Examination room automatic inspection method and device, storage medium and electronic equipment

Abstract

The invention provides an examination room automatic inspection method and device, a storage medium and electronic equipment, and belongs to the field of security and protection. The electronic device obtains a historical time difference sequence of a target examination room; obtaining a time difference threshold value suitable for the target examination room according to the historical time difference sequence; obtaining current display time from the current monitoring picture of the target examination room; acquiring a time deviation between the current display time and the current reference time; and if the time deviation is greater than the time difference threshold, delay alarm information is generated. Therefore, since the historical time difference sequence reflects the deviation change trend between the monitoring picture display time and the reference time in the previous operation process of the examination room, the judgment standard is dynamically set according to the actual operation characteristics of different examination rooms, the real fault and the normal fluctuation can be effectively distinguished, the false alarm rate is reduced, and the examination room safety is improved. And the accuracy and adaptability of the inspection system are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of security and protection, in particular, to an examination room automatic inspection method and device, a storage medium and an electronic equipment. BACKGROUND

[0002] Before the formal start of an education examination, it is a crucial preparatory work to inspect the examination room environment and equipment status. Since the examination process is highly dependent on standardization and fairness, any technical or management oversight that may affect the normal progress of the examination must be excluded in advance. For example, if the time information in the monitoring video stream in the examination room is displayed abnormally, it may cause the examination staff to be unable to accurately check the time point of the event, affecting the reliability of the subsequent tracing; the lack of necessary OSD label information in the video picture will affect the identification and archiving of key attributes such as examination sites, examination rooms, and passages by the remote inspection system; and the desks placed in a non-compliant distance may cause cheating risks or violations of examination discipline. Therefore, it is necessary to comprehensively check the video stream of each examination site and examination room before the examination to ensure that all technical facilities are in compliance and available.

[0003] The current examination room inspection method mainly relies on manual participation combined with simple rule judgment to complete. Specifically, the staff initiates a video stream pull request through a network inspection platform to obtain real-time monitoring pictures of multiple examination rooms and simultaneously displays multiple video images on the client. Then, the staff checks each picture frame by frame to check whether there are any abnormal situations. Such a check includes judging whether the timestamp marked in the monitoring picture is clearly visible, the time delay is too long, the format is correct, the position is within the preset area, the content of the OSD label is complete and meets the naming specification, the video picture has quality problems such as blur, frame loss, color cast, and whether the desk arrangement meets the requirements of examination organization.

[0004] Among them, for the time delay problem in the video transmission process, a fixed threshold is usually set for alarm, such as stipulating that "the video delay should not exceed 5 seconds", and once it exceeds, an alarm is triggered. However, this method ignores the complexity of the actual network environment and cannot adapt to the natural fluctuations caused by different network bandwidths, device performance, encoding methods, and transmission paths, which is easy to misjudge the dynamic delay within the normal range as a fault. SUMMARY

[0005] In order to overcome at least one deficiency in the prior art, the present application provides an examination room automatic inspection method, device, storage medium and electronic equipment, comprising: In a first aspect, the present application provides an examination room automatic inspection method, comprising: obtaining a historical time difference sequence of a target examination room, wherein the historical time difference sequence comprises a plurality of historical time differences arranged in time sequence, each of the historical time differences representing a deviation between a historical display time displayed in a historical monitoring picture of the target examination room and a corresponding historical reference time; obtaining a time difference threshold value suitable for the target examination room according to the historical time difference sequence; obtaining a current display time from a current monitoring picture of the target examination room; obtaining a time deviation between the current display time and a current reference time; generating a delay alarm information if the time deviation is greater than the time difference threshold value.

[0006] In a second aspect, the present application provides an examination room automatic inspection device, which comprises: an information obtaining module, configured to obtain a historical time difference sequence of a target examination room, wherein the historical time difference sequence comprises a plurality of historical time differences arranged in time sequence, each of the historical time differences representing a deviation between a historical display time displayed in a historical monitoring picture of the target examination room and a corresponding historical reference time; a threshold value adjusting module, configured to obtain a time difference threshold value suitable for the target examination room according to the historical time difference sequence; an abnormality identifying module, configured to obtain a current display time from a current monitoring picture of the target examination room, obtain a time deviation between the current display time and a current reference time, and generate a delay alarm information if the time deviation is greater than the time difference threshold value.

[0007] In a third aspect, the present application provides a storage medium, which stores a computer program, wherein the computer program realizes the examination room automatic inspection method when executed by a processor.

[0008] In a fourth aspect, the present application provides an electronic device, which comprises a processor and a memory, wherein the memory stores a computer program, and the computer program realizes the examination room automatic inspection method when executed by the processor.

