Personnel path abnormity early warning system based on laboratory personnel identification and tracking

Abstract

The invention discloses a personnel path abnormity early warning system based on laboratory personnel identification and tracking. The personnel path abnormity early warning system comprises an entrance image acquisition module, an in-laboratory image acquisition module, an intelligent label acquisition module, a data processing module and an information sending module. The entrance image acquisition module is used for acquiring real-time face images of personnel entering a laboratory at a laboratory entrance; the image acquisition module in the laboratory is used for acquiring real-time images of personnel entering the laboratory, namely real-time images of the personnel entering the laboratory; the intelligent label acquisition module is used for acquiring real-time position information through RFID (Radio Frequency Identification Device) equipment worn by a person entering the laboratory; the data processing module is used for processing the real-time face image to obtain personnel verification warning information; and the real-time image of the entering person is processed, and the abnormal approaching warning information of the person is obtained. According to the invention, better personnel path abnormity early warning can be carried out.

Description

Technical Field

[0001] The present invention relates to the field of early warning systems, and in particular to an abnormal personnel path early warning system based on laboratory personnel identification and tracking. Background Art

[0002] With the continuous expansion of laboratory scale and the increase of experimental complexity, laboratory safety management faces many challenges. Traditional management methods mainly rely on manual inspections and simple monitoring equipment, which makes it difficult to monitor personnel behavior and equipment operation status in real time and comprehensively, and is prone to management loopholes and safety hazards. In recent years, with the rapid development of technologies such as the Internet of Things, computer vision, and artificial intelligence, laboratory safety management has ushered in new opportunities.

[0003] There are many potential risks in the laboratory environment, such as leakage of hazardous chemicals, improper equipment operation, and personnel entering dangerous areas by mistake. These risks not only threaten the safety of experimental personnel, but may also cause damage to experimental equipment and loss of experimental data, and even cause serious safety accidents. Therefore, it is particularly important to establish an intelligent safety management system that can monitor personnel behavior, equipment status, and environmental parameters in real time, so the personnel path abnormality warning system will be used.

[0004] The existing personnel path abnormality warning system has a single warning mode and poor overall warning effect, which has a certain impact on the use of the personnel path abnormality warning system. Therefore, a personnel path abnormality warning system based on laboratory personnel identification and tracking is proposed. Summary of the invention

[0005] The technical problem to be solved by the present invention is: how to solve the problem that the existing personnel path abnormality warning system has a single warning mode and poor overall warning effect, which has brought certain impacts on the use of the personnel path abnormality warning system. A personnel path abnormality warning system based on laboratory personnel identification and tracking is provided.

[0006] The present invention solves the above technical problems through the following technical solutions, which include an entrance image acquisition module, an in-laboratory image acquisition module, an intelligent label acquisition module, a data processing module and an information sending module;

[0007] The entrance image acquisition module is used to collect real-time facial images of people entering the laboratory at the laboratory entrance;

[0008] The laboratory image acquisition module is used to collect real-time images of people entering the laboratory, that is, real-time images of people entering the laboratory;

[0009] Smart tag collection module, used to obtain real-time location information through RFID devices worn by personnel entering the laboratory;

[0010] The data processing module is used to process the real-time face image to obtain the personnel verification warning information;

[0011] Process the real-time image of the entering personnel to obtain the warning information of abnormal approaching of personnel;

[0012] Process the real-time position information and the real-time image of the entering personnel to generate the warning information of abnormal personnel path;

[0013] The information sending module is used to send the personnel verification warning information, the warning information of abnormal approaching of personnel and the warning information of abnormal personnel path to the preset receiving terminal after generating them.

[0014] Furthermore, after the entrance image acquisition module acquires the real-time face image of the personnel entering the laboratory, it preprocesses it. The specific acquisition process and preprocessing process are as follows:

[0015] The entrance image acquisition module uses a camera to acquire face images at a frequency of 5-10 frames per second; in order to adapt to different lighting conditions, the camera is equipped with adaptive exposure and white balance adjustment functions, and in low-light environments, near-infrared fill light is automatically turned on;

[0016] Preprocess the acquired face image, including image enhancement, normalization and denoising processing;

[0017] Use the histogram equalization algorithm to enhance the contrast of the image;

[0018] Unify the pixel values of the image to a preset range through normalization operation;

[0019] Use the median filtering algorithm to remove salt-and-pepper noise in the image.

