is 50m. It can be understood that the smaller the distance between the Class II and Class I marker selection boxes, the greater the impact of the construction equipment on the construction personnel. Managers can adjust the preset distance adaptively according to training needs.

[0082] The threshold for the difference in the number of influences of the second-class selection boxes is set to 5. The threshold for the difference in the surrounding area of ​​the first-class selection boxes is set to 15% of the area of ​​the calibrated video frame. Administrators can set these thresholds according to actual needs. It is understood that the larger the difference in the number of influences of the second-class selection boxes or the larger the threshold for the surrounding area difference, the greater the difference in video features. Therefore, the greater the diversity of matching questions, the better it meets the diverse training needs of the questions, avoiding poor training results caused by a single type of question. Thus, the higher the user's demand for training diversity, the larger the values ​​of the thresholds for the number of influences and the surrounding area difference.

[0083] Specifically, if the number of historical training sessions corresponding to the generated request is less than the set threshold for the number of historical training sessions and the effective duration is less than the set threshold for the effective duration, then the test questions will be obtained based on the number of rounds.

[0084] Specifically, question acquisition is based on the number of rounds, including:

[0085] The required number of test questions are selected as the test questions for the current training session, based on the descending order of the number of rotations.

[0086] For a single candidate question, the corresponding rotation number is the total number of generation requests received by the request analysis unit between the current moment of the candidate video and the most recent time it was selected as a training question.

[0087] Please see Figure 5 As shown, this is a schematic diagram of the cloud-based security talent training resource sharing system of the present invention. The present invention provides a cloud-based security talent training resource sharing system applying the aforementioned cloud-based security talent training resource sharing method, comprising:

[0088] The request analysis unit is used to determine the test question acquisition method based on the historical training number and effective duration corresponding to the uploaded generation request, whether it is based on the calibrated training test questions or the number of rotations.

[0089] The historical analysis unit, which is connected to the request analysis unit, is used to determine the selection of calibration training questions based on the label support ratio, the processing evaluation value, or the learning degree of the label.

[0090] The video analysis unit, which is connected to the historical analysis unit, is used to determine the video feature difference degree as the difference in the number of influences of the second type of label boxes or the difference in the surrounding area of ​​the first type of label boxes based on the reference value of the position change of the first type of label box in the test video corresponding to the calibration training test questions.

[0091] The test question search unit, which is connected to the request analysis unit and the video analysis unit respectively, is used to select initial training test questions based on the difference in video features between the test questions to be selected and the calibration training test questions, and to determine whether to perform test question compensation based on the number of initial training test questions, or to select the training test questions for the current session based on the number of rotations corresponding to the test questions to be selected.

[0092] The system also includes a shared cloud storage unit, which is connected to the request analysis unit and the historical analysis unit to store personnel information and historical answer information of each trainee, as well as test questions and corresponding test video.

[0093] Users of this cloud-based security talent training resource sharing system can upload test questions and corresponding test question videos to the shared cloud storage unit. Users can also label the test questions, including but not limited to fire safety violations, electrical safety violations, mechanical safety violations, or high-altitude operation violations. However, due to issues such as users' omission of labels, it cannot be guaranteed that all test questions have corresponding labels.

[0094] Finally, it should be noted that, for the apparatus embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and relevant details can be found in the description of the method embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0095] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, in implementing the present invention, the functions of each unit can be implemented in one or more software and / or hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present invention.