[0009] Compared with the prior art, the present application has the following beneficial effects: The method, device, storage medium and electronic device provided in the application, the electronic device obtains a historical time difference sequence of a target examination room; the historical time difference sequence comprises a plurality of historical time differences arranged in time sequence, and each historical time difference represents a deviation between a historical display time displayed in a historical monitoring picture of the target examination room and a corresponding historical reference time; a time difference threshold value suitable for the target examination room is obtained according to the historical time difference sequence; a current display time is obtained from a current monitoring picture of the target examination room; a time deviation between the current display time and a current reference time is obtained; and if the time deviation is greater than the time difference threshold value, a delay alarm information is generated. In this way, since the historical time difference sequence reflects the deviation trend between the display time and the reference time of the monitoring picture in the past operation process of the examination room, the judgment standard is dynamically set according to the actual operation characteristics of different examination rooms, the real fault and the normal fluctuation can be effectively distinguished, the false alarm rate is reduced, and the accuracy and adaptability of the inspection system are improved. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the application, and therefore should not be regarded as a limitation to the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0011] Figure 1 The flowchart of the examination room automatic inspection method provided by the embodiments of the application; Figure 2 The software system schematic diagram of the examination room automatic inspection method provided by the embodiments of the application; Figure 3 The time format configuration file schematic diagram provided by the embodiments of the application; Figure 4 The structure schematic diagram of the examination room automatic inspection device provided by the embodiments of the application; Figure 5 The structure schematic diagram of the electronic device provided by the embodiments of the application. DETAILED DESCRIPTION

[0012] In order to make the purpose, technical solutions and advantages of the embodiments of the application (hereinafter referred to as the present embodiments) more clear, the technical solutions of the embodiments of the application will be described clearly and completely in the following with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are some of the embodiments of the application, not all the embodiments. The components of the embodiments of the application described and shown in the drawings here can be arranged and designed in various different configurations.

[0013] The following detailed description of embodiments of the application in the depictions is not intended to limit the scope of the application, but rather the application is given only by the claims. Based upon the embodiments of the application described herein, all other embodiments falling within the scope of the application will be obvious to those skilled in the art to which the application pertains. Such

[0014] It should be noted that like reference numerals and letters refer to like items in the several views of the drawings, and as such, definitions and explanations of such items need not be repeated in the several views of the drawings.

[0015] In the description of the application, it should be noted that the terms "first", "second", "third", etc. are used merely to distinguish descriptions and can not be understood as indicating or implying relative importance. In addition, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0016] Based on the above statement, as introduced in the background art, for the delay problem in the video transmission process, a fixed threshold is usually set for alarm, such as stipulating that "the video delay should not exceed 5 seconds", and once it exceeds, the alarm is triggered. However, this way is easy to misjudge the dynamic delay in the normal range as a fault.

[0017] To this end, it is found in the research process that the video transmission delay itself is affected by various objective factors, including network bandwidth fluctuations, router hop count, front-end device coding efficiency, central server load status, etc. These factors together determine the end-to-end delay experienced by each video from collection to platform display. In peak periods or remote areas, even if the device is working normally, a reasonable delay of 10 seconds or even longer may occur. If a static fixed delay threshold is used as the basis for alarm, it will inevitably cause false alarms in a large number of normal scenarios, causing the operation and maintenance personnel to be overwhelmed by false alarms, reducing the credibility and usability of the overall system. More importantly, different types of exams and different network architectures have different tolerance requirements for delay. For example, national exams may require higher synchronization accuracy, while regional simulation exams can be relaxed. In addition, the running environment of the same exam room changes at different times, especially in the night. Due to poor lighting conditions, camera automatic switching to low-illumination mode, coding parameter adjustment, and other factors, image quality may fluctuate, affecting the stability and accuracy of the clock digits recognized by OCR. At this time, if the same error tolerance standard as during the day is still used, it may cause frequent false alarms due to slight jitter in the recognition results.

[0018] It should be noted that the defects in the above prior art solutions are the result of careful research and practice, so the discovery process of the above problems and the solutions proposed by the embodiments of the present application to solve the above problems should be considered as contributions to the present application in the process of invention, and should not be understood as technical content known to those skilled in the art.

[0019] Based on the discovery of the above technical problems, the present embodiment provides an automatic examination room inspection method. As shown in the figure, the method comprises: Figure 1 S1, obtaining a historical time difference sequence of the target examination room.

[0020] The historical time difference sequence comprises a plurality of historical time differences arranged in chronological order, and each historical time difference represents the deviation between the historical display time displayed in the historical monitoring picture of the target examination room and the corresponding historical reference time.

[0021] S2, obtaining a time difference threshold suitable for the target examination room according to the historical time difference sequence.

[0022] S3, obtaining the current display time from the current monitoring picture of the target examination room.

[0023] S4, obtaining the time deviation between the current display time and the current reference time.

[0024] S5, if the time deviation is greater than the time difference threshold, generating a delay alarm information. ​

[0025] Thus, since the historical time difference sequence reflects the change trend of the deviation between the display time of the monitoring picture and the reference time in the past operation of the examination room, the judgment standard is dynamically set according to the actual operation characteristics of different examination rooms, which can effectively distinguish between real faults and normal fluctuations, thereby reducing the false alarm rate and improving the accuracy and adaptability of the inspection system.

[0026] It should be noted that the electronic device implementing the examination room automatic inspection method can be any device with data processing, video analysis and network communication capabilities, specifically, but not limited to, a general server, an edge computing device or a special industrial computer. Taking a server as an example, it is equipped with a central processing unit (CPU) and a graphics processing unit (GPU) to support concurrent processing of multiple examination room monitoring video streams and AI model inference tasks.

[0027] As shown in Figure 2 The system for implementing the examination room automatic inspection method in the server includes a data acquisition layer, a preprocessing layer, an AI processing layer, a rule engine layer and a business presentation layer. The data is transmitted and the functions are called between the layers through a standardized interface to form a complete intelligent inspection processing link. Among them, the data acquisition layer in the architecture is located at the bottom, and is connected with the national education department online inspection platform at the top and the online inspection platform of each region at the middle layer, and covers multiple examination room monitoring channels under each examination point, realizing the unified access to the examination room video resources in the whole network.

