An online examination invigilation system based on image recognition

Abstract

The present invention relates to the field of image recognition technology, and in particular to an online examination invigilation system based on image recognition, comprising: an image acquisition unit; an image analysis unit for determining the video state of a video to be analyzed based on a video duration reference value and a blur reference value, and determining an analysis method based on the video state of the video to be analyzed; an association processing unit for determining the association state of blurry frames based on a distance interval reference value between blurry frames, and determining a frame extraction point setting method based on the association state of the blurry frames; an adjustment analysis unit for determining the video frame category based on a limb point change degree and a limb point change quantity reference value; and an adjustment processing unit for determining an adjustment method based on a video frame aggregation area. The present invention can optimize the video frame selection position in a physical examination with complex movements, so that the number of video frames selected and the image quality effectively meet the examination evaluation requirements.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to an online examination invigilation system based on image recognition. Background Art

[0002] With the rapid development of educational informatization, online examinations have become the new normal. For online physical education exams such as aerobics exams, in an online proctoring environment, due to the need to monitor the examination status of multiple candidates at the same time, system resources and the energy of the proctors are limited. Therefore, it is impossible to monitor every movement of each candidate in real time and in detail. In order to overcome the above limitations, the proctors will select representative video frames for evaluation, but due to the continuity and dynamics of aerobics movements, the selected video frames often have defects such as blurred frames and weak representativeness, which cannot fully reflect the test level of the candidates. Therefore, how to optimize the selection position of video frames in the aerobics exam and select video frames that can reflect the test level of the candidates is a technical problem that needs to be solved urgently by those skilled in the art.

[0003] Chinese patent publication number CN117975333A discloses a physical education examination system for testing students' academic proficiency, including: a video and image acquisition device to capture video data of the examinee during the physical education examination, the shooting range of the video and image acquisition device covering the entire area of the score detection mat; the score detection mat measures the examinee's test score during the examination, and different score lines are marked on the score detection mat; an edge computing device is used to receive the video data captured by the video and image acquisition device, and analyze the video data according to the detection algorithm to obtain the test results, the detection algorithm includes a line stepping detection algorithm, a landing detection algorithm, a human posture detection algorithm and a score calculation, and the test results include fouls and normal scores; a result output device is used to output the test results. It can be seen that the above technical solution has the following problems: the video frame selection position in the physical education examination with complex movements is not optimized, and the number of video frames selected and the image quality of the video frames cannot effectively meet the test evaluation requirements. Summary of the Invention

[0004] To this end, the present invention provides an online examination invigilation system based on image recognition, which is used to overcome the problems in the prior art that the video frame selection position in sports examinations with complex movements is not optimized, and the number of selected video frames and the image quality of the video frames cannot effectively meet the examination evaluation requirements.

[0005] To achieve the above objectives, the present invention provides an online examination invigilation system based on image recognition, comprising:

[0006] An image acquisition unit, used to acquire video frame information of the video to be analyzed;

[0007] an image analysis unit connected to the image acquisition unit, configured to determine a video state of the video to be analyzed based on a video duration reference value and a fuzzy reference value, and to determine an analysis method based on the video state of the video to be analyzed;

[0008] an association processing unit connected to the image acquisition unit and the image analysis unit, configured to determine an association state of the blurred frames according to a reference value of a distance interval between the blurred frames, and determine a frame extraction point setting method according to the association state of the blurred frames;

[0009] an adjustment and analysis unit connected to the image acquisition unit and the image analysis unit, for determining a video frame category according to a limb point change degree and a limb point change quantity reference value;

[0010] An adjustment processing unit is connected to the adjustment analysis unit and is used to determine an adjustment method based on the video frame aggregation area, which is to adjust the number of frame sampling points in the first frame aggregation area according to the similarity of the second frame images closest to the two ends of the first frame aggregation area, or to adjust the frame sampling point interval of the video to be analyzed according to the length similarity of the second frame aggregation area.