[0020] Utilize the convolutional neural network (CNN) architecture to extract a face feature vector containing 200-300 feature points.

[0021] Furthermore, the specific process of the data processing module for processing the real-time face image to obtain the personnel verification warning information is as follows:

[0022] Calculate the cosine similarity between the face feature vector containing 200-300 feature points and the pre-stored legal personnel database;

[0023] Use the KD-tree or locality-sensitive hashing (LSH) algorithm to index the database to reduce the search range;

[0024] At the same time, conduct a confidence evaluation on the similarity calculation result. If the confidence evaluation fails, it means the verification fails;

[0025] The threshold is set at 0.85, but in actual applications, it is dynamically adjusted according to the usage frequency and risk level of personnel. For core personnel who often use the laboratory, the threshold can be appropriately lowered to increase the passing rate of verification; for temporary personnel or personnel in high-risk areas, the threshold is increased to enhance security.

[0026] Liveness detection is performed through blink detection and head pose analysis. Failure in liveness detection indicates verification failure.

[0027] After two consecutive verification failures, the system automatically activates secondary verification (such as entering the work number). The method of secondary verification can be diversified according to the actual situation. In addition to entering the work number, methods such as SMS verification code, fingerprint recognition, or voice recognition can also be used. To improve security, the information for secondary verification is transmitted through an encrypted channel to prevent information leakage.

[0028] During the secondary verification process, the system records the user's operation behavior and time, analyzes the user's input speed and input accuracy characteristics to determine whether there are abnormal behaviors. When abnormalities are found, a personnel verification warning message is immediately generated.

[0029] If there are a cumulative of five verification failures (including secondary verification), a personnel verification warning message is also generated.

[0030] When generating a personnel verification warning message, the time, location, verification method used, and specific reasons for failure of the verification are recorded in detail.

[0031] After generating the personnel verification warning message, the system automatically conducts a risk assessment on this person, determines the risk level of this person based on factors such as the number of failures, reasons for failure, and the person's historical records. For high-risk personnel, their permission to enter the laboratory is restricted, and the laboratory safety management department is notified for further investigation.

[0032] Furthermore, the specific process for obtaining the warning information about a person approaching abnormally is as follows:

[0033] Extract the real-time image of the entering person collected. First, process the real-time image of the entering person to obtain the person's identity information, import the person's identity information into a preset database, and retrieve the information about the experimental equipment that the person is allowed to approach from the preset database.

[0034] Then, process the real-time image of the entering person again, retrieve the image of the experimental equipment that the person is not allowed to approach from the real-time image of the entering person. When the image of the experimental equipment that is not allowed to approach is retrieved, establish a warning range, and then process the real-time image of the entering person to obtain the current position. When the current position is within the warning range.

[0035] Furthermore, the process for establishing the warning range is as follows:

[0036] Extract the image of the experimental equipment, extract its center point from the image of the experimental equipment, draw a circle with the center point as the center and a preset length as the radius to obtain a preliminary area, and then draw a rectangle circumscribing the preliminary area, that is, obtain the finally determined area, and the inner range of the finally determined area is the warning range.

[0037] Furthermore, the specific process of processing the real-time position information and the real-time image of the entering personnel to generate the warning information of abnormal personnel path is as follows:

[0038] First, preprocess the real-time image of the entering personnel, including denoising and enhancement;

[0039] Use computer vision techniques (such as background subtraction method, optical flow method, etc.) to analyze the preprocessed image, extract the contour and motion features of the personnel, and determine the position coordinates of the personnel in the image;

[0040] According to the position coordinates of the personnel in consecutive frames, calculate the motion trajectory of the personnel, that is, the real-time path information;

[0041] Fuse the real-time position information (position coordinates) collected by the RFID device with the real-time path information (motion trajectory) obtained through image analysis;

[0042] Establish a preset safe path model, which is formulated according to factors such as the layout of the laboratory, the location of equipment, and the operation specifications of personnel, and stipulates the normal walking path and activity area of personnel in the laboratory;

[0043] Compare and analyze the fused real-time path information with the safe path model, and use a path matching algorithm (such as dynamic time warping algorithm) to calculate the similarity between the two;

[0044] When the similarity between the fused real-time path information and the safe path model is less than the preset value, it is determined that the path is abnormal.