[0028] Based on the above system architecture, the video streaming function is integrated on the data acquisition layer, which is responsible for initiating and managing all video stream acquisition requests. The AI processing layer receives the processed video frames output from the data acquisition layer and performs multiple artificial intelligence analysis tasks including clock detection, OSD label recognition, desk positioning, etc. The rule engine layer checks and determines the results output by the AI processing layer according to the preset compliance judgment logic. Finally, the business presentation layer receives the inspection results after the rule judgment, and realizes the creation, scheduling and state tracking of the inspection task through the task management module, and displays the results in a visual way or generates a report output.

[0029] The data acquisition layer obtains real-time video streams of all the test rooms in the network through RTSP and HTTP protocols, supports H.264 and H.265 encoding formats, and ensures compatibility with mainstream cameras and educational examination monitoring systems. To adapt to diversified access scenarios, the data acquisition layer also supports stream media protocols such as RTMP and HLS, and can access non-standard video sources such as mobile terminals and wireless devices. The data acquisition layer uses multi-threading and asynchronous IO technology to realize concurrent pulling of no less than 100 video streams, and dynamically allocates resources through an optimized thread scheduling strategy to ensure stable operation under high concurrency. The server performs Gaussian filtering, median filtering, and bilateral filtering on the obtained raw video frames to eliminate various image noises, thereby improving the picture quality; then performs histogram equalization processing to enhance the image contrast, further improves the visual effect under low illumination or backlight conditions, and provides clear and stable input data for the subsequent AI processing layer.

[0030] Based on the above introduction of the system architecture, to make the scheme provided by the embodiment more clear, the following takes the server as an electronic device for implementing the test room automatic inspection method, and combines Figure 1 each step of the method is described in detail. It should be understood that the operations of the flowchart can not be implemented in sequence, and the steps without logical context relationship can be reversed in order or implemented simultaneously. In addition, under the guidance of the content of the present application, one or more other operations can be added to the flowchart, or one or more operations can be removed from the flowchart. Continue to refer to Figure 1 The method comprises: S1, obtaining a historical time difference sequence of the target test room.

[0031] The historical time difference sequence comprises a plurality of historical time differences arranged in chronological order, and each historical time difference represents the deviation between the historical display time displayed in the historical monitoring picture of the target test room and the corresponding historical reference time.

[0032] It can be understood that, by collecting the deviation data between the clock display of a specific test room at different time points and the standard time, a historical record reflecting the time synchronization characteristics of the test room is constructed, and corresponding analysis data is provided for dynamically setting a reasonable compliance judgment threshold in the subsequent process.

[0033] In actual application, the server extracts a plurality of time point monitoring picture frames from the stored historical video records, and the monitoring pictures contain time information automatically generated by the camera or video coding device and superimposed on the picture, i.e. OSD (On-Screen Display) time label. The OSD time label serves as the historical display time, indicating the time stamp recorded at the recording moment of the historical monitoring picture. At the same time, the server also obtains the standard reference time corresponding to the picture frame acquisition moment, which is usually derived from the precise time provided by the network time protocol server. For each picture frame, the time difference between the historical display time and the corresponding historical reference time is calculated to form a historical time difference. All the historical time differences arranged in time sequence form a historical time difference sequence. Therefore, the historical time difference sequence here refers to the collection of a plurality of historical time differences arranged in time sequence.

[0034] Based on the above description of the historical time difference sequence in the embodiment, the step S2 in Figure 1 will be explained as follows: Figure 1 S2, obtaining the time difference threshold applicable to the target examination room according to the historical time difference sequence.

[0035] As an optional implementation, the server can input the historical time difference sequence into a threshold prediction model for processing to obtain the time difference threshold applicable to the target examination room, wherein the threshold prediction model is trained based on a recurrent neural network architecture.

[0036] In summary, the embodiment introduces a recurrent neural network architecture with time series modeling capability, automatically learns the normal fluctuation range of the target examination room by using its historical time deviation data, dynamically generates personalized judgment criteria, and avoids false positives caused by relying on fixed thresholds.

[0037] ​In actual application, the server takes the historical time difference sequence of the target examination room as input data, which contains multiple historical time differences arranged in time sequence and reflects the deviation trend between the time displayed in the monitoring picture and the reference time in different time periods; and inputs the historical time difference sequence as a whole into the threshold prediction model trained in advance for processing. The threshold prediction model refers to a machine learning model (for example, an LSTM model) based on a recurrent neural network architecture, which can capture long-term dependencies and periodic patterns in time series data. In the model inference process, the server analyzes and induces the time sequence features in the historical time difference sequence through the threshold prediction model, and outputs a time difference threshold suitable for the current running state of the target examination room. The time difference threshold can reflect the reasonable delay tolerance range of the examination room under normal working conditions, for example, a slightly larger deviation can be allowed at night or during a period of low network load, thereby improving the accuracy and adaptability of alarm judgment.

[0038] Based on the above description of the time difference threshold of the target examination room in the embodiments, continue to refer to Figure 1 , and the following will continue to explain step S3 in Figure 1 : S3, obtain the current display time from the current monitoring picture of the target examination room.