[0011] Furthermore, the image analysis unit determines an analysis method according to the video state of the video to be analyzed, wherein, in the first video state, the analysis method is a first analysis method that determines a frame extraction point setting method according to the association state of the blurred frame; in the second video state, the analysis method is a second analysis method that determines an adjustment method according to a first-class frame aggregation area and a second-class frame aggregation area.

[0012] Furthermore, the image analysis unit determines the video state of the video to be analyzed based on the video duration reference value and the fuzzy reference value;

[0013] The video status of the video to be analyzed includes a first video status in which the video length reference value is greater than a preset video length reference value or the blur reference value is greater than a preset blur reference value, and a second video status in which the video length reference value is less than or equal to the preset video length reference value and the blur reference value is less than or equal to the preset blur reference value.

[0014] Furthermore, the association processing unit responds to the first analysis mode and determines a frame extraction point setting mode according to the association state of the fuzzy frames;

[0015] If the association state is the first preset association state, the association analysis unit determines that the frame extraction point setting method is to segment the video to be analyzed according to the fuzzy frame position to obtain a plurality of associated fuzzy frame segments, and perform frame extraction point compensation on each associated fuzzy frame segment;

[0016] If the association state is the second preset association state, the association analysis unit determines that the frame extraction point setting method is to perform frame extraction point compensation for the video segment between adjacent blurred frames.

[0017] Furthermore, the association processing unit performs segmentation processing on the video to be analyzed according to the fuzzy frame positions to obtain a plurality of associated fuzzy frame segments, and a time interval reference value between each fuzzy frame in any single associated fuzzy frame segment is less than a preset time interval reference value;

[0018] The segmentation processing includes: performing segmentation analysis on each blurred frame in sequence according to the time sequence of the video to be analyzed; when performing segmentation analysis on a single blurred frame, recording the blurred frame as a target blurred frame; detecting the time interval reference value between each blurred frame after the time sequence of the target blurred frame; if the time interval reference value between the blurred frame after the time sequence of the target blurred frame and the blurred frame is less than a preset time interval reference value, recording it as an associated blurred frame of the target blurred frame; if the time interval reference value between the blurred frame after the time sequence of the target blurred frame and the blurred frame is greater than or equal to the preset time interval reference value, stopping the segmentation analysis of the target blurred frame, and uniformly recording the target blurred frame and its corresponding associated blurred frames as an associated blurred frame segment.

[0019] Furthermore, the association processing unit determines the association state of the blurred frames according to the distance interval reference value between the blurred frames;

[0020] The first preset association state is that there are a number of adjacent blurred frames in the video to be analyzed whose distance interval reference value is equal to the frame sampling point interval;

[0021] The second preset association state is that the reference values of the distance intervals of the blurry frames in the video to be analyzed are all greater than the frame extraction point interval;

[0022] The frame sampling point interval is the time length between two adjacent frame sampling points in the video to be analyzed.

[0023] Furthermore, the adjustment analysis unit determines the video frame category based on the limb point change degree and the limb point change quantity reference value. The video frame category includes a type of frame in which the limb point change degree is less than or equal to the preset limb point change degree and the limb point change quantity reference value is less than or equal to the preset limb point change quantity reference value, and a type of frame in which the limb point change degree is greater than the preset limb point change degree or the limb point change quantity reference value is greater than the preset limb point change quantity reference value.

[0024] Furthermore, the adjustment processing unit determines an adjustment method according to the video frame aggregation area in response to the second analysis method, the adjustment method including a first adjustment method for adjusting the number of frame extraction points in the first frame aggregation area according to the similarity between the two types of frame images closest to both ends of the first frame aggregation area, and a second adjustment method for adjusting the frame extraction point interval of the video to be analyzed according to the length similarity between the two types of frame aggregation areas;

[0025] The video frame aggregation area includes a first-class frame aggregation area and a second-class frame aggregation area. The first-class frame aggregation area and the second-class frame aggregation area are confirmed as follows:

[0026] When multiple consecutive adjacent video frames are all class I frames, these video frames are recorded as target class I frames, and the class I frame aggregation area is the video area between the starting target class I frame and the ending target class I frame in the time sequence;

[0027] When multiple consecutive adjacent video frames are all Class II frames, these video frames are recorded as target Class II frames, and the Class II frame aggregation area is the video area between the starting target Class II frame and the ending target Class II frame in time sequence.

