A machine room scene inspection personnel track tracking and identification method

Abstract

This invention provides a method for tracking and identifying the trajectory of inspection personnel in a data center, comprising: performing facial recognition on staff at the entrance of the data center and determining the staff's identity information from a cloud material library based on the recognition results; determining the time and location of the staff in the data center based on multiple cameras deployed in the data center and the identity information, and uploading the data to the cloud; the cloud sorting the locations based on the time of the staff in the data center and automatically generating a dynamic travel trajectory map of the personnel based on the sorting results; when equipment in the data center is lost or damaged, the system can quickly determine the list of personnel whose movement trajectories involve the original location of the lost equipment based on the historical records in the cloud, thereby eliminating possibilities one by one and quickly and accurately locating the responsible person, achieving intelligent supervision of the data center.

Description

Technical Field

[0001] This invention relates to the field of data center inspection technology, and in particular to a method for tracking and recognizing the trajectory of inspection personnel in a data center setting. Background Technology

[0002] As the physical storage location for major network equipment and processors, the data center is constantly exchanging data in real time. To ensure the security of equipment and processors and maintain the data center environment, it is necessary to monitor personnel entry and exit and activity areas around the clock to effectively manage the smart data center. Real-time monitoring of the movement trajectories of personnel entering the data center allows administrators to trace responsibility based on cloud-recorded personnel movement trajectories when equipment or environmental problems occur, reducing investigation costs and ensuring the convenience and security of data center maintenance.

[0003] Existing technologies mostly rely on cameras to detect personnel activities, but they cannot detect the order and time of personnel arriving at activity locations. When equipment in the computer room is lost or damaged, the only way to trace responsibility is to search all monitoring records, which consumes a lot of manpower, material resources and financial resources and is inefficient. This invention solves the problems of existing technologies. Summary of the Invention

[0004] This invention provides a method for tracking and identifying the trajectory of inspection personnel in a data center, enabling real-time monitoring and rapid, accurate positioning of staff, thereby improving the accuracy and efficiency of identifying responsible persons.

[0005] A method for tracking and recognizing the trajectory of inspection personnel in a data center setting, comprising:

[0006] S1: Perform facial recognition on staff at the entrance of the server room and determine the staff's identity information from the cloud material library based on the recognition results;

[0007] S2: Based on the multiple cameras deployed in the computer room and the aforementioned identity information, determine the time and location of the staff member in the computer room, and upload the information to the cloud;

[0008] S3: The cloud platform sorts the locations based on the time points when the staff are in the computer room, and automatically generates a dynamic travel trajectory map of the staff based on the sorting results.

[0009] Preferably, in S1, the staff's identity information is determined from the cloud material library based on the recognition result, including:

[0010] The faces of the staff are compared with the faces of the materials in the cloud material library to obtain a set of materials that meet the first comparison requirements;

[0011] A second comparison is performed between the staff member's face and the aforementioned set of materials to determine whether there are target materials that meet the second comparison requirements;

[0012] If so, the identity information of the target material pre-registered in the cloud material library is determined as the identity information of the staff member;

[0013] Otherwise, the staff member will be deemed an irrelevant person.

[0014] Preferably, prior to S2, the method also includes: deploying multiple cameras inside the server room according to the server room layout, specifically:

[0015] Obtain the layout diagram of the computer room, determine the rack intersections in the layout diagram, and mark the rack intersections to obtain the marked points;

[0016] The camera field of view is simulated at the marked point, and the field of view range of the camera is determined to determine whether the total field of view range of all cameras covers the entire computer room.

[0017] If so, deploy a camera at the marked point;

[0018] Otherwise, identify the missing field of view, set the corresponding deployment point for it, and deploy the camera at the marked point and the deployment point.

[0019] Preferably, in step S2, based on the multiple cameras deployed in the server room and the aforementioned identity information, the time and location of the staff member in the server room are determined and uploaded to the cloud, including:

[0020] Based on the layout diagram of the computer room, name the multiple deployed cameras to obtain the camera names;

[0021] The system detects personnel arriving at the camera deployment point, obtains detection images, and uploads the detection images, along with the corresponding camera names and time points, to the cloud.