[0039] As an optional implementation, this embodiment automatically extracts the time information contained in the monitoring picture by using advanced optical character recognition technology, converts the clock numbers or screen superimposed time labels in the image into comparable structured time data.

[0040] In actual application, the server first obtains the current monitoring picture of the target examination room as input data, which is a video frame collected by the examination room camera device in real time and transmitted to the patrol platform; then, performs optical character recognition (Optical Character Recognition, OCR) processing on the current monitoring picture to locate and identify the time content displayed in the picture. The current display time refers to the time value presented by the on-screen display (On-Screen Display, OSD) time label automatically generated by the camera.

[0041] In the OCR processing stage, the server can employ an OCR model based on a Convolutional Recurrent Neural Network (CRNN) architecture. This model can extract image features through convolutional layers, capture the context relationship of character sequences using recurrent layers, and implement end-to-end text recognition combined with a Connectionist Temporal Classification (CTC) loss function. It can effectively recognize continuous clock digits. To further improve recognition accuracy and robustness, the server can also employ an OCR model based on a Transformer architecture. This model enhances the modeling ability for long sequences and complex background text through a self-attention mechanism, making it suitable for scenarios with uneven lighting, diverse fonts, or slight blurring. Additionally, during the training process of the OCR model, an adversarial training strategy is introduced to simulate noise, occlusion, and low resolution, improving the model's adaptability in real complex environments.

[0042] Optionally, the server can also use a Long Short-Term Memory (LSTM) network to verify the coherence of the identified time sequence, ensuring that the output current display time is logically reasonable, stable, and reliable.

[0043] Based on the above description of the current display time recognition method in the embodiments, continue to refer to Figure 1 , and next continue to describe step S4 in Figure 1 : S4, obtain the time deviation between the current display time and the current reference time.

[0044] In practice, it is found that when judging time compliance, the recognized time in the monitoring screen is usually directly compared with the standard time without verifying whether the time information appears in the specified position or conforms to the preset format. This method is prone to misidentifying numbers in non-clock regions as display times or causing invalid time to participate in comparison due to label format errors, thereby causing false alarms. In view of this, as an optional implementation, before step S4, the server obtains the display position of the current display time; and the display position is consistent with the reference display position recorded in the rule file, and the current display time is consistent with the time format recorded in the rule file, as the condition for executing step S4.

[0045] It can be understood that this embodiment introduces a configurable rule file to double-constrain the presentation of time information, ensuring that only when the time content appears in the specified area and conforms to the preset format does the deviation calculation start.

[0046] In actual application, the server obtains the current display time from the current monitoring picture of the target examination room, and synchronously extracts the coordinate position of the current display time in the picture, i.e., the display position of the current display time. The display position here refers to the picture rectangular region where the current display time is located, which is usually represented by the coordinates of the upper left corner and the lower right corner. Then, the server matches and checks the display position with the reference display position recorded in the rule file, and the reference display position is the region where the legal time label is located, which is defined in advance through the rule file. As shown in Figure 3 , multiple reference display positions can coexist in the same rule configuration.

[0047] The server further checks whether the current display time is consistent with the time format recorded in the rule file, where the time format refers to the expression of the time string, such as "HH:MM:SS" or "hh:mm:ss AM / PM". The rule file can be written in a structured format such as JSON or YAML, which supports flexible definition according to the manufacturer and type of the connected camera, multiple combination rules, and parameters such as position, format, and error threshold. For example, the rule file can record the corresponding matching rules in the following manner: { "clock_rules": [ {"position": [0, 0, 100, 100], "format": "HH:MM:SS", "error_threshold": 5}, {"position": [10, 10, 110, 110], "format": "hh:mm:ss AM / PM", "error_threshold": 3} ] } It should be noted that "error_threshold" in the above rule file represents the time difference threshold, which is obtained based on the historical time difference sequence.

[0048] Based on the above rule file, the server can realize dynamic loading and matching judgment of rules through the Drools rule engine. As long as the display position of the current display time is consistent with the reference display position recorded in any rule, and the current display time is consistent with the time format recorded in the rule, it can meet the condition for executing subsequent time deviation calculation. This design avoids the problem of misidentification caused by differences in camera installation or changes in picture layout.

[0049] Therefore, when the server detects that the format of the current display time and the display position are consistent with the matching rule recorded in the rule file, the time deviation between the current display time and the current reference time is calculated.

[0050] Based on the time deviation obtained in the above embodiment, referring to FIG. 5, the step S5 is described as follows: Figure 1 S5, if the time deviation is greater than the time difference threshold, a delay alarm information is generated.

[0051] In actual application, the server first obtains the time deviation between the current display time and the current reference time, and compares it with the time difference threshold suitable for the target examination room obtained according to the historical time difference sequence. The time deviation here refers to the difference between the time displayed in the current monitoring picture and the standard time, and the time difference threshold is a tolerance range dynamically generated based on the historical time synchronization characteristics of the examination room. If the time deviation is greater than the time difference threshold, the server determines that there is an abnormal delay in the examination room, and generates a delay alarm information. The delay alarm information can be pushed to the invigilation terminal or the management platform, reminding the operation and maintenance personnel to handle it in time, so as to ensure the accuracy and reliability of time synchronization during the examination.