[0028] Furthermore, the adjustment processing unit responds to the first adjustment mode and increases or decreases the number of frame extraction points in the first frame aggregation area according to the similarity between the second frame images closest to both ends of the first frame aggregation area;

[0029] There is a negative correlation between the number of frame extraction points in the first-class frame aggregation area and the similarity between the second-class frame images at both ends of the first-class frame aggregation area.

[0030] Furthermore, the adjustment processing unit responds to the second adjustment mode, and if there are multiple second-type frame clustering regions with length similarity greater than a preset length similarity in the video to be analyzed, the frame sampling point interval of the video to be analyzed is reduced;

[0031] The relationship between the reduction value of the frame extraction point interval and the representativeness of the second type of frame image is positively correlated.

[0032] Compared with the prior art, the beneficial effect of the present invention lies in that, in the technical solution of the present invention, the image analysis unit effectively reflects the video status of the video to be analyzed based on the video length reference value and the blur reference value, and effectively reflects the complexity of the physical examination movements in the video to be analyzed through the video status of the video to be analyzed, and adaptively selects different analysis methods, so that the selection of analysis method is more in line with the actual work scenario, avoiding the problem that the number of selected video frames and the image quality cannot effectively meet the test evaluation requirements due to the video to be analyzed being too long and having too many blurry frames, thereby improving the accuracy and representativeness of the test scores.

[0033] Furthermore, the association processing unit in the present invention determines the association status of the blurred frames based on the reference value of the distance interval between each blurred frame, effectively reflects the distribution of blurred frames in the video to be analyzed according to the association status of the blurred frames, and selects different frame extraction point setting methods according to actual conditions, thereby avoiding the problem that the number of video frames cannot effectively meet the actual examination requirements, thereby improving the accuracy of physical examination scoring.

[0034] Furthermore, the associated processing unit in the present invention performs segmented processing on the video to be analyzed according to the position of the fuzzy frame, ensuring that when performing frame extraction point compensation on the associated fuzzy frame segment, different numbers of frame extraction points can be set according to the adaptive selection of the length of the associated fuzzy frame segment, thereby making the number of selected video frames effectively meet the test evaluation requirements.

[0035] Furthermore, the adjustment analysis unit in the present invention determines the video frame category based on the limb point change degree and the limb point change quantity reference value, determines the first type of frame aggregation area and the second type of frame aggregation area through the video frame category, and determines different adjustment methods based on the first type of frame aggregation area and the second type of frame aggregation area, thereby avoiding the problem that the selection position of the video frame cannot effectively meet the test evaluation requirements, thereby improving the representativeness of the video frame.

[0036] Furthermore, the adjustment processing unit in the present invention increases or decreases the number of frame extraction points in the first type of frame aggregation area according to the similarity of the second type of frame images closest to both ends of the first type of frame aggregation area, so that the system can flexibly adjust the number of frame extraction points according to the similarity of the second type of frame images closest to both ends of the first type of frame aggregation area, ensuring that the repeated extraction of redundant information can be avoided when the similarity of the two types of frame images is high, and also ensuring that the representative information of the examinee's gymnastics is not omitted when the similarity of the two types of frame images is low, thereby improving the representativeness of the video frame selection.

[0037] Furthermore, the adjustment processing unit in the present invention reduces the frame sampling point interval of the video to be analyzed, increases the number of selected video frames, avoids the problem of insufficient representativeness of video frames caused by the high similarity of the two types of frame clustering areas in the video to be analyzed, and thus effectively meets the test evaluation requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a unit connection diagram of the online examination invigilation system based on image recognition of the present invention;

[0039] Figure 2 This is a flow chart of the present invention for determining an analysis method according to the video status of a video to be analyzed;

[0040] Figure 3 This is a flow chart of the present invention for determining the video state of a video to be analyzed based on a video duration reference value and a fuzzy reference value;