[0022] Preferably, in S3, the cloud sorts the locations according to the time points when staff are in the computer room, and automatically generates a dynamic travel trajectory map of the personnel based on the sorting results, including:

[0023] The detection images captured by the camera are identified, and head and shoulder features are extracted from the detection images. The recognition image at the entrance of the computer room is obtained, and the contour features of the recognition image are extracted.

[0024] Based on the head features, shoulder features, and contour features, the personnel information in the detected image is determined;

[0025] The location of the detected image is marked according to the camera name, and the time of the detected image is marked according to the shooting time.

[0026] Based on personnel information, target detection images that satisfy the target identity are obtained from detection images from multiple cameras;

[0027] The target detection images are sorted according to the time stamp, and the dynamic travel trajectory map of the personnel is determined based on the sorting result.

[0028] Preferably, based on the head features, shoulder features, and contour features, the personnel information in the detected image is determined, including:

[0029] The head and shoulder features are matched with the contour features to determine the personnel information in the detected image;

[0030] The detection time point of the detected image and the recognition time point of the recognized image are compared to determine whether the time difference meets the preset time difference requirement.

[0031] If so, the detected images are labeled with the identities of the individuals in them;

[0032] Otherwise, the head and shoulder features are rematched with the contour features until the detected image meets the preset time difference requirement to obtain the final personnel information.

[0033] Preferably, the target detection images are sorted according to time stamps, and a dynamic travel trajectory map of the personnel is determined based on the sorting results, including:

[0034] The target detection images are sorted according to the time stamps, and a preliminary trajectory animation is generated based on the sorting results and the location stamps.

[0035] Based on the layout diagram of the computer room, constraint rules for the trajectory are generated, and it is determined whether the preliminary trajectory dynamic diagram satisfies the constraint rules.

[0036] If so, the preliminary trajectory animation is determined to be the personnel dynamic travel trajectory animation of the staff.

[0037] Otherwise, extract trajectory points that do not meet the constraint rules from the initial trajectory dynamic map, and correct the trajectory points according to the estimated points before and after the trajectory points, and the time and location of the trajectory points to obtain corrected trajectory points. Then, generate a dynamic travel trajectory map of the staff based on the corrected trajectory points.

[0038] Preferably, it also includes: after problems occur in the computer room environment and equipment, identifying the target staff member based on the personnel dynamic travel trajectory map, specifically:

[0039] Based on the abnormal time point, computer room location, and specific information of the problem, a third dynamic travel trajectory map is determined;

[0040] Based on specific information, anomaly attributes are determined. Based on the anomaly attributes and the location of the computer room, the trajectory of the target staff member is predicted to obtain the predicted trajectory.

[0041] The third dynamic travel trajectory map is compared with the predicted trajectory to obtain the trajectory similarity. The suspicious staff are then sorted from largest to smallest according to the trajectory similarity to obtain the sorting result.

[0042] Based on the comparative state travel trajectory map related to the third dynamic travel trajectory map in the personnel dynamic travel trajectory map, the sorting results are weighted and calculated to obtain a weighted sorting result, and the final target staff are determined.

[0043] Preferably, based on the abnormal time point where the problem occurred, the location of the data center, and specific information, a third dynamic travel trajectory map is determined, including:

[0044] Determine the abnormal time point and the location of the data center where the problem occurred, obtain a first dynamic travel trajectory map before a preset time point of the abnormal time point, and obtain a second dynamic travel trajectory map related to the location of the data center from the first dynamic travel trajectory map;

[0045] Based on the second dynamic travel trajectory map, obtain the dwell time and number of times all staff members stayed at the computer room location;

[0046] Based on the specific information of the problem, the duration and number of stays are filtered to select suspicious staff members who meet the specific information regarding the duration and number of stays from all target personnel, and the third dynamic travel trajectory map corresponding to the suspicious staff members is obtained from the dynamic second travel trajectory map.

