Power grid operation trajectory archive construction method, system, device and medium

Abstract

The application discloses a kind of power grid operation track archives construction methods, it is related to operation process safety control technical field, for solving the problem of existing operation site control difficulty, the method includes the following steps: calculating operation personnel positioning and carrying out operation personnel identity identification;According to video data, detection analysis is carried out, the detection includes safety measure target, operation personnel target and operation personnel safety dress condition;According to the personnel positioning, the identity identification and the detection analysis, operation process control judgment is carried out;The personnel positioning, the identity identification, the detection analysis and the process control judgment result are combined according to time sequence, and form multidimensional track archives.The application further discloses a kind of power grid operation track archives construction system, electronic equipment and computer storage medium.The application obtains the multidimensional track archives of operation personnel by detecting analysis, position identification to operation site.

Description

Technical Field

[0001] This invention relates to the field of work process safety management technology, and in particular to a method, system, equipment and medium for constructing power grid work trajectory files. Background Technology

[0002] Safety accidents at power grid work sites can result in significant personal and property losses. Therefore, it is essential to implement safety controls over power grid work processes to effectively reduce the incidence of accidents, minimize worker injuries and equipment damage, and simultaneously increase business efficiency and better fulfill production targets.

[0003] Currently, process control methods at power grid operation sites are mostly based on post-event analysis, which lacks timeliness for real-time risks. Some sites use a combination of video surveillance and manual spot checks, which fails to guarantee full coverage of the operation site in terms of time and space. Furthermore, real-time manual monitoring of the operation site is prone to overlooking critical risks due to human error. In addition, manual control incurs high labor costs. Moreover, current control methods struggle to digitally record on-site operations, making it impossible to create operation files and hindering subsequent safety inspections and analysis.

[0004] To ensure safety at work sites and to implement safety management of work site processes, it is necessary to address the existing technical problems of untimely management, poor management effectiveness, high management costs, and difficulty in achieving multi-dimensional and comprehensive management. Therefore, it is urgent to construct a multi-dimensional trajectory archive construction method and system for power grid operation safety. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, one of the objectives of this invention is to provide a method for constructing a power grid operation trajectory archive, which records the power grid operation process in multiple dimensions to form a multi-dimensional trajectory archive.

[0006] One of the objectives of this invention is achieved through the following technical solution:

[0007] A method for constructing power grid operation trajectory files includes the following steps:

[0008] Based on ultra-bandwidth and inertial measurement calculations, the location of workers is determined and their identities are identified.

[0009] Acquire on-site monitoring video data, and perform detection and analysis based on the video data. The detection includes safety measure targets, worker targets, and worker safety attire.

[0010] Based on the personnel location, identity recognition, and detection analysis, the operation process control judgment is made;

[0011] The personnel location, identity recognition, detection analysis, and process control judgment results are combined in chronological order to form a multi-dimensional trajectory file.

[0012] Furthermore, the positioning of operators based on ultra-bandwidth and inertial measurement calculations includes the following steps:

[0013] The coordinate information of the workers in the three-dimensional coordinate system is obtained through ultra-broadband base stations;

[0014] Solve the position equation based on the distance measurement equation, satisfying:

[0015]

[0016] Among them, P x P y and P z Let x represent the coordinates of the position at time k in the X, Y, and Z directions, respectively. i y i and z i Let d represent the coordinates of the i-th ultra-wideband base station relative to the origin. i Indicates the distance between equipment and personnel;

[0017] Based on inertial measurement data, state prediction is performed, satisfying:

[0018]

[0019] in, Let F be the predicted system state value at time k+1 for the i-th Sigma sampling point, F be the state transition matrix, B be the input control matrix, and u be the input control matrix. k|k It is a control input;

[0020] The observed values ​​are calculated based on the observation equation, satisfying the following:

[0021]

[0022] Among them, The observation value refers to the i-th Sigma sampling point, and H is the observation matrix;

[0023] Calculate the Kalman gain, satisfying:

[0024]

[0025] Among them, P zk, Let P be the autocovariance matrix of the observation vector. xk, Represents the cross-covariance matrix between the observation vector and the state vector;

[0026] Update the system state and covariance matrix to obtain the personnel location results, satisfying:

[0027]

[0028]

[0029] in, The system mean is obtained by weighting the Sigma sampling points in the predicted stage. The observed mean is calculated by weighting the Sigma sampling points.