[0041] Figure 4 This is a flow chart of the present invention for determining an adjustment method based on a video frame aggregation area. DETAILED DESCRIPTION

[0042] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0043] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0044] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0045] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0046] See also Figures 1 to 4 As shown, the present invention provides an online examination invigilation system based on image recognition, comprising:

[0047] An image acquisition unit, used to acquire video frame information of the video to be analyzed;

[0048] an image analysis unit connected to the image acquisition unit, configured to determine a video state of the video to be analyzed based on a video duration reference value and a fuzzy reference value, and to determine an analysis method based on the video state of the video to be analyzed;

[0049] an association processing unit connected to the image acquisition unit and the image analysis unit, configured to determine an association state of the blurred frames according to a reference value of a distance interval between the blurred frames, and determine a frame extraction point setting method according to the association state of the blurred frames;

[0050] an adjustment and analysis unit connected to the image acquisition unit and the image analysis unit, for determining a video frame category according to a limb point change degree and a limb point change quantity reference value;

[0051] An adjustment processing unit is connected to the adjustment analysis unit and is used to determine an adjustment method based on the video frame aggregation area, which is to adjust the number of frame sampling points in the first frame aggregation area according to the similarity of the second frame images closest to the two ends of the first frame aggregation area, or to adjust the frame sampling point interval of the video to be analyzed according to the length similarity of the second frame aggregation area.

[0052] The present invention is applied to online examinations for prescribed movements in physical education exams. The initial frame extraction rule is as follows: A frame extraction point is set every t seconds in chronological order for the video to be analyzed. The value of t can be set by the user based on actual needs. The greater the user's requirement for video scoring accuracy, the smaller the value of t. A value of t is provided, t = 10. For a single frame extraction point, the video image extracted at that extraction point is the video frame corresponding to that extraction point. A blurred frame is a video frame with a blur greater than a preset blurriness. Each video frame is converted to a grayscale image, and then a Laplacian operator template is used to convolve the grayscale image to obtain a new image. The sum of the absolute values of each pixel in the new image is used as the blurriness. The preset blurriness value can be determined by the user based on historical usage records. The greater the user's requirement for video scoring accuracy, the larger the preset blurriness value. A preset blurriness value is provided, which is the maximum blurriness of blurred frames in videos whose scoring accuracy meets the user's requirements during historical usage.

[0053] Specifically, the image analysis unit determines the analysis method according to the video state of the video to be analyzed, wherein, in the first video state, the analysis method is a first analysis method that determines the frame extraction point setting method according to the association state of the blurred frame; in the second video state, the analysis method is a second analysis method that determines the adjustment method according to the first type of frame aggregation area and the second type of frame aggregation area.

[0054] Specifically, the image analysis unit determines the video state of the video to be analyzed according to the video duration reference value and the fuzzy reference value;

[0055] The video status of the video to be analyzed includes a first video status in which the video length reference value is greater than a preset video length reference value or the blur reference value is greater than a preset blur reference value, and a second video status in which the video length reference value is less than or equal to the preset video length reference value and the blur reference value is less than or equal to the preset blur reference value.

[0056] The video duration reference value is the time length of the video to be analyzed, and the blur reference value = the number of blur frames of the video to be analyzed / the total number of video frames of the video to be analyzed;

[0057] The values of the preset video length reference value and the preset fuzzy reference value can be determined by the user based on the historical usage records of the system. The greater the user's demand for the accuracy of the score of the video to be analyzed, the smaller the values of the preset video length reference value and the preset fuzzy reference value. A value of a preset video length reference value is provided, and the preset video length reference value is 15 minutes. A value of a preset fuzzy reference value is provided. The fuzzy reference values corresponding to videos with the same video length as the video to be analyzed during historical usage are detected, and the average value of the fuzzy reference values whose score accuracy meets the user's needs is recorded as the preset fuzzy reference value.