[0045] The present invention has the following advantages compared with the prior art: The personnel path abnormal warning system based on laboratory personnel identification and tracking collects face images at a frequency of 5-10 frames per second through the high-definition camera of the entrance image acquisition module, and performs preprocessing such as image enhancement, normalization, and denoising, and then uses a convolutional neural network to extract a face feature vector containing 200-300 feature points, which greatly improves the accuracy of personnel identification and can effectively avoid misidentification caused by factors such as poor lighting conditions and poor image quality, and reduces the false alarm rate.

[0046] Verify the personnel identity by using a variety of means such as cosine similarity calculation, KD-tree or locality-sensitive hashing algorithm indexing, confidence evaluation, liveness detection, and secondary verification.

[0047] When generating the personnel verification warning information, the system records in detail the time, location, verification method used, and specific reasons for verification failure, and conducts a risk assessment on the personnel to determine their risk level. This helps the management personnel to timely understand the specific details of abnormal situations, facilitating subsequent investigations and handling. At the same time, it also provides important data support for the safety management of the laboratory.

[0048] The in-lab image acquisition module can acquire personnel images in real time. The data processing module obtains personnel identity information by analyzing the images, retrieves the information of experimental equipment that the personnel are allowed to approach from a preset database, and simultaneously retrieves the images of experimental equipment that the personnel are not allowed to approach. When a person approaches equipment that they are not allowed to approach, it can be detected in time and a warning is issued.

[0049] Fuse the real-time position information collected by RFID devices with the real-time path information obtained through image analysis, combining the high-precision positioning of RFID technology and the trajectory tracking advantages of computer vision technology, improving the reliability and accuracy of path monitoring, and being able to effectively handle errors or failures that may occur in a single technology;

[0050] Establish a preset safe path model, and use a path matching algorithm to compare and analyze the fused real-time path information with the safe path model. When the similarity is less than the preset value, it is determined that the path is abnormal and a warning message is immediately issued. This intelligent analysis method can quickly and accurately identify abnormal behaviors of personnel, gain valuable processing time for management personnel, realize the identification, tracking, and path monitoring of laboratory personnel through intelligent means, reduce the workload of manual inspections, and improve management efficiency. At the same time, information such as detailed records and risk assessment reports automatically generated by the system also provides strong support for management personnel, enabling them to make more scientific decisions and management, making this system more worthy of popularization and use. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is the system block diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] The following details the embodiments of the present invention. These embodiments are implemented on the premise of the technical solution of the present invention, and provide detailed implementation manners and specific operation processes. However, the protection scope of the present invention is not limited to the following embodiments.

[0053] As Figure 1 shown, this embodiment provides a technical solution: a personnel path anomaly warning system based on laboratory personnel identification and tracking, including an entrance image acquisition module, an in-lab image acquisition module, an intelligent tag acquisition module, a data processing module, and an information sending module;

[0054] The entrance image acquisition module is used to acquire the real-time face images of the personnel entering the laboratory at the entrance of the laboratory;

[0055] The in-laboratory image acquisition module is used to acquire the real-time images of the personnel entering the laboratory, that is, the real-time images of the entering personnel;

[0056] The intelligent tag acquisition module is used to obtain the real-time location information through the RFID device worn by the personnel entering the laboratory;

[0057] The data processing module is used to process the real-time face images to obtain the personnel verification warning information;

[0058] Process the real-time images of the entering personnel to obtain the personnel abnormal approach warning information;

[0059] Process the real-time location information and the real-time images of the entering personnel to generate the personnel path abnormal warning information;

[0060] The information sending module is used to send the personnel verification warning information, the personnel abnormal approach warning information and the personnel path abnormal warning information to the preset receiving terminal after they are generated.