[0030] Furthermore, the identification of the operator includes: obtaining the device ID identification code of the ultra-wideband device, matching the identification code with the employee number in the database, and obtaining the identification result of the operator.

[0031] Furthermore, on-site monitoring video data is acquired, and detection and analysis are performed based on the video data. The detection includes safety measure targets, worker targets, and the worker's safety attire, including the following steps:

[0032] The surveillance video data is preprocessed to obtain an RGB image sequence;

[0033] The input RGB image is used to detect safety measures and personnel targets at the work site using the DETR target detection algorithm, and the target category and corresponding detection box are output.

[0034] The RGB image is used to detect safe attire using the QueryDet small object detection algorithm, and the attire category and corresponding detection box are output.

[0035] Furthermore, the work process includes a start-up meeting, site deployment, work operation, and a end-of-shift meeting.

[0036] Furthermore, the work process control judgment includes:

[0037] The data for the initial class meeting were obtained by fitting a straight line using the least squares algorithm.

[0038] Detect preset targets in the site to obtain site deployment data;

[0039] The safety attire worn on site was inspected to obtain operational data;

[0040] The data from the final station meeting were obtained by fitting a straight line using the least squares algorithm.

[0041] Furthermore, the time is a directly obtained 19-bit timestamp.

[0042] The second objective of this invention is to provide a relay armature motion state detection system, which uses multiple devices combined with multi-dimensional behavior calculation and detection to obtain the multi-dimensional trajectory of the operator.

[0043] The second objective of this invention is achieved by the following technical solution:

[0044] A power grid operation trajectory archive construction system includes: a GPS device, a camera, an ultra-wideband base station, and an edge computing device; the GPS is used for operator positioning, the camera is used for on-site monitoring video recording, the ultra-wideband base station is used for operator positioning calculation and operator identification, and the edge computing device is used for processing data to obtain and store multi-dimensional trajectory archives. When the power grid operation trajectory archive construction system is running, the above-mentioned power grid operation trajectory archive construction method is executed.

[0045] A third objective of this invention is to provide an electronic device for performing one of the objectives of the invention, comprising a processor, a storage medium, and a computer program, wherein the computer program is stored in the storage medium, and when the computer program is executed by the processor, it implements the above-mentioned method for constructing power grid operation trajectory files.

[0046] The fourth objective of this invention is to provide a computer-readable storage medium storing one of the objectives of the invention, wherein a computer program is stored thereon, and when the computer program is executed by a processor, it implements the above-mentioned method for constructing power grid operation trajectory files.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] This invention uses ultra-wideband combined with inertial measurement for worker positioning and identification, achieving accurate results and real-time worker location. By detecting and analyzing real-time video data, the safety situation at the work site can be identified. In addition, this invention also identifies the work process and integrates all data to form a multi-dimensional trajectory archive of the work site. This not only achieves real-time data recording, providing a foundation for safety protection at the work site, but also creates archives that meet archiving and traceability requirements. Attached Figure Description

[0049] Figure 1 This is a flowchart of the method for constructing power grid operation trajectory files in Implementation Example 1;

[0050] Figure 2 This is a schematic diagram of the power grid operation trajectory file construction system in Example 2;

[0051] Figure 3 This is a structural block diagram of the electronic device in Embodiment 3. Detailed Implementation

[0052] The present invention will now be described in more detail with reference to the accompanying drawings. It should be noted that the following description of the present invention with reference to the accompanying drawings is merely illustrative and not restrictive. Various embodiments can be combined with each other to form other embodiments not shown in the following description.