[0058] Specifically, the association processing unit responds to the first analysis method and determines the frame extraction point setting method according to the association state of the fuzzy frame;

[0059] If the association state is the first preset association state, the association analysis unit determines that the frame extraction point setting method is to segment the video to be analyzed according to the fuzzy frame position to obtain a plurality of associated fuzzy frame segments, and perform frame extraction point compensation on each associated fuzzy frame segment;

[0060] If the association state is the second preset association state, the association analysis unit determines that the frame extraction point setting method is to perform frame extraction point compensation for the video segment between adjacent blurred frames.

[0061] The frame point compensation is to increase the number of frame points and reduce the frame point interval by increasing the number of frame points, so that the number of video frames can meet the actual test needs. The higher the user's demand for the scoring accuracy of video actions in the test, the smaller the value of the frame point interval set during the frame point compensation. The relationship between the reduction value of the frame point interval and the scoring accuracy of video actions in the test is positively correlated.

[0062] Specifically, the association processing unit performs segmentation processing on the video to be analyzed according to the fuzzy frame positions to obtain a plurality of associated fuzzy frame segments, and a time interval reference value between each fuzzy frame in any single associated fuzzy frame segment is less than a preset time interval reference value;

[0063] The segmentation processing includes: performing segmentation analysis on each blurred frame in sequence according to the time sequence of the video to be analyzed; when performing segmentation analysis on a single blurred frame, recording the blurred frame as a target blurred frame; detecting the time interval reference value between each blurred frame after the time sequence of the target blurred frame; if the time interval reference value between the blurred frame after the time sequence of the target blurred frame and the blurred frame is less than a preset time interval reference value, recording it as an associated blurred frame of the target blurred frame; if the time interval reference value between the blurred frame after the time sequence of the target blurred frame and the blurred frame is greater than or equal to the preset time interval reference value, stopping the segmentation analysis of the target blurred frame, and uniformly recording the target blurred frame and its corresponding associated blurred frames as an associated blurred frame segment.

[0064] When confirmation of a related fuzzy frame segment is completed, segmentation analysis is continued for a fuzzy frame that is closest to the related fuzzy frame segment in time sequence.

[0065] The time interval reference value is the time length between the extraction point of the blurred frame and the extraction point of the adjacent blurred frame in the time sequence;

[0066] The user can determine the value of the preset time interval reference value based on the historical usage records of the system. The greater the user's demand for the accuracy of the score of the video to be analyzed, the smaller the value of the preset time interval reference value. A value of the preset time interval reference value is provided. The preset time interval reference value is the maximum value of the time interval reference value in the associated blurred frame segments of each video in the historical usage records whose score accuracy meets the user's demand.

[0067] Specifically, the association processing unit determines the association state of the blurred frames according to the distance interval reference value between the blurred frames;

[0068] The first preset association state is that there are a number of adjacent blurred frames in the video to be analyzed whose distance interval reference value is equal to the frame sampling point interval;

[0069] The second preset association state is that the reference values of the distance intervals of the blurry frames in the video to be analyzed are all greater than the frame extraction point interval;

[0070] The frame sampling point interval is the time length between two adjacent frame sampling points in the video to be analyzed.

[0071] The reference value of the distance interval between blur frames is the interval length between the frame extraction points corresponding to two adjacent blur frames.

[0072] Specifically, the adjustment analysis unit determines the video frame category based on the limb point change degree and the limb point change quantity reference value. The video frame category includes a type of frame in which the limb point change degree is less than or equal to the preset limb point change degree and the limb point change quantity reference value is less than or equal to the preset limb point change quantity reference value, and a type of frame in which the limb point change degree is greater than the preset limb point change degree or the limb point change quantity reference value is greater than the preset limb point change quantity reference value.

[0073] It can be understood that the video frame is an image with a rectangular shape, and a rectangular coordinate system is established with the lower left corner of the rectangle as the coordinate origin, the straight line extending from the coordinate origin along the bottom edge of the rectangle to the right as the x-axis, and the straight line extending from the coordinate origin along the left side of the rectangle upward as the y-axis. The limb point position is the coordinate of the limb point in the rectangular coordinate system. For a single limb point, the shortest distance between the limb point position and the corresponding limb point position in the standard state is recorded as the change distance. The standard state is the state in which the examinee in the video to be identified maintains a military posture and stands. The limb point change degree is the sum of the change distances of each limb point. The reference value of the limb point change number is the number of limb points whose limb point positions have changed compared with the limb point positions in the standard state; a point is randomly selected on each joint of the human body, and these points are limb points, including the head, shoulders, elbows, wrists, hips, knees and ankles; the identification of limb points can be determined by machine vision and deep learning networks. This is content that is easy for technicians in this field to understand and will not be elaborated here.