[0061] After the entrance image acquisition module acquires the real-time face images of the personnel entering the laboratory, it preprocesses them. The specific acquisition process and preprocessing process are as follows:

[0062] The entrance image acquisition module uses a camera to acquire face images at a frequency of 5-10 frames per second; in order to adapt to different lighting conditions, the camera is equipped with an adaptive exposure and white balance adjustment function, and in low-light environments, near-infrared fill light is automatically turned on;

[0063] Preprocess the acquired face images, including image enhancement, normalization and denoising processing;

[0064] Use the histogram equalization algorithm to enhance the contrast of the image;

[0065] Unify the pixel values of the image to a preset range through normalization operation;

[0066] Use the median filter algorithm to remove the salt-and-pepper noise in the image.

[0067] Utilize the convolutional neural network (CNN) architecture to extract a face feature vector containing 200-300 feature points;

[0068] The acquisition and preprocessing process of the above-mentioned entrance image acquisition module has significant advantages. First of all, the camera captures face images at a frequency of 5-10 frames per second and is equipped with adaptive exposure and white balance adjustment functions, which can adapt to different lighting conditions to ensure clear images. In low light conditions, near-infrared fill light is automatically turned on to further ensure the acquisition effect, enabling the system to operate stably in various environments. Secondly, in the preprocessing stage, histogram equalization is used to enhance the image contrast and highlight face details; normalization operations unify the pixel value range to improve the system's compatibility and stability; median filtering removes salt-and-pepper noise to enhance the image quality. These processing steps effectively solve problems such as lighting changes and noise interference, providing high-quality images for subsequent recognition. Finally, a face feature vector with 200-300 feature points is extracted using a convolutional neural network (CNN), which automatically learns complex features to handle expression and pose changes and improve the recognition accuracy. Overall, this process significantly improves the robustness, accuracy, and adaptability of face acquisition and recognition, providing a solid foundation for the identification and management of laboratory personnel and effectively ensuring the safe and efficient operation of the laboratory.

[0069] The specific process for the data processing module to process real-time face images to obtain personnel verification warning information is as follows:

[0070] Calculate the cosine similarity between the face feature vector containing 200-300 feature points and the pre-stored legal personnel database;

[0071] Use the KD-tree or locality-sensitive hashing (LSH) algorithm to index the database to reduce the search range;

[0072] At the same time, conduct a confidence assessment on the similarity calculation result. If the confidence assessment fails, it indicates that the verification fails;

[0073] The threshold is set to 0.85, but in actual applications, it is dynamically adjusted according to the usage frequency and risk level of personnel. For core personnel who often use the laboratory, the threshold can be appropriately lowered to increase the verification passing rate; for temporary personnel or personnel in high-risk areas, the threshold is increased to enhance security.

[0074] Conduct a live detection through blink detection and head pose analysis. If the live detection fails, it indicates that the verification fails;

[0075] After two consecutive verification failures, the system automatically activates secondary verification (such as entering the work number). The method of secondary verification can be diversified according to the actual situation. In addition to entering the work number, it can also use methods such as SMS verification codes, fingerprint recognition, or voice recognition. To improve security, the information for secondary verification is transmitted through an encrypted channel to prevent information leakage.

[0076] During the secondary verification process, the system will record the user's operation behavior and time, analyze the user's input speed and input accuracy characteristics to determine whether there is any abnormal behavior. When an abnormality is found, a personnel verification warning message will be generated immediately.

[0077] If there are 5 cumulative verification failures (including secondary verification), a personnel verification warning message will also be generated;

[0078] When generating personnel verification warning information, the time, location, verification method used and specific reasons for failure of verification shall be recorded in detail;

[0079] After generating the personnel verification warning information, the system automatically conducts a risk assessment on the personnel and determines the risk level of the personnel based on the number of failures, reasons for failures and the personnel's historical records. For high-risk personnel, their access to the laboratory is restricted and the laboratory safety management department is notified for further investigation;

[0080] By calculating the cosine similarity between the facial feature vector and the legal personnel database, and combining the KD-tree or local sensitive hashing (LSH) algorithm to index the database, the comparison can be completed quickly and accurately, the search scope can be reduced, the verification efficiency can be greatly improved, and the efficient management of laboratory personnel entering and exiting can be ensured.