[0053] Example 1

[0054] Example 1 provides a method for constructing power grid operation trajectory files, which constructs power grid operation trajectory files from multiple dimensions based on an algorithm for operator positioning and identification that integrates ultra-wideband (UWB) and inertial measurement unit (IMU) and an algorithm for operation process detection based on surveillance video.

[0055] Please refer to Figure 1 As shown, a method for constructing a power grid operation trajectory file includes the following steps:

[0056] S1. Based on ultra-bandwidth and inertial measurement, calculate the location of the workers and identify their identities;

[0057] To improve positioning accuracy, this embodiment uses ultra-wideband (UWB) and inertial measurement (IMU) data fusion to calculate the location of the workers. Specifically, a UWB base station is used as the far point of the coordinate system, and one UWB base station is set up in each of the X, Y, and Z directions of the origin to cover the entire work area, thereby obtaining three-dimensional distance information for subsequent coordinate positioning calculations. During high-frequency IMU data processing, forward recursion is performed to obtain short-term personnel position changes. The worker positioning result is obtained by fusing the UWB global position with the IMU high-frequency information. The worker positioning calculation based on ultra-wideband and inertial measurement includes the following steps:

[0058] The coordinate information of the workers in the three-dimensional coordinate system is obtained through ultra-broadband base stations;

[0059] Solve the position equation based on the distance measurement equation, satisfying:

[0060]

[0061] Among them, P x P y and P z Let x represent the coordinates of the position at time k in the X, Y, and Z directions, respectively. i y i and z i Let d represent the coordinates of the i-th ultra-wideband base station relative to the origin. i Indicates the distance between equipment and personnel;

[0062] Further optimize and correct the recursive results of the IMU data, eliminate the cumulative error of IMU integration and obtain a more accurate positioning result. Based on the inertial measurement data, perform state prediction to satisfy:

[0063]

[0064] in, It is the predicted system state (pose and velocity) at time k+1 for the i-th Sigma sampling point, F is the state transition matrix, B is the input control matrix, and u k|k It is the control input (IMU input); the Sigma sampling points mentioned above refer to points randomly sampled in the algorithm, which is a term in unscented Kalman filtering;

[0065] The mean of the system state is obtained by weighting multiple Sigma sampling points. The covariance matrix P k+1|k According to the mean The covariance matrix P k+1|k The Sigma sampling points are updated using the UT transform. Bin calculates the observed values ​​according to the observation equation, which satisfies:

[0066]

[0067] Among them, The observation value refers to the i-th Sigma sampling point, and H is the observation matrix;

[0068] The system observation mean was obtained by weighting the Sigma sampling points again. and the observed covariance matrix P zk, The Kalman gain K is obtained, which satisfies:

[0069]

[0070] Among them, P zk, Let P be the autocovariance matrix of the observation vector. xk, Represents the cross-covariance matrix between the observation vector and the state vector;

[0071] Finally, the updated values ​​of the system state and covariance matrix at time k+1 are obtained. The system state and covariance matrix are then updated to obtain the personnel positioning result, satisfying the following:

[0072]

[0073]

[0074] in, The system mean is obtained by weighting the Sigma sampling points in the predicted stage. The observed mean is calculated by weighting the Sigma sampling points.

[0075] The operator's location result, i.e., system status X k+1|k+1 An estimate was obtained.

[0076] Each UWB device has a unique ID identification code. This code, along with UWB signals and IMU data, can be packaged and sent to a multi-dimensional trajectory file processing server. The identification code is then matched against employee ID numbers in a database to identify the operator. The operator identification process includes: obtaining the UWB device ID identification code, matching the code against employee ID numbers in the database, and obtaining the operator identification result.

[0077] S2. Acquire on-site monitoring video data, and perform detection and analysis based on the video data. The detection includes safety measure targets, worker targets, and worker safety attire.

