CN112766129A - Power grid worker behavior real-time monitoring method and device - Google Patents
Power grid worker behavior real-time monitoring method and device Download PDFInfo
- Publication number
- CN112766129A CN112766129A CN202110036945.9A CN202110036945A CN112766129A CN 112766129 A CN112766129 A CN 112766129A CN 202110036945 A CN202110036945 A CN 202110036945A CN 112766129 A CN112766129 A CN 112766129A
- Authority
- CN
- China
- Prior art keywords
- human body
- key point
- body part
- coordinates
- power grid
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 129
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000002372 labelling Methods 0.000 claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 10
- 238000013507 mapping Methods 0.000 claims description 15
- 238000010586 diagram Methods 0.000 claims description 7
- 238000012806 monitoring device Methods 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 abstract description 5
- 238000000605 extraction Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013139 quantization Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Alarm Systems (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a method and a device for monitoring the behavior of a power grid worker in real time, wherein in the method, a power grid real-time monitoring video is input to a human body posture 3D model which is established in advance; the training set used by the human body posture 3D model comprises a historical monitoring image data set, wherein the historical monitoring image data set comprises a historical monitoring image and a plurality of human body part labeling information on the historical monitoring image; acquiring human body part key point information output by the human body posture 3D model, wherein the human body part key point information comprises head key point information and foot key point information; judging whether the distance between the head of the worker and the power grid equipment is smaller than a preset distance or not; judging whether the worker has a falling risk, namely judging whether the distance between the head and the feet of the worker in the vertical direction is smaller than a preset distance; this application can whether the real-time supervision electric wire netting staff has the abnormal conditions and send out the police dispatch newspaper, the error that appears when avoiding artifical monitoring.
Description
Technical Field
The application mainly relates to the power system technology, in particular to a method and a device for monitoring the behavior of a power grid worker in real time.
Background
Some power systems are limited by production environments and site locations and are arranged in remote places, so that the safety of workers in a power grid is very important. In the working process of the power grid workers, abnormal conditions such as the safety helmet is not worn, the accident falls, the illegal gathering, the too close distance from the power grid equipment and the like can occur, and potential safety hazards exist, so that the workers need to be monitored, the safety of the workers is protected, and the stable operation of the power grid is maintained.
At present, when a power grid worker is detected, cameras are installed at all positions of a working environment, and monitoring personnel judge whether abnormal conditions exist in the worker or not through monitoring videos; however, this manual monitoring method is prone to misjudgment, resulting in irreparable loss.
Disclosure of Invention
In order to solve the problem that manual detection is easy to make a judgment error when monitoring power grid workers, the application discloses a real-time monitoring method and a real-time monitoring device for the behavior of the power grid workers through the following embodiments.
The application discloses in a first aspect a real-time monitoring method for the behavior of a power grid worker, comprising the following steps:
acquiring a real-time monitoring video of a power grid;
inputting the power grid real-time monitoring video into a pre-established human body posture 3D model; the training set used by the human body posture 3D model comprises a historical monitoring image data set, wherein the historical monitoring image data set comprises a historical monitoring image and a plurality of human body part labeling information on the historical monitoring image;
acquiring human body part key point information output by the human body posture 3D model, wherein the human body part key point information comprises head key point information and foot key point information;
judging whether the distance between the head of the worker and the power grid equipment is smaller than a preset distance or not according to the head key point information, and if so, giving an alarm;
and judging whether the distance between the head and the feet of the worker in the vertical direction is smaller than a preset distance or not according to the head key point information and the foot key point information, if so, judging that the worker has a falling risk, and giving an alarm at the moment, wherein the vertical direction is the direction vertical to the ground.
Optionally, the method further comprises pre-establishing a historical monitoring image dataset by:
classifying a plurality of human body part marking information on the historical monitoring image according to different human body parts, wherein the classification result comprises head marking information and foot marking information;
and corresponding the coordinates of the plurality of human body part labeling information with the coordinates of the 3D basic model, and establishing the corresponding relation between the human body part labeling and the 3D basic model.