[0052] In addition to identifying and processing the time in the form of OSB in the current monitoring picture, the server also uses an improved clock detection model to accurately position and identify the clock area in the current monitoring picture. Specifically, the server inputs the current monitoring picture into a clock detection model based on the improved YOLOv5 architecture to determine the position of the clock area. The model uses CSPDarknet53 as the backbone network and integrates attention mechanism, which can effectively focus on the key feature areas such as clock edge and dial, thereby improving the detection accuracy in complex background. At the same time, the model is equipped with multi-scale detection head, including the design of 3x3 and 5x5 convolution kernel size, which supports adaptive identification of different sizes and types of clocks (such as wall clock, electronic screen time), and enhances the compatibility of multiple scenes. In addition, in order to improve the generalization performance of the model in the actual deployment environment, image rotation, affine transformation and other data enhancement techniques are introduced in the training stage, which significantly improves the stability of the model under non-ideal conditions such as camera angle deviation and local occlusion.

[0053] ​After completing the clock area positioning, the server further extracts the on-site time of the target examination room through optical character recognition technology, and detects the periodic signal (for example, the ticking sound of a mechanical clock) in the environmental sound, which is used to assist in verifying whether the clock is running normally. When the image recognition result is abnormal but the audio signal is stable, or the image shows that the time is stagnant while the audio beat disappears, the server can comprehensively judge whether there is a real failure. For example, if the server confirms that the clock has stopped running or the countdown has appeared abnormally, it will immediately trigger a real-time alarm function and generate corresponding state abnormal alarm information.

[0054] In this way, the false alarm risk caused by single visual recognition failure is effectively reduced, and the intelligent level and fault tolerance of the overall inspection system are enhanced.

[0055] The research also found that existing examination room inspection mainly relies on manual inspection of whether the desks are placed in accordance with examination specifications, and it is difficult to achieve real-time monitoring on a large scale and with high efficiency. In view of this, the examination room automatic inspection method provided in the embodiment further includes: S6, obtaining the coordinates of the multiple desks in the current monitoring picture.

[0056] As an optional implementation manner, the server can input the current monitoring picture into the backbone network of the desk detection model to obtain the first image feature of the current monitoring picture, wherein the backbone network is constructed based on the CSPDarknet53 structure; input the first image feature into the neck network of the desk detection model to obtain the second image feature of the current monitoring picture, wherein the neck network is constructed based on the PANet structure and the E-ELAN structure; input the second image feature into the self-attention network of the desk detection model to obtain the third image feature of the current monitoring picture, wherein the self-attention network is constructed based on the CBAM structure; and input the third image feature into the detection head of the desk detection model to obtain the coordinates of the multiple desks.

[0057] It can be understood that the embodiment realizes high-precision and robust positioning of the coordinates of the multiple desks in the current monitoring picture by constructing a desk detection model that integrates advanced network structures and attention mechanisms.

[0058] In actual application, the server first inputs the current monitoring picture into the backbone network of the desk detection model for preliminary feature extraction. The backbone network here is a deep convolutional neural network module constructed based on the CSPDarknet53 structure, which is based on the principle of introducing a cross-stage partial connection structure to divide the input feature map into two branches in the channel dimension. One branch is subjected to deep feature transformation through a plurality of convolution blocks, and the other branch directly transmits the original features; then, the outputs of the two branches are fused in the subsequent stage, and finally the first image feature of the current monitoring picture is output.

[0059] Based on the first image feature, the server inputs the first image feature into a neck network of the desk detection model, the neck network being jointly constructed based on a Path Aggregation Network (PANet) structure and an Expanded Efficient Layer Aggregation Network (E-ELAN) structure, and being capable of enhancing the information flow and fusion capability between different scale features, thereby generating a second image feature with higher discriminability.

[0060] On this basis, the server further inputs the second image feature into a self-attention network of the desk detection model, the self-attention network being an attention mechanism module constructed based on a Convolutional Block Attention Module (CBAM) structure, and being capable of focusing on key regions such as desk edges through a channel attention and a spatial attention, thereby significantly improving the recognition capability in a complex background or a partial occlusion condition, and outputting a third image feature.

[0061] Finally, the server inputs the third image feature into a detection head of the desk detection model, completes a target positioning task, and obtains the coordinates of the multiple desks in the current monitoring picture.

[0062] Therefore, in the process of recognizing the desk coordinates, the overall architecture design of the improved YOLOv5 is combined, the CBAM attention mechanism is integrated in the Neck layer to strengthen the key region response, and the SPPF (Spatial Pyramid Pooling Fast) structure in YOLOv8 is introduced to enhance the extraction capability of small target edge features by using pooling operations of multiple sizes such as 5×5, 9×9, and 13×13, thereby improving the detection precision of small target desks with a size less than 100 pixels. In addition, data enhancement techniques such as image rotation and affine transformation are used in the training process, thereby improving the adaptability of the model to the differences in camera installation angles and the occlusion conditions.

[0063] Based on the coordinates of the multiple desks obtained according to the above embodiment, the automatic examination room inspection method further includes: S7. According to the coordinates of the multiple desks, multiple adjacent distances between the multiple desks are calculated.

[0064] In actual application, the server can determine the two-dimensional position of each desk in the current monitoring picture based on the coordinates of the plurality of desks obtained in the preceding step. The coordinates of the plurality of desks refer to the data points output by the desk detection model and representing the position of each desk in the image plane, which can be in the form of pixel coordinates of the center point or corner point of the bounding box. The server calculates the straight-line distance between any two adjacent desks according to the coordinate data using a geometric distance algorithm to form a plurality of adjacent distances. The plurality of adjacent distances refer to the set of distance values calculated between pairs of desks that are adjacent in spatial distribution.