[0078] The specific calculation of S2 in this embodiment uses the DETR algorithm. DETR, short for Detection Transformer, was proposed in 2020 and is the first algorithm in the field of computer vision to utilize the Transformer structure for object detection. Compared with previous object detection algorithms, DETR avoids various complex manually designed mechanisms, solves the problem of a large amount of repetition in the prediction results, and does not require post-processing processes such as non-maximum suppression (NMS), realizing an end-to-end object detection process. Furthermore, relying on the advantages of the Transformer structure, DETR achieves detection accuracy exceeding that of many mainstream detection algorithms based on convolutional neural networks.

[0079] The specific method by which the DETR algorithm performs detection is as follows:

[0080] a. Convert the RGB image output by the video preprocessing unit to X img The input feature extraction network ResNet50 is used for feature extraction to obtain the corresponding feature map F. res The formula is: F res =ResNet50(X img );

[0081] b. Transfer the feature map F res Conv through 1×1 convolution 1×1 Dimensionality reduction is performed, then the data is reshaped into sequence data, and positional encoding information (Embedding) is added. pos The feature F with added location information is obtained. pos The formula is: F pos =Reshape(Conv 1×1 (Fres ))+Embedding pos ;

[0082] c. Feature F pos The encoded feature F is obtained by encoding using six repeated Transformer Encoders. encoder The formula is: F pos =Encoder ×6 (F pos )

[0083] d. Encode the feature F encoder With learnable target query features object The decoded feature F is obtained by decoding using six repeated Transformer Decoders. decoder The formula is: F decoder =Decoder ×6 (F encoder Query object );

[0084] e. Input the decoded features into a prediction head composed of multiple feedforward neural networks (FFNs) to perform target category prediction and bounding box prediction, using the formula: <(c1,b1),…,(c i ,b i ),…,9c n ,b n )>=FFNs(F decoder ), where c i For the predicted category of the i-th target, b n The detection bounding box for the predicted i-th target, including the center point coordinates, height, and width of the box, is denoted as: b i =(x i ,y i ,w i ,h i );

[0085] The training data for the DETR target detection model includes four types of targets: workers and safety signs, grounding wires, and safety railings. After training, the target detection unit can accurately and quickly detect these four types of targets.

[0086] QueryDet is an improved object detection algorithm based on RetinaNet, proposed in 2021. Compared with general object detection algorithms, it introduces a cascaded sparse query mechanism, which significantly improves the accuracy and efficiency of small object detection. During the operation, the safety clothing worn by the workers is mostly small in size and has a small pixel area in the monitoring image. Therefore, the clothing detection unit using the QueryDet algorithm can achieve more efficient and accurate detection of safety clothing targets.

[0087] The specific method for clothing detection using the QueryDet algorithm is as follows:

[0088] a. The target detection box of the operator output by the target detection unit (b) i The corresponding image X person From the original RGB image X img Crop the data from the middle and resize it to a uniform size using the formula: X person =Resize(Crop(X) img ,b i ));

[0089] b. X-ray the image of the operator person Feature maps P at seven resolution scales were obtained using the ResNet50 feature extraction network and the feature pyramid structure FPN. i＝1,2,3,4,5,6, =FPN(ResNet50(X person ));

[0090] c. Add a query header branch (Head) for coarse localization of small targets to the original RetinaNet prediction header. query Low-resolution feature map P l via Head query Generate a probability map of the existence of small targets, and generate a key location map (Key) of the possible existence of small targets by using a threshold (Threshold). l Positions with a probability greater than the threshold δ are set to 1, and all others are set to 0. The formula is: Key l =Threshold δ (Head query (P l ));

[0091] d. Map the key locations of small targets. l Mapping to a higher-resolution scale involves taking the four nearest neighbors of a location marked as 1 in the low-resolution location map as key locations at that resolution, and generating a higher-resolution key location lookup map for the small target.