[0047] Preferably, based on the comparative state travel trajectory map related to the third dynamic travel trajectory map in the personnel dynamic travel trajectory map, the ranking result is weighted to obtain a weighted ranking result, and the final target staff member is determined, including:

[0048] Obtain a comparative travel trajectory map related to the third dynamic travel trajectory map from the personnel dynamic travel trajectory map, and determine the daily trajectory characteristics of the suspicious staff member based on the comparative travel trajectory map;

[0049] The single trajectory features of the third dynamic travel trajectory map are compared with the daily trajectory features to obtain the feature similarity. The ranking results are then weighted based on the value of the feature similarity to obtain the weighted ranking result.

[0050] Based on the weighted sorting results, retrieve the surveillance records of the suspicious staff members to determine the final target staff member.

[0051] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings.

[0052] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0053] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0054] Figure 1 This is a flowchart of a method for tracking and recognizing the trajectory of inspection personnel in a computer room scenario, as described in an embodiment of the present invention.

[0055] Figure 2 This is a flowchart illustrating the process of determining the identity information of staff members in an embodiment of the present invention;

[0056] Figure 3 This is a flowchart illustrating the process of determining the time and location of staff members in the computer room, as described in an embodiment of the present invention. Detailed Implementation

[0057] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0058] Example 1

[0059] A method for tracking and recognizing the trajectory of inspection personnel in a data center setting, such as Figure 1 As shown, it includes:

[0060] S1: Perform facial recognition on staff at the entrance of the server room and determine the staff's identity information from the cloud material library based on the recognition results;

[0061] S2: Based on the multiple cameras deployed in the computer room and the aforementioned identity information, determine the time and location of the staff member in the computer room, and upload the information to the cloud;

[0062] S3: The cloud platform sorts the locations of staff members in the computer room based on their time points and automatically generates a dynamic travel trajectory map of the staff members based on the sorting results.

[0063] In this embodiment, a facial recognition device is pre-installed at the entrance of the computer room, and the facial and identity information of the staff is pre-collected to form the cloud material library.

[0064] In this embodiment, cameras are pre-deployed at the intersections of various server racks and on the ceiling of the server room, such as... Figure 2 The effect is shown, and the corresponding cameras are named according to their placement locations. When personnel enter the equipment room and reach the positioned cameras, the cameras can vertically detect the person's head and shoulders, and transmit the recorded time and location to the cloud.

[0065] The beneficial effects of the aforementioned solution are as follows: by using facial recognition to identify staff at the entrance of the computer room, timely alerts can be issued for unauthorized intrusions, preventing damage to the computer room's property. By recording the location and time of personnel's activities in the cloud, a movement trajectory map can be generated. When equipment in the computer room is lost or damaged, the original location of the lost equipment can be quickly identified based on the historical records in the cloud, allowing for the elimination of individuals one by one and rapid, accurate identification of the responsible party, thus achieving intelligent supervision of the computer room.

[0066] Example 2

[0067] Based on Example 1, this embodiment of the invention provides a method for tracking and recognizing the trajectory of inspection personnel in a data center scenario, such as... Figure 2 As shown, in S1, the staff's identity information is determined from the cloud material library based on the recognition result, including:

[0068] S11: Perform a first comparison between the staff's face and the face data in the cloud material library to obtain a set of materials that meet the first comparison requirements;

[0069] S12: Perform a second comparison between the staff member's face and the set of materials to determine whether there is a target material that meets the second comparison requirements;

[0070] S13: If so, determine the identity information of the target material pre-registered in the cloud material library as the identity information of the staff;

[0071] Otherwise, the staff member will be deemed an irrelevant person.

[0072] In this embodiment, the first comparison is a preliminary comparison with low accuracy, and the set of materials that are determined to be matched consists of multiple materials.

[0073] In this embodiment, the second comparison is a precise comparison with high accuracy, ensuring that the matched target material is unique.

[0074] In this embodiment, after determining that the target material's pre-registered identity information in the cloud material library is the staff member's identity information, the method further includes allowing the staff member to enter the computer room.

[0075] In this embodiment, if a staff member is determined to be an unauthorized person, they are not allowed to enter the computer room, and a warning is issued.

[0076] The beneficial effects of the above design scheme are: by using facial recognition to identify the identity information of personnel entering the computer room, a basis for trajectory determination is provided, and timely alarms are issued for unauthorized personnel intrusion, thus preventing the security of computer room property from being compromised.