[0065] S8, if at least one of the plurality of adjacent distances is less than the distance threshold, generating distance alarm information.

[0066] In actual application, the server compares the plurality of adjacent distances calculated in the preceding step with the preset distance threshold one by one. The plurality of adjacent distances refer to the set of spatial distances between pairs of adjacent desks determined based on the coordinates of the plurality of desks, while the distance threshold is a preset minimum allowed distance used to measure whether the placement of desks conforms to the examination specification requirements. The server determines whether there is at least one distance in the plurality of adjacent distances that is less than the distance threshold. If there is, it is determined that there is a situation where desks are too close to each other in the current examination room, which may affect the examination order or increase the risk of cheating.

[0067] Under this condition, the server generates distance alarm information and pushes the distance alarm information to the invigilation management platform or the operation and maintenance terminal to remind relevant personnel to check and handle in a timely manner.

[0068] It is also found in the research process that existing examination room monitoring systems usually use fixed and unchanged alarm thresholds to judge various quality indicators when performing video quality evaluation, and cannot dynamically adjust the discrimination standard according to the actual type of monitoring fault. It should be noted that the static threshold mechanism is prone to false positives or false negatives in specific fault scenarios, such as triggering unnecessary alarms in acceptable abnormal situations such as changes in illumination or slight occlusion, or failing to identify real quality problems in complex interference.

[0069] In view of this, the examination room automatic inspection method provided by the embodiment further comprises: S9, determining the current monitoring fault according to the current monitoring picture.

[0070] As an optional implementation manner, the server can input the current monitoring picture into a pre-trained fault recognition model to obtain the current monitoring fault.

[0071] S10, obtaining a plurality of quality evaluation indicators of the current monitoring picture.

[0072] Among them, the plurality of quality evaluation indicators are respectively configured with alarm thresholds.

[0073] S11, determining a target quality evaluation index from the plurality of quality evaluation indexes according to the monitored fault.

[0074] The target quality evaluation index represents a quality evaluation index affected by the monitored fault. S12, adjusting an alarm threshold of the target quality evaluation index to a new alarm threshold according to the monitored fault.

[0075] S13, generating a picture quality anomaly alarm if the new alarm threshold indicates that the target quality evaluation index is abnormal.

[0076] It can be understood that the embodiment realizes adaptive detection of video quality anomaly by dynamically adjusting the alarm threshold of the related quality index according to the specific fault type existing in the current monitoring picture, and improves the accuracy and scene adaptability of the alarm judgment.

[0077] In actual application, the server first determines the current monitoring fault according to the current monitoring picture. The current monitoring fault here refers to the specific image anomaly type identified by analyzing the current monitoring picture, for example, blur, color cast, brightness anomaly, stripe interference, etc. The identification process can be realized based on a pre-trained fault identification model. The model combines the imaging characteristic model learned from historical video samples and the problem pattern library, and combines the convolutional neural network (CNN) to perform deep analysis on the picture features, so as to accurately determine the main fault type existing at present.

[0078] After determining the current monitoring fault, the server obtains a plurality of quality evaluation indexes of the current monitoring picture. These quality evaluation indexes include peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), video multimethod assessment fusion (VMAF), and frame rate, definition, and color cast degree parameters. Each quality evaluation index is configured with an initial alarm threshold for judging whether it is abnormal under normal conditions. For example, if PSNR < 25 dB for 5 consecutive frames, it is marked as "blur".

[0079] Based on multiple quality evaluation indicators of the current monitoring picture, the server screens out a target quality evaluation indicator most significantly affected by the current monitoring fault from the multiple quality evaluation indicators. For example, when a "blurring" type fault is detected, the target quality evaluation indicator can be determined as PSNR or definition score; if color distortion is detected, the target quality evaluation indicator is color cast indicator. The server adaptively adjusts the alarm threshold of the target quality evaluation indicator according to the current monitoring fault, and generates a new alarm threshold. The adjustment process considers scene complexity factors such as light change intensity or motion blur degree, and dynamically optimizes parameters by modeling using historical data. For another example, when short-term blurring caused by slight motion but the overall content is still recognizable, the server can moderately increase the alarm threshold of PSNR to reduce the false alarm probability.

[0080] Finally, the server determines whether the new alarm threshold indicates that the target quality evaluation indicator is abnormal, and if the triggering condition is met, generates a picture quality abnormality alarm and pushes it to the invigilation terminal. In this way, the accuracy and environmental adaptability of the inspection system are improved.

[0081] Based on the same inventive concept as the examination room automatic inspection method provided in the embodiment, the embodiment also provides an examination room automatic inspection device. The device includes at least one software function module stored in the memory in the form of software or solidified in the electronic device. The processor in the electronic device is used to execute the executable modules stored in the memory. For example, the device includes software function modules and computer programs. Please refer to Figure 4 From the functional point of view, the device can include: The information acquisition module 11 is configured to acquire a historical time difference sequence of the target examination room, wherein the historical time difference sequence includes a plurality of historical time differences arranged in time sequence, and each historical time difference represents a deviation between a historical display time displayed in a historical monitoring picture of the target examination room and a corresponding historical reference time. The threshold adjustment module 12 is configured to obtain a time difference threshold suitable for the target examination room according to the historical time difference sequence. The abnormality identification module 13 is configured to acquire a current display time from a current monitoring picture of the target examination room; acquire a time deviation between the current display time and a current reference time; and generate a delay alarm information if the time deviation is greater than the time difference threshold.