[0081] Flexible threshold adjustment and adaptability The threshold is adjusted dynamically and flexibly set according to the frequency of use and risk level of personnel. The threshold of core personnel is appropriately lowered to improve the pass rate and ensure the normal operation of the laboratory; the threshold of temporary or high-risk personnel is increased to enhance safety. This flexible threshold strategy ensures safety while taking into account the efficient operation of the laboratory.

[0082] Liveness detection and secondary verification ensure security. Liveness detection is combined with blink detection and head posture analysis to effectively prevent non-live attacks such as photos and videos, and prevent illegal personnel from using technical means to get through from the source. Secondary verification is automatically enabled after two consecutive verification failures, and the secondary verification methods are diversified, such as entering the work number, SMS verification code, fingerprint recognition or voice recognition, etc., to further ensure the reliability of verification. At the same time, the secondary verification information is transmitted through an encrypted channel to prevent information leakage and comprehensively ensure the safety of the laboratory.

[0083] During the secondary verification process of abnormal behavior monitoring and risk assessment, the system records the user's operation behavior and time, analyzes the input speed and accuracy, judges abnormal behavior and generates warning information in a timely manner. If the verification fails five times in a row, a warning information will be generated, and the time, location, method and reason of the failure will be recorded in detail. After that, the system automatically conducts a risk assessment on the personnel, determines the risk level according to the number of failures, reasons and historical records, restricts the permissions of high-risk personnel and notifies the security department, which helps to detect potential security hazards in a timely manner and take targeted measures. Generally speaking, this process greatly improves the accuracy, security and management efficiency of laboratory personnel verification through multi-dimensional verification means, flexible threshold strategies, abnormal behavior monitoring and risk assessment, etc., providing a strong guarantee for the safe operation of the laboratory.

[0084] The specific process of obtaining the warning information when the person is abnormally close is as follows:

[0085] Extract the real-time image of the entering person that has been collected. First, process the real-time image of the entering person to obtain the person's identity information, import the person's identity information into a preset database, and retrieve the information of the experimental equipment that the person is allowed to approach from the preset database;

[0086] After that, process the real-time image of the entering person again, retrieve the image of the experimental equipment that the person is not allowed to approach from the real-time image of the entering person. When the image of the experimental equipment that is not allowed to approach is retrieved, establish a warning range, and then process the real-time image of the entering person to obtain the current position. When the current position is within the warning range.

[0087] Furthermore, the process of establishing the warning range is as follows:

[0088] Extract the image of the experimental equipment, extract its center point from the image of the experimental equipment, draw a circle with the center point as the center and a preset length as the radius to obtain a preliminary area, and then draw a rectangle circumscribing the preliminary area to obtain the finally determined area. The inner area of the finally determined area is the warning range;

[0089] First, obtain the person's identity information through the real-time image, and then retrieve the information of the equipment that the person is allowed to approach from the database, making the management more targeted, accurately distinguishing the permissions of different personnel, and effectively avoiding personnel's misoperations or entering dangerous areas.

[0090] Retrieve the image of the equipment that is not allowed to approach and establish a warning range, and issue a warning in a timely manner when the person approaches, which can effectively prevent the person from approaching dangerous equipment due to negligence or illegal operations and reduce the risk of laboratory safety accidents.

[0091] Draw a circle with the center point of the device image as the center, and then draw a circumscribed rectangle to determine the warning range. The method is scientific and reasonable, which not only ensures the accuracy of the warning range, but also avoids misjudgment or omission caused by too large or too small a range, making the warning more scientific and effective.

[0092] Through the above process, the laboratory can monitor the personnel dynamics in real time, discover and correct abnormal approaching behaviors in time, ensure the safety of experimental personnel, and at the same time reduce the experiment interruption and resource waste caused by accidents, improving the overall operation efficiency of the laboratory.