[0077] Example 3

[0078] Based on Example 1, this embodiment of the invention provides a method for tracking and recognizing the trajectory of inspection personnel in a data center scene. Before S2, it further includes: deploying multiple cameras inside the data center according to the layout of the data center, specifically:

[0079] Obtain the layout diagram of the computer room, determine the rack intersections in the layout diagram, and mark the rack intersections to obtain the marked points;

[0080] The camera field of view is simulated at the marked point, and the field of view range of the camera is determined to determine whether the total field of view range of all cameras covers the entire computer room.

[0081] If so, deploy a camera at the marked point;

[0082] Otherwise, identify the missing field of view, set the corresponding deployment point for it, and deploy the camera at the marked point and the deployment point.

[0083] The beneficial effects of the above design scheme are: deploying corresponding cameras at the intersection of server racks ensures a wide field of view for the cameras, facilitating the recording of staff trajectories; at the same time, deploying cameras in areas not covered by the cameras at the server rack intersections ensures that the deployed cameras can capture the entire scene inside the server room, providing a basis for determining staff trajectories.

[0084] Example 4

[0085] Based on Embodiment 1, this embodiment of the invention provides a method for tracking and recognizing the trajectory of inspection personnel in a data center scenario, such as... Figure 3 As shown, in S2, based on the multiple cameras deployed in the server room and the aforementioned identity information, the time and location of the staff member in the server room are determined and uploaded to the cloud, including:

[0086] S21: Based on the layout diagram of the computer room, name the multiple deployed cameras to obtain the camera names;

[0087] S22: Detect the personnel who arrive at the camera deployment point, obtain the detection image, and upload the detection image, the corresponding camera name, and the corresponding time point to the cloud.

[0088] The beneficial effects of the above design scheme are: by naming the cameras according to their location in the computer room and uploading the detection images and time points captured by the cameras to the cloud, a data foundation is provided for determining the trajectory of staff in the cloud.

[0089] Example 5

[0090] Based on Embodiment 1, this embodiment of the invention provides a method for tracking and recognizing the trajectory of inspection personnel in a data center scenario. In S3, the cloud platform sorts the locations of the personnel's time points in the data center and automatically generates a dynamic travel trajectory map of the personnel based on the sorting results, including:

[0091] The detection images captured by the camera are identified, and head and shoulder features are extracted from the detection images. The recognition image at the entrance of the computer room is obtained, and the contour features of the recognition image are extracted.

[0092] The head and shoulder features are matched with the contour features to determine the personnel information in the detected image;

[0093] The detection time point of the detected image and the recognition time point of the recognized image are compared to determine whether the time difference meets the preset time difference requirement.

[0094] If so, the detected images are labeled with the identities of the individuals in them;

[0095] Otherwise, the head and shoulder features are rematched with the contour features until the preset time difference requirement is met, and the final personnel information is obtained.

[0096] The location of the detected image is marked according to the camera name, and the time of the detected image is marked according to the shooting time.

[0097] Based on the identity marker, obtain the target detection image that meets the target identity from the detection images of multiple cameras;

[0098] The target detection images are sorted according to the time stamps, and a preliminary trajectory animation is generated based on the sorting results and the location stamps.

[0099] Based on the layout diagram of the computer room, constraint rules for the trajectory are generated, and it is determined whether the preliminary trajectory dynamic diagram satisfies the constraint rules.

[0100] If so, the preliminary trajectory animation is determined to be the personnel dynamic travel trajectory animation of the staff.

[0101] Otherwise, extract trajectory points that do not meet the constraint rules from the initial trajectory dynamic map, and correct the trajectory points according to the estimated points before and after the trajectory points, and the time and location of the trajectory points to obtain corrected trajectory points. Then, generate a dynamic travel trajectory map of the staff based on the corrected trajectory points.

[0102] In this embodiment, the detection image mainly captures the head and shoulders of the staff, so it is necessary to match the head and shoulder features in the detection image with the contour features in the recognition image.

[0103] In this embodiment, the detection time point of the detection image and the recognition time point of the recognition image are compared by time difference. For example, the recognition time point of the recognition image is 15:35, while the detection time point of the detection image is 15:40. According to the camera location of the detection image, it takes at least 10 minutes to get from the door of the computer room to the camera location. Therefore, it is determined that the detection time point of the detection image is inaccurate.