[0082] In the embodiment, the information acquisition module 11 is configured to implement step S1 in Figure 1 The threshold adjustment module 12 is configured to implement step S2 in Figure 1 The abnormality identification module 13 is configured to implement steps S3, S4 and S5 in Figure 1 Therefore, the detailed description of each module can be found in the specific embodiments of the corresponding steps.

[0083] Since the same inventive concept as the automatic examination room inspection method, the device can also implement other steps or sub-steps of the method through the above modules.

[0084] Optionally, the threshold adjustment module 12 is further specifically used for: inputting the historical time difference sequence into a threshold prediction model for processing to obtain a time difference threshold suitable for the target examination room, wherein the threshold prediction model is obtained by training based on a recurrent neural network architecture.

[0085] Optionally, after obtaining the current display time from the current monitoring picture of the target examination room, the information acquisition module 11 is further used for: obtaining the display position of the current display time; wherein the condition for performing the step of obtaining the time deviation between the current display time and the current reference time includes that the display position is consistent with the reference display position recorded in the rule file, and the current display time is consistent with the time format recorded in the rule file.

[0086] Optionally, the information acquisition module 11 is further used for: obtaining the coordinates of the multiple desks in the current monitoring picture; calculating multiple adjacent distances between the multiple desks according to the coordinates of the multiple desks; The abnormality identification module 13 is further used for generating distance alarm information if at least one of the multiple adjacent distances is less than the distance threshold.

[0087] Optionally, the information acquisition module 11 is further specifically used for: inputting the current monitoring picture into a backbone network of a target detection model to obtain first image features of the current monitoring picture, wherein the backbone network is constructed based on a CSPDarknet53 structure; inputting the first image features into a neck network of the target detection model to obtain second image features of the current monitoring picture, wherein the neck network is constructed based on a PANet structure and an E-ELAN structure; inputting the second image features into a self-attention network of the target detection model to obtain third image features of the current monitoring picture, wherein the self-attention network is constructed based on a CBAM structure; inputting the third image features into a detection head of the target detection model to obtain the coordinates of the multiple desks.

[0088] Optionally, the threshold adjustment module 12 is further used for: determining a current monitoring fault according to the current monitoring picture; obtaining multiple quality evaluation indexes of the current monitoring picture, wherein the multiple quality evaluation indexes are respectively configured with alarm thresholds; According to the monitored fault, a target quality evaluation index is determined from a plurality of quality evaluation indexes, wherein the target quality evaluation index represents a quality evaluation index affected by the monitored fault; According to the monitored fault, an alarm threshold of the target quality evaluation index is adjusted to a new alarm threshold; The abnormality identification module 13 is further configured to generate a picture quality abnormality alarm if the new alarm threshold indicates that the target quality evaluation index is abnormal.

[0089] Optionally, the threshold adjustment module 12 is further configured to: input the current monitoring picture into the pre-trained fault identification model to obtain the current monitored fault.

[0090] In addition, each functional module in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0091] It should also be understood that the above embodiments, if realized in the form of software functional modules and sold or used as independent products, can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the parts that make contributions to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application.

[0092] Therefore, the present embodiment also provides a storage medium, which is a computer readable storage medium. The storage medium stores a computer program, and the computer program is executed by a processor to implement the examination room automatic inspection method provided by the present embodiment. The storage medium can be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0093] Please refer to Figure 5 The present embodiment also provides an electronic device for implementing the examination room automatic inspection method. The electronic device can include a processor 22 and a memory 21. The memory 21 stores a computer program, and the processor reads and executes the computer program corresponding to the above embodiments in the memory 21 to implement the examination room automatic inspection method provided by the present embodiment.

[0094] Please continue to refer to Figure 5The electronic device further includes a communication unit 23. The memory 21, the processor 22, and the communication unit 23 are electrically connected to each other directly or indirectly through a system bus 24 to enable data transmission or interaction.

[0095] The memory 21 can be an information recording device based on any electronic, magnetic, optical, or other physical principle for recording execution instructions, data, and the like. In some embodiments, the memory 21 can be, but is not limited to, a volatile memory, a non-volatile memory, a storage drive, and the like.

[0096] In some embodiments, the volatile memory can be a random access memory (RAM); in some embodiments, the non-volatile memory can be a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electric erasable programmable read-only memory (EEPROM), a flash memory, and the like; in some embodiments, the storage drive can be a disk drive, a solid state drive, any type of storage disk (such as an optical disk, a DVD, and the like), or a similar storage medium, or a combination thereof, and the like.

[0097] The communication unit 23 is configured to transmit and receive data via a network. In some embodiments, the network can include a wired network, a wireless network, a fiber optic network, a telecommunications network, an intranet, the Internet, a Local Area Network (LAN), a Wide Area Network (WAN), a Wireless Local Area Network (WLAN), a Metropolitan Area Network (MAN), a Wide Area Network (WAN), a Public Switched Telephone Network (PSTN), a Bluetooth network, a ZigBee network, or a Near Field Communication (NFC) network, etc., or any combination thereof. In some embodiments, the network can include one or more network access points. For example, the network can include wired or wireless network access points, such as base stations and / or network switching nodes, through which one or more components of the service request processing system can connect to the network to exchange data and / or information.