[0092] e. At high resolution scales, for feature map P l-1 By utilizing sparse convolution SpConv, the graph is queried only at key locations of small targets. The calculation is performed at positions with a value of 1, using the predicted head. pred The target detection result is output using the following formula: Where c i For the predicted category of the i-th target, b n The detection bounding box for the predicted i-th target, including the center point coordinates, height, and width of the box, is denoted as: b i =(x i ,y i ,w i ,h i Simultaneously, a key location map (Key) is generated at this resolution. l-1 For small target detection at a higher resolution;

[0093] The training data for the QueryDet clothing detection model includes five categories of safety clothing targets: insulating gloves, insulating shoes, work clothes, goggles, and safety helmets. After training, the clothing detection unit can accurately and quickly detect these five types of targets.

[0094] In summary, S2 specifically includes the following steps:

[0095] The surveillance video data is preprocessed to obtain an RGB image sequence;

[0096] The input RGB image is used to detect safety measures and personnel targets at the work site using the DETR target detection algorithm, and the target category and corresponding detection box are output.

[0097] The RGB image is used to detect safe attire using the QueryDet small object detection algorithm, and the attire category and corresponding detection box are output.

[0098] S3. Based on the personnel location, the identity recognition, and the detection analysis, make judgments on work process control;

[0099] The work process in S3 includes a start shift meeting, site deployment, work operation, and end shift meeting. Before S3 is executed, a work process table needs to be generated. The work process table is generated based on the input work information such as the number of workers, work area, work duration, and work content.

[0100] Specifically, it includes:

[0101] The data for the initial class meeting were obtained by fitting a straight line using the least squares algorithm.

[0102] A start-up meeting will be conducted within the first 30 minutes of the operation. First, the number of on-site personnel N will be calculated based on the UWB signal recognition data. r and the number of workers N set. s Compare and confirm attendance. Next, determine the ground position (x, y) of all personnel except the person in charge. i ,y i ), i = 1, 2, ..., N r -1. Using the least squares algorithm to fit a straight line, the total error T is obtained. r .

[0103] T r Compared with the set threshold T1, when T is satisfied three times consecutively r <When the shift meeting starts at time T1, record the start time and save the camera image at that time as process control data;

[0104] T r Compared with the set threshold T2, when T is satisfied three times consecutively r > If T2 and the start of the shift meeting is true, then the end of the shift meeting is true. Record the end time and save the camera image at that moment as process control data.

[0105] The system detects pre-defined targets in the testing area to obtain site deployment data. It also checks for the presence of necessary safety measures such as grounding wires and safety fences, comparing this data with the work flow chart. If no necessary target is detected for 10 consecutive minutes, the system records the detection status and time, and saves an image of that moment as process control data.

[0106] The safety attire on site is inspected to obtain operational data. The prescribed safety attire includes insulated gloves, insulated shoes, work clothes, and a safety helmet, meeting the personnel safety protection needs of electrical work scenarios. Specifically, it confirms whether any attire is missing. When a missing attire is detected, such as not wearing a safety helmet, the inspection time, inspection details, and camera images at that moment are recorded as process control data.

[0107] The data for the final station meeting was obtained by fitting a straight line using the least squares algorithm; please refer to the instructions for the initial station meeting data for this process.

[0108] S4. Combine the personnel location, identity recognition, detection analysis, and process control judgment results according to the time sequence to form a multi-dimensional trajectory file.

[0109] The timestamps mentioned above are 19-bit timestamps directly obtained from the trajectory archive construction system. Personnel positioning also includes absolute position, which can be obtained by acquiring data from the GPS device at the work site; the two-dimensional coordinate position is obtained from S1.

[0110] Example 2

[0111] Example 2 is an explanation and description of each device in the on-site system.

[0112] Please refer to Figure 2 The system diagram shown illustrates a power grid operation trajectory file construction system, which includes: a GPS device, a camera, an ultra-wideband base station, and an edge computing device; the GPS is used for operator positioning, the camera is used for on-site monitoring video recording, the ultra-wideband base station is used for operator positioning calculation and operator identification, and the edge computing device is used to process data to obtain multi-dimensional trajectory files. When the power grid operation trajectory file construction system is running, it executes the power grid operation trajectory file construction method described in Embodiment 1.