Optionally, the method further includes pre-building a human body posture 3D model by:
acquiring the historical monitoring image data set;
extracting the characteristics of the historical monitoring images in the historical monitoring image data set through a characteristic pyramid network to obtain a characteristic diagram;
carrying out dimension unification on the feature graphs;
carrying out classification regression on the feature map with unified dimension to obtain all pixel coordinates belonging to human body parts;
acquiring the coordinates of key points of the human body part according to the classification result based on the pixel coordinates of all the human body parts; the human body part key point coordinates comprise coordinates of a head key point and coordinates of a foot key point;
converting the coordinates of the key points of the human body parts into space coordinates, and mapping the key points of the human body parts on a 3D basic model according to the corresponding relation between the pre-established human body part labels and the 3D basic model;
and mapping key points of human body parts of all historical monitoring images in the historical monitoring image data set on the 3D basic model to obtain a human body posture 3D model.
Optionally, the classifying and regressing the feature map with unified dimensions to obtain all pixel coordinates belonging to the human body part includes:
judging whether a certain pixel in the feature map with unified dimension is from a background or a human body part;
for a pixel from a human body part, pixel coordinates of the pixel in the human body part are acquired.
The second aspect of the present application discloses a real-time monitoring device for behavior of power grid staff, the device is used for executing the real-time monitoring method for behavior of power grid staff disclosed in the first aspect of the present application, the device includes:
the video acquisition module is used for acquiring a real-time monitoring video of the power grid;
the video input module is used for inputting the real-time monitoring video of the power grid to a pre-established human body posture 3D model; the training set used by the human body posture 3D model comprises a historical monitoring image data set, wherein the historical monitoring image data set comprises a historical monitoring image and a plurality of human body part labeling information on the historical monitoring image;
the key point information acquisition module is used for acquiring key point information of a human body part output by the human body posture 3D model, and the key point information of the human body part comprises head key point information and foot key point information;
the first distance judgment module is used for judging whether the distance between the head of a worker and the power grid equipment is smaller than a preset distance or not according to the head key point information, and if so, giving an alarm;
and the second distance judgment module is used for judging whether the distance between the head and the foot of the worker in the vertical direction is smaller than a preset distance according to the head key point information and the foot key point information, if so, judging that the worker has a falling risk, and sending an alarm at the moment, wherein the vertical direction is vertical to the horizontal direction of the ground.
Optionally, the apparatus further includes a data set creating module for creating a historical monitoring image data set in advance, where the data set creating module includes:
the labeling information classifying unit is used for classifying a plurality of human body part labeling information on the historical monitoring image according to different human body parts, and the classification result comprises head labeling information and foot labeling information;
and the corresponding relation establishing unit is used for corresponding the coordinates of the plurality of human body part labeling information with the coordinates of the 3D basic model and establishing the corresponding relation between the human body part labeling and the 3D basic model.
Optionally, the apparatus further includes a model building module for building a human body posture 3D model in advance, where the model building module includes:
a historical monitoring image data set acquisition unit for acquiring the historical monitoring image data set;
the characteristic map acquisition unit is used for extracting the characteristics of the historical monitoring images in the historical monitoring image data set through a characteristic pyramid network to acquire a characteristic map;
the image dimension unifying unit is used for unifying the dimensions of the feature maps;
the pixel coordinate acquisition unit is used for carrying out classification regression on the feature map with unified dimensionality and acquiring pixel coordinates of all parts belonging to the human body;
a key point coordinate obtaining unit, configured to obtain, based on the pixel coordinates of all the human body parts, key point coordinates of the human body parts according to the classification result; the human body part key point coordinates comprise coordinates of a head key point and coordinates of a foot key point;
the coordinate conversion mapping unit is used for converting the human body part key point coordinates into space coordinates, and mapping the human body part key points on the 3D basic model according to the corresponding relation between the pre-established human body part label and the 3D basic model;
and the model acquisition unit is used for acquiring a human body posture 3D model by mapping the human body part key points of all historical monitoring images in the historical monitoring image data set on the 3D basic model.
Optionally, the pixel coordinate obtaining unit includes:
the characteristic image judging subunit is used for judging whether a certain pixel in the characteristic image with unified dimensionality is from a background or a human body part;
and the pixel coordinate acquisition subunit is used for acquiring the pixel coordinate of the pixel in the human body part from the pixel of the human body part.