[0093] The specific process of processing the real-time position information and the real-time image of the entering personnel to generate the warning information of abnormal personnel path is as follows:

[0094] First, preprocess the real-time image of the entering personnel, including denoising and enhancement;

[0095] Use computer vision techniques (such as background subtraction method, optical flow method, etc.) to analyze the preprocessed image, extract the contour and motion features of the personnel, and determine the position coordinates of the personnel in the image;

[0096] According to the position coordinates of the personnel in consecutive frames, calculate the motion trajectory of the personnel, that is, the real-time path information;

[0097] Fuse the real-time position information (position coordinates) collected by the RFID device with the real-time path information (motion trajectory) obtained through image analysis;

[0098] Establish a preset safe path model, which is formulated according to factors such as the layout of the laboratory, the position of equipment, and the operation specifications of personnel, and stipulates the normal walking path and activity area of personnel in the laboratory;

[0099] Compare and analyze the fused real-time path information with the safe path model, and use a path matching algorithm (such as dynamic time warping algorithm) to calculate the similarity between the two;

[0100] When the similarity between the fused real-time path information and the safe path model is less than the preset value, it is determined that the path is abnormal;

[0101] By fusing the real-time position information collected by the RFID device and the real-time path information obtained through image analysis, it combines the advantages of high-precision positioning and intuitive trajectory tracking, effectively making up for the deficiencies of single technologies. Using computer vision techniques such as background subtraction method and optical flow method, accurately extract the contour and motion features of personnel and determine the position coordinates, and can accurately track even in complex environments. At the same time, using path matching algorithms such as dynamic time warping algorithm to compare the real-time path with the safe path model can handle time differences and local deformations, further improving the monitoring accuracy.

[0102] Strengthen laboratory safety management. This process realizes real-time monitoring of the movement trajectories of personnel. Once an anomaly is detected, a warning is immediately issued to promptly stop any violations or dangerous operations, preventing safety accidents caused by entering dangerous areas by mistake or violating operating procedures. The safety path model is formulated based on the laboratory layout, equipment locations, and operating specifications, fitting the actual needs and adapting to complex environments to ensure the safety and controllability of personnel activities.

[0103] Improve management efficiency. The decision support system records the real-time path information of personnel and the comparison results, providing detailed activity records and analysis reports for management personnel to help identify potential safety hazards and promptly adjust safety strategies and operating specifications. This process realizes the automation and intelligence of personnel path monitoring, reducing the workload of manual inspections and subjective errors, and improving management efficiency. The automatically generated warning information and data analysis reports provide strong support for management decisions, facilitating the scientific management of laboratory safety and personnel scheduling. In short, through multi-source data fusion, the application of advanced algorithms, and personalized safety path models, this process significantly improves the accuracy and reliability of monitoring, strengthens laboratory safety management, improves management efficiency and decision support capabilities, and provides strong guarantees for the safe operation of the laboratory.

[0104] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of the present invention, "a plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0105] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0106] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A personnel path anomaly warning system based on laboratory personnel identification and tracking, characterized in that, It includes an entrance image acquisition module, an in-laboratory image acquisition module, an intelligent tag acquisition module, a data processing module, and an information sending module; The entrance image acquisition module is used to acquire the real-time face image of the person entering the laboratory at the laboratory entrance; The in-laboratory image acquisition module is used to acquire the real-time image of the person entering the laboratory, that is, the real-time image of the entering person; The intelligent tag acquisition module is used to obtain the real-time position information through the RFID device worn by the person entering the laboratory; The data processing module is used to process the real-time face image to obtain the personnel verification warning information; Process the real-time image of the entering person to obtain the warning information of abnormal approach of the person; Process the real-time position information and the real-time image of the entering person to generate the warning information of abnormal personnel path; The information sending module is used to send the personnel verification warning information, the warning information of abnormal approach of the person, and the warning information of abnormal personnel path to the preset receiving terminal after they are generated.