[0113] For example, deploy one GPS device, two high-definition cameras, multiple Ultra Wideband (UWB) base stations, and one edge computing device at the work site.

[0114] Example 3

[0115] Figure 3 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention, as shown below. Figure 3 As shown, the electronic device includes a processor 210, a memory 220, an input device 230, and an output device 240; the number of processors 210 in the computer device can be one or more. Figure 3 Taking a processor 210 as an example; the processor 210, memory 220, input device 230, and output device 240 in the electronic device can be connected via a bus or other means. Figure 3 Taking the example of a connection between China and Israel via a bus.

[0116] The memory 220, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules. The processor 210 executes various functional applications and data processing of the electronic device by running the software programs, instructions, and modules stored in the memory 220, thereby implementing the power grid operation trajectory file construction method of Embodiment 1 above.

[0117] The memory 220 may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on terminal usage. Furthermore, the memory 220 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory, or other non-volatile solid-state storage device. In some instances, the memory 220 may further include memory remotely located relative to the processor 210, which can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0118] Input device 230 can be used to receive input user identity information, video data, and location data, etc. Output device 240 may include display devices such as a display screen.

[0119] Example 4

[0120] Embodiment 4 of the present invention also provides a storage medium containing computer-executable instructions, which can be used by a computer to execute a method for constructing a power grid operation trajectory file, the method comprising:

[0121] Based on ultra-bandwidth and inertial measurement calculations, the location of workers is determined and their identities are identified.

[0122] Acquire on-site monitoring video data, and perform detection and analysis based on the video data. The detection includes safety measure targets, worker targets, and worker safety attire.

[0123] Based on the personnel location, identity recognition, and detection analysis, the operation process control judgment is made;

[0124] The personnel location, identity recognition, detection analysis, and process control judgment results are combined in chronological order to form a multi-dimensional trajectory file.

[0125] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause an electronic device (which may be a mobile phone, personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0126] It is worth noting that in the embodiments of the above-mentioned power grid operation trajectory archive construction method and device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0127] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this invention.

Claims

1. A method of building a power grid operation trajectory archive, characterized by, The method comprises the following steps: calculating the position of the worker based on ultra-wideband and inertial measurement, and identifying the identity of the worker; acquiring on-site monitoring video data, and performing detection analysis based on the video data, wherein the detection includes safety measure targets, worker targets, and the safety dress of the worker; acquiring on-site monitoring video data, and performing detection analysis based on the video data, wherein the detection includes safety measure targets, worker targets, and the safety dress of the worker, and the method comprises the following steps: performing video sequence preprocessing on the monitoring video data to obtain an RGB image sequence; inputting the RGB image, detecting the safety measure targets and worker targets in the work site by using a DETR target detection algorithm, and outputting the target category and the corresponding detection frame; detecting the safety dress of the worker by using a QueryDet small target detection algorithm, and outputting the dress category and the corresponding detection frame; the detection by using the DETR algorithm comprises the following steps: a. the RGB image X output by the video pre-processing unit img The input feature extraction network ResNet50 is used for feature extraction, and a corresponding feature map F is obtained res ; b. Feature map F res By 1x1 convolution Conv 1×1 Dimensionality reduction, convert to sequence data, and add position encoding information Embedding pos , get feature F pos with position information c. the feature F pos is encoded by 6 repeated Transformer Encoders to obtain the encoded feature F encoder ; d. The encoded feature F encoder with the learnable target query feature Query object The decoded feature F is obtained by decoding through 6 repeated Transformer Decoders decoder ; e. inputting the decoded features into a prediction head composed of multiple feedforward neural networks (FFNs) to perform target category prediction and boundary frame prediction; the dress detection by using the QueryDet algorithm comprises the following steps: a. The job worker target detection frame output by the target detection unit b i The corresponding image X person The original RGB image X img is cropped out and scaled to a uniform size; b. the work personnel image X person Through the feature extraction network ResNet50 and the feature pyramid structure FPN, feature maps P of 7 resolution scales are obtained i＝1,2,3,4,5,6,7 = FPN(ResNet50(X person )) c. Add a query head branch Head for rough positioning of small targets based on the original prediction head of RetinaNet query , low-resolution feature map P l Generate a small target existence probability map through Head query Generate a key position map where small targets may exist through threshold Threshold l The position value is 1 if the probability is greater than threshold δ, and 0 otherwise. d. mapping the small target key position map Key l to a higher resolution scale, specifically, taking the 4 nearest neighbors of the position corresponding to 1 in the low resolution position map as the key positions at the higher resolution, to generate a small target key position query map at the higher resolution e.At a high resolution scale, for the feature map P l-1 , using sparse convolution SpConv, only the positions with value 1 in the small target key position query map are calculated, and the target detection result is output through the prediction head Head pred ; performing work flow control judgment based on the worker positioning, the identity identification, and the detection analysis; combining the worker positioning, the identity identification, the detection analysis, and the work flow control judgment result according to the time sequence to form a multi-dimensional trajectory archive.