The application discloses a method and a device for monitoring the behavior of a power grid worker in real time, wherein in the method, a power grid real-time monitoring video is input to a human body posture 3D model which is established in advance; the training set used by the human body posture 3D model comprises a historical monitoring image data set, wherein the historical monitoring image data set comprises a historical monitoring image and a plurality of human body part labeling information on the historical monitoring image; acquiring human body part key point information output by the human body posture 3D model, wherein the human body part key point information comprises head key point information and foot key point information; judging whether the distance between the head of the worker and the power grid equipment is smaller than a preset distance or not according to the head key point information, and if so, giving an alarm; and judging whether the distance between the head and the feet of the worker in the vertical direction is smaller than a preset distance or not according to the head key point information and the foot key point information, if so, judging that the worker has a falling risk, and giving an alarm at the moment, wherein the vertical direction is the direction vertical to the ground.
This application can whether the real-time supervision electric wire netting staff has the abnormal conditions and send out the police dispatch newspaper, the error that appears when avoiding artifical monitoring.
Drawings
In order to more clearly explain the technical solution of the present application, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious to those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic workflow diagram of a method for monitoring behavior of a power grid worker in real time according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of a device for monitoring behavior of a power grid worker in real time according to an embodiment of the present application.
Detailed Description
In order to solve the problem that manual detection is easy to make a judgment error when monitoring power grid workers, the application discloses a real-time monitoring method and a real-time monitoring device for the behavior of the power grid workers through the following embodiments.
The first embodiment of the present application discloses a method for monitoring behavior of a power grid worker in real time, referring to a work flow diagram shown in fig. 1, the method includes:
and S101, acquiring a real-time monitoring video of the power grid.
And S102, inputting the real-time monitoring video of the power grid to a pre-established human body posture 3D model. The training set used by the human body posture 3D model comprises a historical monitoring image data set, wherein the historical monitoring image data set comprises a historical monitoring image and a plurality of human body part labeling information on the historical monitoring image.
Step S103, obtaining the human body part key point information output by the human body posture 3D model, wherein the human body part key point information comprises head key point information and foot key point information.
And step S104, judging whether the distance between the head of the worker and the power grid equipment is smaller than a preset distance or not according to the head key point information, and if so, giving an alarm.
And S105, judging whether the distance between the head and the feet of the worker in the vertical direction is smaller than a preset distance or not according to the head key point information and the foot key point information, if so, judging that the worker has a falling risk, and sending an alarm at the moment, wherein the vertical direction is the direction vertical to the ground.
The method comprises the steps that whether a worker has a falling risk or not can be judged by determining a rectangular frame of a human body through a frame output by a human body posture 3D model, the rectangular frame is vertical when the worker normally works, when the vertical distance between the rectangular frame and the ground is smaller than a preset distance, the worker is judged to have the falling risk, and an alarm is given at the moment.
The method can also judge whether the worker wears the safety helmet or not, the head area of the worker is displayed according to the head key point information, and the monitoring worker judges whether the worker wears the safety helmet or not.
The method is also used for judging whether the workers have the conditions of illegal aggregation and fighting, acquiring the head key point information of a plurality of workers, and monitoring and giving an alarm when the distance between the heads of any two of five or more workers is smaller than the preset distance.
The application discloses a method and a device for monitoring the behavior of a power grid worker in real time, wherein in the method, a power grid real-time monitoring video is input to a human body posture 3D model which is established in advance; the training set used by the human body posture 3D model comprises a historical monitoring image data set, wherein the historical monitoring image data set comprises a historical monitoring image and a plurality of human body part labeling information on the historical monitoring image; acquiring human body part key point information output by the human body posture 3D model, wherein the human body part key point information comprises head key point information and foot key point information; judging whether the distance between the head of the worker and the power grid equipment is smaller than a preset distance or not according to the head key point information, and if so, giving an alarm; and judging whether the distance between the head and the feet of the worker in the vertical direction is smaller than a preset distance or not according to the head key point information and the foot key point information, if so, judging that the worker has a falling risk, and giving an alarm at the moment, wherein the vertical direction is the direction vertical to the ground.
This application can whether the real-time supervision electric wire netting staff has the abnormal conditions and send out the police dispatch newspaper, the error that appears when avoiding artifical monitoring.
Further, the method further comprises pre-establishing a historical monitoring image dataset by:
and classifying the plurality of human body part marking information on the historical monitoring image according to different human body parts, wherein the classification result comprises head marking information and foot marking information.