[0074] The values of the preset limb point change degree and the preset limb point change number reference value can be determined by the user based on the historical usage records of the system. The greater the user's demand for the accuracy of the video score to be analyzed, the smaller the values of the preset limb point change degree and the preset limb point change number. A value of the preset limb point change degree is provided, and the preset limb point change degree is 0.5m. A value of the preset limb point change number reference value is provided, and the preset limb point change number reference value is 2.

[0075] Specifically, the adjustment processing unit responds to the second analysis method and determines an adjustment method according to the video frame aggregation area. The adjustment method includes a first adjustment method for adjusting the number of frame extraction points in the first frame aggregation area according to the similarity between the two types of frame images closest to both ends of the first frame aggregation area, and a second adjustment method for adjusting the frame extraction point interval of the video to be analyzed according to the length similarity of the two types of frame aggregation areas.

[0076] The video frame aggregation area includes a first-class frame aggregation area and a second-class frame aggregation area. The first-class frame aggregation area and the second-class frame aggregation area are confirmed as follows:

[0077] When multiple consecutive adjacent video frames are all class I frames, these video frames are recorded as target class I frames, and the class I frame aggregation area is the video area between the starting target class I frame and the ending target class I frame in the time sequence;

[0078] When multiple consecutive adjacent video frames are all Class II frames, these video frames are recorded as target Class II frames, and the Class II frame aggregation area is the video area between the starting target Class II frame and the ending target Class II frame in time sequence.

[0079] When the video frame aggregation area is a first-type frame aggregation area, the adjustment method is the first adjustment method; when the video frame aggregation area is a second-type frame aggregation area, the adjustment method is the second adjustment method.

[0080] The calculation formula of the similarity φ of the two types of frame images closest to the two ends of the clustered area of the first type of frame is: φ = α × ω1 + β × ω2, where α is the proportion of the same changed limb points, ω1 is the weight coefficient corresponding to the proportion of the same changed limb points, β is the difference in the degree of change of the limb points, and ω2 is the weight coefficient corresponding to the difference in the degree of change of the limb points;

[0081] The proportion of the same changed limb points α = the reference value of the number of the same changed limb points / the maximum value of the reference value of the number of changed limb points in the two-type frame images closest to the two ends of the clustering area of the first type frame;

[0082] The reference value of the number of identical changed limb points is the number of changed limb points with identical coordinates in the standard state in the two types of frame images closest to both ends of the first type of frame aggregation area; the reference value of the number of changed limb points is the number of limb points whose positions in the video frame are different from the corresponding limb point positions in the standard state;

[0083] The limb point change difference β is the absolute value of the difference between the limb point change of the second-class frame image closest to the left end of the first-class frame aggregation area and the limb point change of the second-class frame image closest to the right end of the first-class frame aggregation area;

[0084] It can be understood that the weight coefficients corresponding to the proportion of limb points with the same change and the weight coefficients corresponding to the difference in limb point change are set in advance by the user, and one value of ω1 and ω2 is provided, ω1=0.5, ω2=-0.5.

[0085] The length similarity of the second-class frame aggregation area = the maximum value of the length of each second-class frame aggregation area - the minimum value of the length of each second-class frame aggregation area. The length of the second-class frame aggregation area is the time length of the video area between the starting target second-class frame and the ending target second-class frame in the time sequence.

[0086] Specifically, the adjustment processing unit responds to the first adjustment mode and increases or decreases the number of frame extraction points in the first frame aggregation area according to the similarity between the two types of frame images closest to the two ends of the first frame aggregation area;

[0087] There is a negative correlation between the number of frame extraction points in the first-class frame aggregation area and the similarity between the second-class frame images at both ends of the first-class frame aggregation area.