2. The personnel path anomaly warning system based on laboratory personnel identification and tracking according to claim 1, characterized in that: After the entrance image acquisition module acquires the real-time face image of the person entering the laboratory, it preprocesses it. The specific acquisition process and preprocessing process are as follows: The entrance image acquisition module uses a camera to acquire face images at a frequency of 5-10 frames per second; in order to adapt to different lighting conditions, the camera is equipped with an adaptive exposure and white balance adjustment function, and in low-light environments, near-infrared fill light is automatically turned on; Preprocess the acquired face image, including image enhancement, normalization, and noise reduction processing; Use the histogram equalization algorithm to enhance the contrast of the image; Unify the pixel values of the image to a preset range through normalization operation; Use the median filter algorithm to remove salt-and-pepper noise in the image; Utilize the convolutional neural network architecture to extract a face feature vector containing 200-300 feature points.

3. The personnel path anomaly warning system based on laboratory personnel identification and tracking according to claim 2, wherein: The specific process of the data processing module for processing the real-time face image to obtain the personnel verification warning information is as follows: Calculate the cosine similarity between the face feature vector containing 200-300 feature points and the pre-stored legal personnel database; Use the KD-tree or locality-sensitive hashing algorithm to index the database to reduce the search range; At the same time, conduct a confidence evaluation on the similarity calculation result. If the confidence evaluation fails, it means the verification fails; Conduct a live detection through blink detection and head pose analysis. If the live detection fails, it means the verification fails; After 2 consecutive verification failures, the system automatically starts the secondary verification. During the secondary verification process, the system will record the user's operation behavior and time, analyze the user's input speed and input accuracy characteristics to determine whether there is abnormal behavior. When abnormal behavior is found, the personnel verification warning information is immediately generated; If the cumulative number of verification failures reaches 5 times, the personnel verification warning information is also generated; When generating the personnel verification warning information, record in detail the time, location, verification method used, and specific failure reasons of the verification failure; After generating the personnel verification warning information, the system automatically conducts a risk assessment on this person, and determines the risk level of this person according to factors such as the number of failures, failure reasons, and the person's historical records.

4. A personnel path anomaly warning system based on laboratory personnel identification and tracking according to claim 1, characterized in that: The specific process of obtaining the warning information for the abnormal approach of the personnel is as follows: Extract the real-time image of the entering personnel that has been collected. First, process the real-time image of the entering personnel to obtain the personnel identity information, import the personnel identity information into the preset database, and retrieve the information of the experimental equipment that the personnel is allowed to approach from the preset database; After that, process the real-time image of the entering personnel again, retrieve the image of the experimental equipment that the person is not allowed to approach from the real-time image of the entering personnel. When the image of the experimental equipment that is not allowed to approach is retrieved, establish a warning range, and then process the real-time image of the entering personnel to obtain the current position. When the current position is within the warning range.

5. The personnel path anomaly warning system based on laboratory personnel identification and tracking according to claim 4, wherein: The process of establishing the warning range is as follows: Extract the image of the experimental equipment, extract its center point from the image of the experimental equipment, draw a circle with the center point as the center and a preset length as the radius to obtain a preliminary area, and then draw a rectangle circumscribing the preliminary area, that is, obtain the finally determined area. The inner area of the finally determined area is the warning range.

6. The personnel path anomaly warning system based on laboratory personnel identification and tracking according to claim 1, wherein: The specific process of processing the real-time position information and the real-time image of the entering personnel to generate the warning information for abnormal personnel path is as follows: First, preprocess the real-time image of the entering personnel, including denoising and enhancement; Use computer vision technology to analyze the preprocessed image, extract the contour and motion features of the personnel, and determine the position coordinates of the personnel in the image; According to the position coordinates of the personnel in consecutive frames, calculate the motion trajectory of the personnel, that is, the real-time path information; Fuse the real-time position information collected by the RFID device with the real-time path information obtained through image analysis; Establish a preset safe path model, compare and analyze the fused real-time path information with the safe path model, and use a path matching algorithm to calculate the similarity between the two; When the similarity between the fused real-time path information and the safe path model is less than the preset value, it is determined that the path is abnormal.

Images