[0096] Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for sharing security talent training resources based on cloud computing, characterized in that, include: For the uploaded generation request, the method of obtaining test questions is determined based on the historical training frequency and effective duration of the personnel information corresponding to the generation request; When acquiring test questions based on calibrated training test questions, the selection of calibrated training test questions is based on the proportion of tag support, the evaluation value, or the learning degree under the tag. Based on the reference value of the positional change of the first type of selection box in the test video corresponding to the calibration training test, the video feature difference is determined as the difference in the number of influences of the second type of selection box or the difference in the surrounding area of ​​the first type of selection box. The initial training questions are selected based on the difference in video features between the questions to be selected and the calibrated training questions, and the number of initial training questions determines whether to perform question compensation. Test question compensation includes: Historical test questions with error rates exceeding the error rate threshold under the detection tag are randomly selected as test questions for the current training session, and the number of historical test questions with compensation quantities is randomly selected. When the test questions are obtained based on the number of rotations, the test questions for the current training session are selected according to the number of rotations corresponding to the test questions to be selected. The confirmation methods for calibration training test questions include: The percentage of tags supported by the historical answer information corresponding to the generated request is detected. If the percentage of labels supported is less than the threshold set for the percentage of labels supported, then the training test questions will be historical test questions whose processing evaluation value is greater than the threshold set for the processing evaluation value. If the percentage of tags supported is greater than or equal to the threshold set for the percentage of tags supported, then the training questions will be historical questions under the tags whose learning degree is less than the threshold set for the learning degree. Test item acquisition is based on calibration training test items, including: The candidate training questions whose video feature difference is greater than the set difference threshold are recorded as the initial training questions. The required number of initial training questions are selected as the training questions for the current training session in descending order of video feature difference. If the number of initial training test questions is less than the required number, test question compensation is required. The method for confirming video feature differences is as follows: The position change reference value corresponding to the first type of bounding box is detected. If the position change reference value is greater than the set threshold of the position change reference value, the video feature difference is the difference in the number of influences of the second type of bounding box. If the position change reference value is less than or equal to the position change reference value set threshold, then the video feature difference is the difference in the area surrounding the first type of bounding box. The first type of selection box is for identifying construction personnel, and the second type of selection box is for identifying construction equipment; The difference in area around a type of selection box is |S1-S2|. For the test questions to be selected, video frames are extracted uniformly to obtain a number of candidate video frames. The number of candidate video frames is the same as the number of calibration video frames. The average area of ​​the smallest circle that can cover all type-1 selection boxes in each candidate video frame is detected and denoted as S1. The average area of ​​the smallest circle that can cover all type-1 selection boxes in the calibration video frame is denoted as S2. The difference in the number of Class II label selection boxes is calculated as |N1-N2|. For a single candidate video frame, the number of Class II label selection boxes whose distance from the Class I label selection box is less than a preset distance is detected and recorded as the number of sub-influences. The average number of sub-influences corresponding to each candidate video frame is recorded as N1, and the number of sub-influences corresponding to the labeled video frame is recorded as N2.

2. The cloud-based security talent training resource sharing method according to claim 1, characterized in that, If the number of historical training sessions corresponding to the generated request is greater than or equal to the historical training session threshold or the effective duration is greater than or equal to the effective duration threshold, then the test questions will be obtained based on the calibrated training test questions.

3. The cloud-based security talent training resource sharing method according to claim 2, characterized in that, The evaluation value is determined based on the processing time and number of reviews for the historical test questions. The processing evaluation value is positively correlated with the processing time and the number of reviews.

4. The cloud-based security talent training resource sharing method according to claim 3, characterized in that, If the number of historical training sessions corresponding to the generated request is less than the set threshold for the number of historical training sessions and the effective duration is less than the set threshold for the effective duration, then the test questions will be obtained based on the number of rounds.

5. The cloud-based security talent training resource sharing method according to claim 4, characterized in that, Question acquisition based on round count includes: The required number of test questions are selected as the test questions for the current training session, based on the descending order of the number of rotations.

6. A system applying the cloud computing-based security talent training resource sharing method according to any one of claims 1 to 5, characterized in that, include: The request analysis unit is used to determine the test question acquisition method based on the historical training number and effective duration corresponding to the uploaded generation request, whether it is based on the calibrated training test questions or the number of rotations. The historical analysis unit, which is connected to the request analysis unit, is used to determine the selection of calibration training questions based on the label support ratio, the processing evaluation value, or the learning degree of the label. The video analysis unit, which is connected to the historical analysis unit, is used to determine the video feature difference degree as the difference in the number of influences of the second type of label boxes or the difference in the surrounding area of ​​the first type of label boxes based on the reference value of the position change of the first type of label box in the test video corresponding to the calibration training test questions. The test question search unit, which is connected to the request analysis unit and the video analysis unit respectively, is used to select initial training test questions based on the difference in video features between the test questions to be selected and the calibration training test questions, and to determine whether to perform test question compensation based on the number of initial training test questions, or to select the training test questions for the current session based on the number of rotations corresponding to the test questions to be selected.

Images