[0104] In this embodiment, the constraint rule of the trajectory is, for example, that the previous time point corresponds to the previous location and the next time point corresponds to the next location. Based on this, the possible locations corresponding to the current time point are determined. If it is detected that the current location corresponding to the current time point in the preliminary trajectory animation is not among the possible locations, then it is determined that the preliminary trajectory animation does not satisfy the constraint rule.

[0105] The beneficial effects of the above design scheme are: by marking the detected images with time, identity, and location, and sorting them according to the time and location, the accuracy of the final dynamic travel trajectory map of personnel is guaranteed. This provides a data foundation for quickly and accurately locating the responsible person and realizing intelligent supervision of the computer room.

[0106] Example 6

[0107] Based on Embodiment 1, this embodiment of the invention provides a method for tracking and identifying the trajectory of inspection personnel in a data center scenario, further comprising: determining the target personnel based on the personnel dynamic travel trajectory map after problems occur in the data center environment and equipment, specifically:

[0108] Determine the abnormal time point and the location of the data center where the problem occurred, obtain a first dynamic travel trajectory map before a preset time point of the abnormal time point, and obtain a second dynamic travel trajectory map related to the location of the data center from the first dynamic travel trajectory map;

[0109] Based on the second dynamic travel trajectory map, obtain the dwell time and number of times all staff members stayed at the computer room location;

[0110] Based on the specific information of the problem, the duration and number of stays are filtered to select suspicious staff members who meet the specific information regarding the duration and number of stays from all target personnel, and the third dynamic travel trajectory map corresponding to the suspicious staff members is obtained from the dynamic second travel trajectory map.

[0111] Based on specific information, anomaly attributes are determined. Based on the anomaly attributes and the location of the computer room, the trajectory of the target staff member is predicted to obtain the predicted trajectory.

[0112] The third dynamic travel trajectory map is compared with the predicted trajectory to obtain the trajectory similarity. The suspicious staff are then sorted from largest to smallest according to the trajectory similarity to obtain the sorting result.

[0113] Obtain a comparative travel trajectory map related to the third dynamic travel trajectory map from the personnel dynamic travel trajectory map, and determine the daily trajectory characteristics of the suspicious staff member based on the comparative travel trajectory map;

[0114] The single trajectory features of the third dynamic travel trajectory map are compared with the daily trajectory features to obtain the feature similarity. The ranking results are then weighted based on the value of the feature similarity to obtain the weighted ranking result.

[0115] Based on the weighted sorting results, retrieve the surveillance records of the suspicious staff members to determine the final target staff member.

[0116] In this embodiment, if the specific information of the problem is, for example, a device malfunction, then the operating time of the device is determined based on the cause of the malfunction, and then the possible duration and number of times the staff may stay when the cause of the malfunction occurs are determined.

[0117] In this embodiment, the abnormal attribute is, for example, a device malfunction or a device loss.

[0118] The beneficial effects of the above design scheme are as follows: Based on the specific information of the problem in the computer room, suspicious staff members are first screened from all staff members according to the dynamic travel trajectory map of the personnel. Then, based on the similarity between the trajectory of the suspicious staff members and the predicted trajectory and daily trajectory characteristics, the investigation order of the suspicious staff members is determined. Staff members with a high degree of suspicion determined by the dynamic trajectory are checked one by one, which makes it easier to quickly and accurately locate the responsible person, improve the visibility of the computer room and reduce labor costs.

[0119] Example 7

[0120] Based on Embodiment 5, this embodiment of the invention provides a method for tracking and recognizing the trajectory of inspection personnel in a data center scenario. According to the layout diagram of the data center, constraint rules for the trajectory are generated, and it is determined whether the preliminary trajectory dynamic diagram satisfies the constraint rules, including:

[0121] Based on the layout diagram of the computer room, camera layout points are determined, and based on personnel trajectory characteristics, single constraint values ​​are generated for each camera layout point and other camera layout points. Based on the single constraint values, the total constraint value for each layout point is determined.