[0088] When the similarity between the two types of frame images closest to the two ends of the one type of frame aggregation area is greater than the preset image similarity, the number of frame extraction points in the one type of frame aggregation area is reduced;

[0089] When the similarity between the two types of frame images closest to the two ends of the first type of frame aggregation area is less than the preset image similarity, the number of frame extraction points in the first type of frame aggregation area is increased;

[0090] The value of the preset image similarity can be determined by the user based on the historical usage records of the system. The greater the user's demand for the accuracy of the video score to be analyzed, the greater the value of the preset image similarity. A value of the preset image similarity is provided. The preset image similarity is the average value of the similarity of the two types of frame images closest to the two ends of the first type frame aggregation area of each video whose scoring accuracy meets the user's requirements in the historical usage records.

[0091] Specifically, the adjustment processing unit responds to the second adjustment mode, and if there are multiple second-type frame clustering regions with length similarity greater than a preset length similarity in the video to be analyzed, the frame sampling point interval of the video to be analyzed is reduced;

[0092] The relationship between the reduction value of the frame extraction point interval and the representativeness of the second type of frame image is positively correlated.

[0093] The value of the preset length similarity can be determined by the user based on the historical usage records of the system. The greater the user's demand for the accuracy of the video score to be analyzed, the larger the value of the preset length similarity is. A value of the preset length similarity is provided. The preset length similarity is the average value of the length similarity of the second type of frame clustering areas of each video whose scoring accuracy meets the user's requirements in the historical usage records.

[0094] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0095] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. An online examination invigilation system based on image recognition, characterized in that: include: An image acquisition unit, used to acquire video frame information of the video to be analyzed; an image analysis unit connected to the image acquisition unit, configured to determine a video state of the video to be analyzed based on a video duration reference value and a fuzzy reference value, and to determine an analysis method based on the video state of the video to be analyzed; an association processing unit connected to the image acquisition unit and the image analysis unit, configured to determine an association state of the blurred frames according to a reference value of a distance interval between the blurred frames, and determine a frame extraction point setting method according to the association state of the blurred frames; an adjustment and analysis unit connected to the image acquisition unit and the image analysis unit, for determining a video frame category according to a limb point change degree and a limb point change quantity reference value; an adjustment processing unit connected to the adjustment analysis unit and configured to determine, based on the video frame aggregation region, an adjustment method, which is to adjust the number of frame extraction points in the first frame aggregation region according to the similarity between the two types of frame images closest to both ends of the first frame aggregation region or to adjust the frame extraction point interval of the video to be analyzed according to the length similarity between the two types of frame aggregation regions; The image analysis unit determines an analysis method according to the video state of the video to be analyzed, wherein, in the first video state, the analysis method is a first analysis method of determining a frame extraction point setting method according to the correlation state of the blurred frames; In the second video state, the analysis method is a second analysis method that determines the adjustment method according to the first type of frame aggregation area and the second type of frame aggregation area.

2. The online examination invigilation system based on image recognition according to claim 1, characterized in that: The image analysis unit determines the video state of the video to be analyzed according to the video duration reference value and the fuzzy reference value; The video status of the video to be analyzed includes a first video status in which the video length reference value is greater than a preset video length reference value or the blur reference value is greater than a preset blur reference value, and a second video status in which the video length reference value is less than or equal to the preset video length reference value and the blur reference value is less than or equal to the preset blur reference value.

3. The online examination invigilation system based on image recognition according to claim 2, characterized in that: The association processing unit responds to the first analysis mode and determines the frame extraction point setting mode according to the association state of the fuzzy frames; If the association state is the first preset association state, the association analysis unit determines that the frame extraction point setting method is to segment the video to be analyzed according to the fuzzy frame position to obtain a plurality of associated fuzzy frame segments, and perform frame extraction point compensation on each associated fuzzy frame segment; If the association state is the second preset association state, the association analysis unit determines that the frame extraction point setting method is to perform frame extraction point compensation for the video segment between adjacent blurred frames.