[0098] The processor 22 can be an integrated circuit chip that has the ability to process signals and can include one or more processing cores (e.g., a single-core processor or a multi-core processor). By way of example only, the processor can include a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), an Application Specific Instruction-set Processor (ASIP), a Graphics Processing Unit (GPU), a Physics Processing Unit (PPU), a Digital Signal Processor (DSP), a Field Programmable Gate Array (FPGA), a Programmable Logic Device (PLD), a controller, a microcontroller unit, a Reduced Instruction Set Computing (RISC), or a microprocessor, etc., or any combination thereof.

[0099] It can be understood that, Figure 5The structure shown is for illustrative purposes only. Electronic devices may also have more advanced features. Figure 5 Showing more or fewer components, or having with Figure 5 The different configurations shown. Figure 5 The components shown can be implemented using hardware, software, or a combination thereof.

[0100] It should be understood that the apparatus and methods disclosed in the above embodiments can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0101] The above descriptions are merely various embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An automatic examination room inspection method, characterized in that, The method includes: Obtain the historical time difference sequence of the target examination room, wherein the historical time difference sequence includes multiple historical time differences arranged in chronological order, and each historical time difference represents the deviation between the historical display time shown in the historical monitoring screen of the target examination room and the corresponding historical reference time; Based on the historical time difference sequence, a time difference threshold applicable to the target examination room is obtained; Obtain the current display time from the current monitoring screen of the target examination room; Obtain the time deviation between the current displayed time and the current reference time; If the time deviation is greater than the time difference threshold, a delay alarm message is generated.

2. The automatic examination room inspection method according to claim 1, characterized in that, Based on the historical time difference sequence, a time difference threshold applicable to the target examination room is obtained, including: The historical time difference sequence is input into the threshold prediction model for processing to obtain the time difference threshold applicable to the target examination room, wherein the threshold prediction model is trained based on a recurrent neural network architecture.

3. The automatic examination room inspection method according to claim 1, characterized in that, After obtaining the current display time from the current monitoring screen of the target examination room, the method further includes: Obtain the display position of the current display time; The conditions for obtaining the time deviation between the current display time and the current reference time include that the display position is consistent with the reference display position recorded in the rule file, and that the current display time is consistent with the time format recorded in the rule file.

4. The automatic examination room inspection method according to any one of claims 1-3, characterized in that, The method further includes: Obtain the coordinates of multiple desks in the current monitoring screen; Calculate multiple adjacent spacings between the multiple desks based on their coordinates; If at least one of the plurality of adjacent spacings is less than the spacing threshold, a spacing alarm message is generated.

5. The automatic examination room inspection method according to claim 4, characterized in that, Obtain the coordinates of multiple desks in the current monitoring screen, including: The current monitoring screen is input into the backbone network of the target detection model to obtain the first image features of the current monitoring screen, wherein the backbone network is constructed based on the CSPDarknet53 structure; The first image features are input into the neck network of the target detection model to obtain the second image features of the current monitoring screen, wherein the neck network is constructed based on the PANet structure and the E-ELAN structure; The second image feature is input into the self-attention network of the target detection model to obtain the third image feature of the current monitoring screen, wherein the self-attention network is constructed based on the CBAM structure; The third image feature is input into the detection head of the target detection model to obtain the coordinates of the multiple desks.

6. The automatic examination room inspection method according to claim 1, characterized in that, The method further includes: Based on the current monitoring screen, determine the current monitoring fault; The system acquires multiple quality assessment indicators of the current monitoring screen, wherein each of the multiple quality assessment indicators is configured with an alarm threshold. Based on the monitoring failure, a target quality assessment indicator is determined from the plurality of quality assessment indicators, wherein the target quality assessment indicator represents the quality assessment indicator affected by the monitoring failure. Based on the monitoring fault, the alarm threshold of the target quality assessment indicator will be adjusted to a new alarm threshold. If the new alarm threshold indicates that the target quality assessment index is abnormal, an image quality abnormality alarm will be generated.

7. The automatic examination room inspection method according to claim 6, characterized in that, Based on the current monitoring screen, the current monitoring fault is determined, including: The current monitoring screen is input into a pre-trained fault identification model to obtain the current monitoring fault.

8. An automatic examination room inspection device, characterized in that, The device includes: The information acquisition module is used to acquire the historical time difference sequence of the target examination room, wherein the historical time difference sequence includes multiple historical time differences arranged in chronological order, and each historical time difference represents the deviation between the historical display time shown in the historical monitoring screen of the target examination room and the corresponding historical reference time; The threshold adjustment module is used to obtain a time difference threshold suitable for the target examination room based on the historical time difference sequence. An anomaly detection module is used to obtain the current display time from the current monitoring screen of the target examination room; obtain the time deviation between the current display time and the current reference time; and generate a delay alarm message if the time deviation is greater than the time difference threshold.

9. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the automatic examination room inspection method according to any one of claims 1-7.

10. An electronic device, characterized in that, The electronic device includes a processor and a memory, the memory storing a computer program, which, when executed by the processor, implements the automatic examination room inspection method according to any one of claims 1-7.

Images