[0122] The formula for calculating the total constraint value K of the current layout points is as follows:

[0123]

[0124] Where β represents the importance of the current layout point in the layout diagram, where β is (1, 2), n represents the number of other camera layout points in the layout diagram excluding the current layout point, and ε represents the importance of the current layout point. i h represents the single constraint value between the current layout point and the i-th other camera layout point. i ε represents the distance between the current layout point and the i-th other camera layout point, m represents the number of closely spaced camera layout points whose single constraint value is greater than the preset single constraint value, and ε represents the distance between the current layout point and the i-th other camera layout point. j This represents the single constraint value between the current layout point and the j-th closely spaced camera layout point;

[0125] Based on the total constraint value of each layout point, the trajectory constraint value of the preliminary trajectory dynamic map is determined;

[0126] The formula for calculating the trajectory constraint value D of the preliminary trajectory dynamic graph is as follows:

[0127]

[0128] Where h represents the number of camera layout points in the initial trajectory animation, β ω K represents the importance of the ω-th camera placement point in the initial trajectory animation. ω This represents the total constraint value of the ω-th camera layout point in the initial trajectory animation;

[0129] The trajectory constraint values ​​based on the preliminary trajectory dynamic graph are compared with the standard constraint values ​​in the constraint rules;

[0130] If the trajectory constraint value is less than the standard constraint value, it is determined that the preliminary trajectory dynamic graph satisfies the constraint rule.

[0131] Otherwise, it is determined that the preliminary trajectory animation does not satisfy the constraint rules.

[0132] In this embodiment, the personnel trajectory features are determined based on the personnel's movement characteristics.

[0133] In this embodiment, the importance of the current layout point in the layout diagram is related to the number of times the current layout point appears in the historical trajectory; the more times it appears, the greater the importance value.

[0134] In this embodiment, the larger the constraint value, the higher the requirements for the layout points.

[0135] The beneficial effects of the above design scheme are: by judging whether the preliminary trajectory dynamic image meets the constraints based on the layout diagram of the computer room and the camera layout points, the accurate analysis of the preliminary trajectory dynamic image is achieved, providing a foundation for the subsequent processing of the preliminary trajectory dynamic image.

[0136] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for tracking and recognizing the trajectory of inspection personnel in a computer room scenario, characterized in that, include: S1: Perform facial recognition on staff at the entrance of the server room and determine the staff's identity information from the cloud material library based on the recognition results; S2: Based on the multiple cameras deployed in the computer room and the aforementioned identity information, determine the time and location of the staff member in the computer room, and upload the information to the cloud; S3: The cloud platform sorts the locations based on the time points when the staff are in the computer room, and automatically generates a dynamic travel trajectory map of the staff based on the sorting results; This also includes: identifying target staff members based on the aforementioned personnel dynamic travel trajectory map after problems occur in the computer room environment and equipment, specifically: Determine the abnormal time point and the location of the data center where the problem occurred, obtain a first dynamic travel trajectory map before a preset time point of the abnormal time point, and obtain a second dynamic travel trajectory map related to the location of the data center from the first dynamic travel trajectory map; Based on the second dynamic travel trajectory map, obtain the dwell time and number of times all staff members stayed at the computer room location; Based on the specific information of the problem, the duration and number of stays are filtered to select suspicious staff members who meet the specific information regarding the duration and number of stays from all target personnel, and the third dynamic travel trajectory map corresponding to the suspicious staff members is obtained from the dynamic second travel trajectory map. Based on specific information, anomaly attributes are determined. Based on the anomaly attributes and the location of the computer room, the trajectory of the target staff member is predicted to obtain the predicted trajectory. The third dynamic travel trajectory map is compared with the predicted trajectory to obtain the trajectory similarity. The suspicious staff are then sorted from largest to smallest according to the trajectory similarity to obtain the sorting result. Obtain a comparative travel trajectory map related to the third dynamic travel trajectory map from the personnel dynamic travel trajectory map, and determine the daily trajectory characteristics of the suspicious staff member based on the comparative travel trajectory map; The single trajectory features of the third dynamic travel trajectory map are compared with the daily trajectory features to obtain the feature similarity. The ranking results are then weighted based on the value of the feature similarity to obtain the weighted ranking result. Based on the weighted sorting results, retrieve the surveillance records of the suspicious staff members to determine the final target staff member.