2. The power grid work trajectory profile building method of claim 1, wherein, calculating the position of the worker based on ultra-wideband and inertial measurement, and identifying the identity of the worker, comprises the following steps: acquiring the coordinate information of the worker in a three-dimensional coordinate system by using an ultra-wideband base station; solving a position equation according to a ranging equation, which satisfies: where P x , P y , and P z represent coordinate values in X, Y, and Z directions at time k, x i , y i , and z i represent coordinates of the i-th super-bandwidth base station relative to the origin, and d i represents the distance between the device and the person. performing state prediction according to the inertial measurement data, which satisfies: wherein, is the system state prediction of the i-th Sigma sample point at time k + 1, F is the state transition matrix, B is the input control matrix, u k|k is the control input; calculating an observation value according to an observation equation, which satisfies: wherein, wherein wherein, wherein is the observation value of the i-th Sigma sample point, and H is the observation matrix. calculating a Kalman gain, which satisfies: where P zk,zk represents the auto-covariance matrix of the observation vector, P xk,zk represents the cross-covariance matrix of the observation vector and the state vector; updating the system state and the covariance matrix to obtain the worker positioning result, which satisfies: wherein, is the system mean value predicted from the Sigma sample point weighted phase, is the observation mean value calculated from the Sigma sample point weighted observation.

3. The power grid work trajectory profile building method of claim 1, wherein, the worker identity identification comprises: acquiring the device ID identification code of the ultra-wideband, matching the identification code with the worker ID in the database, and obtaining the worker identity identification result.

4. The power grid work trajectory profile building method of claim 1, wherein, the work flow comprises a start shift meeting, site deployment, work operation, and an end shift meeting.

5. The power grid operations trajectory archive building method of claim 4, wherein, the work flow control judgment comprises: performing straight line fitting by using a least square algorithm to obtain start shift meeting data; detecting preset targets in the site to obtain site deployment data; detecting the safety dress in the site to obtain work operation data; performing straight line fitting by using a least square algorithm to obtain end shift meeting data.

6. The power grid work trajectory profile building method of claim 1, wherein, the time is a 19-bit time stamp directly acquired.

7. A power grid operation trajectory profile building system, characterized by, It comprises: a GPS device, a camera, an ultra-wideband base station, and an edge computing device; the GPS is used for worker positioning, the camera is used for on-site monitoring video recording, the ultra-wideband base station is used for worker positioning calculation and worker identity identification, and the edge computing device is used for processing data to obtain a multi-dimensional trajectory archive; when the power grid work trajectory archive construction system is running, the power grid work trajectory archive construction method in any one of claims 1-6 is executed.

8. An electronic device comprising a processor, a storage medium, and a computer program stored in the storage medium, characterized in that, The computer program, when executed by a processor, implements the power grid operation trajectory archive construction method according to any one of claims 1 to 6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by a processor, implements the power grid operation trajectory archive construction method according to any one of claims 1 to 6.

Images