4. The online examination invigilation system based on image recognition according to claim 3 is characterized in that: The association processing unit performs segmentation processing on the video to be analyzed according to the fuzzy frame positions to obtain a plurality of associated fuzzy frame segments, wherein the time interval reference value between each fuzzy frame in any single associated fuzzy frame segment is less than a preset time interval reference value; The segmentation processing includes: performing segmentation analysis on each blurred frame in sequence according to the time sequence of the video to be analyzed; when performing segmentation analysis on a single blurred frame, recording the blurred frame as a target blurred frame; detecting the time interval reference value between each blurred frame after the time sequence of the target blurred frame; if the time interval reference value between the blurred frame after the time sequence of the target blurred frame and the blurred frame is less than a preset time interval reference value, recording it as an associated blurred frame of the target blurred frame; if the time interval reference value between the blurred frame after the time sequence of the target blurred frame and the blurred frame is greater than or equal to the preset time interval reference value, stopping the segmentation analysis of the target blurred frame, and uniformly recording the target blurred frame and its corresponding associated blurred frames as an associated blurred frame segment.

5. The online examination invigilation system based on image recognition according to claim 3 is characterized in that: The association processing unit determines the association state of the blurred frames according to the distance interval reference value between the blurred frames; The first preset association state is that there are a number of adjacent blurred frames in the video to be analyzed whose distance interval reference value is equal to the frame sampling point interval; The second preset association state is that the reference values of the distance intervals of the blurry frames in the video to be analyzed are all greater than the frame extraction point interval; The frame sampling point interval is the time length between two adjacent frame sampling points in the video to be analyzed.

6. The online examination invigilation system based on image recognition according to claim 2, characterized in that: The adjustment analysis unit determines the video frame category based on the limb point change degree and the limb point change quantity reference value. The video frame category includes a type of frame in which the limb point change degree is less than or equal to the preset limb point change degree and the limb point change quantity reference value is less than or equal to the preset limb point change quantity reference value, and a type of frame in which the limb point change degree is greater than the preset limb point change degree or the limb point change quantity reference value is greater than the preset limb point change quantity reference value.

7. The online examination invigilation system based on image recognition according to claim 6, characterized in that: The adjustment processing unit responds to the second analysis method and determines an adjustment method based on the video frame aggregation area, the adjustment method including a first adjustment method for adjusting the number of frame extraction points in the first frame aggregation area based on the similarity between the two types of frame images closest to both ends of the first frame aggregation area, and a second adjustment method for adjusting the frame extraction point interval of the video to be analyzed based on the length similarity between the two types of frame aggregation areas; The video frame aggregation area includes a first-class frame aggregation area and a second-class frame aggregation area. The first-class frame aggregation area and the second-class frame aggregation area are confirmed as follows: When multiple consecutive adjacent video frames are all class I frames, these video frames are recorded as target class I frames, and the class I frame aggregation area is the video area between the starting target class I frame and the ending target class I frame in the time sequence; When multiple consecutive adjacent video frames are all Class II frames, these video frames are recorded as target Class II frames, and the Class II frame aggregation area is the video area between the starting target Class II frame and the ending target Class II frame in time sequence.

8. The online examination invigilation system based on image recognition according to claim 6, characterized in that: The adjustment processing unit responds to the first adjustment mode and increases or decreases the number of frame extraction points in the first frame aggregation area according to the similarity between the two types of frame images closest to the two ends of the first frame aggregation area; There is a negative correlation between the number of frame extraction points in the first-class frame aggregation area and the similarity between the second-class frame images at both ends of the first-class frame aggregation area.

9. The online examination invigilation system based on image recognition according to claim 6, characterized in that: The adjustment processing unit responds to the second adjustment mode, and if there are multiple second-type frame clustering regions with length similarity greater than a preset length similarity in the video to be analyzed, the frame sampling point interval of the video to be analyzed is reduced; The relationship between the reduction value of the frame extraction point interval and the representativeness of the second type of frame image is positively correlated.

Images