2. The method for tracking and recognizing the trajectory of inspection personnel in a computer room scenario according to claim 1, characterized in that, In S1, the staff's identity information is determined from the cloud material library based on the recognition results, including: The faces of the staff are compared with the faces of the materials in the cloud material library to obtain a set of materials that meet the first comparison requirements; A second comparison is performed between the staff member's face and the aforementioned set of materials to determine whether there are target materials that meet the second comparison requirements; If so, the identity information of the target material pre-registered in the cloud material library is determined as the identity information of the staff member; Otherwise, the staff member will be deemed an irrelevant person.

3. The method for tracking and recognizing the trajectory of inspection personnel in a computer room scenario according to claim 1, characterized in that, Prior to S2, it also included: deploying multiple cameras inside the server room according to the server room layout, specifically: Obtain the layout diagram of the computer room, determine the rack intersections in the layout diagram, and mark the rack intersections to obtain the marked points; The camera field of view is simulated at the marked point, and the field of view range of the camera is determined to determine whether the total field of view range of all cameras covers the entire computer room. If so, deploy a camera at the marked point; Otherwise, identify the missing field of view, set the corresponding deployment point for it, and deploy the camera at the marked point and the deployment point.

4. The method for tracking and recognizing the trajectory of inspection personnel in a computer room scenario according to claim 1, characterized in that, In S2, based on multiple cameras deployed in the server room and the aforementioned identity information, the time and location of the staff member in the server room are determined and uploaded to the cloud, including: Based on the layout diagram of the computer room, name the multiple deployed cameras to obtain the camera names; The system detects personnel arriving at the camera deployment point, obtains detection images, and uploads the detection images, along with the corresponding camera names and time points, to the cloud.

5. The method for tracking and recognizing the trajectory of inspection personnel in a computer room scenario according to claim 1, characterized in that, In S3, the cloud platform sorts the locations based on the time points when staff members are in the computer room, and automatically generates a dynamic travel trajectory map of the personnel based on the sorting results, including: The detection images captured by the camera are identified, and head and shoulder features are extracted from the detection images. The recognition image at the entrance of the computer room is obtained, and the contour features of the recognition image are extracted. Based on the head features, shoulder features, and contour features, the personnel information in the detected image is determined; The location of the detected image is marked according to the camera name, and the time of the detected image is marked according to the shooting time. Based on personnel information, target detection images that satisfy the target identity are obtained from detection images from multiple cameras; The target detection images are sorted according to the time stamp, and the dynamic travel trajectory map of the personnel is determined based on the sorting result.

6. The method for tracking and recognizing the trajectory of inspection personnel in a computer room scenario according to claim 5, characterized in that, Based on the head features, shoulder features, and contour features, the personnel information in the detected image is determined, including: The head and shoulder features are matched with the contour features to determine the personnel information in the detected image; The detection time point of the detected image and the recognition time point of the recognized image are compared to determine whether the time difference meets the preset time difference requirement. If so, the detected images are labeled with the identities of the individuals in them; Otherwise, the head and shoulder features are rematched with the contour features until the detected image meets the preset time difference requirement to obtain the final personnel information.

7. The method for tracking and recognizing the trajectory of inspection personnel in a computer room scenario according to claim 5, characterized in that, The target detection images are sorted according to time stamps, and a dynamic travel trajectory map of the personnel is determined based on the sorting results, including: The target detection images are sorted according to the time stamps, and a preliminary trajectory animation is generated based on the sorting results and the location stamps. Based on the layout diagram of the computer room, constraint rules for the trajectory are generated, and it is determined whether the preliminary trajectory dynamic diagram satisfies the constraint rules. If so, the preliminary trajectory animation is determined to be the personnel dynamic travel trajectory animation of the staff. Otherwise, extract trajectory points that do not meet the constraint rules from the initial trajectory dynamic map, and correct the trajectory points according to the estimated points before and after the trajectory points, and the time and location of the trajectory points to obtain corrected trajectory points. Then, generate a dynamic travel trajectory map of the staff based on the corrected trajectory points.

Images