And corresponding the coordinates of the plurality of human body part labeling information with the coordinates of the 3D basic model, and establishing the corresponding relation between the human body part labeling and the 3D basic model.
Further, the method also comprises the following steps of establishing a human body posture 3D model in advance:
and acquiring the historical monitoring image data set. The adopted data set also comprises a Dense Pose-COCO data set established by researchers, and comprises 5 ten thousand human images, and more than 500 ten thousand corresponding relations are constructed in the images.
And extracting the characteristics of the historical monitoring images in the historical monitoring image data set through a characteristic pyramid network to obtain a characteristic diagram. Performing feature extraction through the improved feature pyramid network, wherein the feature extraction comprises the following steps: extracting features of images in the data set from the bottom through a main network, namely a residual network, obtaining feature maps with different resolutions and channel numbers, respectively representing output features of conv2, conv3, conv4 and conv5 as C1, C2, C3 and C4, and respectively representing step sizes of the input images as widths of 4, 8, 16 and 32 pixel values; taking four-layer network features { C4, C3, C2 and C1} in a top-down sequence, performing 2 times of upsampling, compressing the number of channels to 256 dimensions, and overlapping the last three-layer features obtained by upsampling with features obtained by performing 1 × 1 convolution dimensionality reduction processing on the network features { C1, C2 and C3} to obtain features with strong semantic features and detailed position information; and respectively carrying out parallel residual error layer processing on the obtained network characteristics to obtain the output characteristics of the improved characteristic pyramid network. The parallel residual error layer is an important detector of the network, and the structure of the parallel residual error layer mainly comprises two branches. The first branch consists of a bottleneck module and consists of three convolution layers, a normalization layer and an activation layer: the first 1 x 1 convolution kernel is used for reducing the dimension of the characteristic dimension, the second 3 x 3 convolution kernel is used for down-sampling the characteristic, and the third 1 x 1 convolution kernel is used for increasing the dimension of the characteristic dimension; the second branch consists of three convolution layers, a normalization layer, an activation layer and an upper sampling layer and a lower sampling layer; and finally, carrying out residual error connection on the results output by the two parallel branches to obtain new characteristics. And residual connecting sign behavior. And inputting the characteristic vector [ d, wc, hc ] as the characteristic of the c-th layer of the original characteristic pyramid, and outputting the characteristic as a new characteristic through PRL. The PRL ensures that the feature maps output by different convolution operations have the same size so as to facilitate feature splicing; after the original features are subjected to 1 × 1 convolution kernel dimensionality reduction, the number of channels is reduced, and after 3 × 3 convolution operation, the network effectively trains data and extracts features; the original dimensions of the features are restored by a 1 x 1 convolution kernel, and the final two-fold upsampling increases the feature resolution by two-fold. The original features are subjected to a PRL serial structure to obtain network features output by the improved feature pyramid, and the extraction performance of the geometric and textural features of key point parts is improved.
And carrying out dimension unification on the feature graph. And the dimensions of the output network feature maps with different sizes are unified through the ROI Align pooling layer, and a feature map with a fixed size is generated for each ROI, so that subsequent operation is facilitated. The ROI Align pooling layer uses a bilinear interpolation method to replace the prior quantization operation, effectively eliminates quantization errors, avoids the occurrence of a mismatch problem, and plays an important role in the prediction of a subsequent human mask.
And performing classification regression on the feature map with unified dimensionality to obtain all pixel coordinates belonging to the human body part. And (3) sending the feature map output by the ROI Align into a cross cascade framework, and performing mutual fusion training through a Dense Pose and two auxiliary networks (mask and keypoint).
And acquiring the coordinates of key points of the human body part according to the classification result based on the pixel coordinates of all the human body parts. The human body part key point coordinates comprise coordinates of a head key point and coordinates of a foot key point.
And converting the coordinates of the key points of the human body parts into space coordinates, and mapping the key points of the human body parts on the 3D basic model according to the corresponding relation between the pre-established human body part labels and the 3D basic model.
And mapping key points of human body parts of all historical monitoring images in the historical monitoring image data set on the 3D basic model to obtain a human body posture 3D model.
Further, the classifying and regressing the feature map with unified dimensions to obtain all pixel coordinates belonging to the human body part includes:
and judging whether a certain pixel in the feature map with unified dimension is from the background or a human body part.
For a pixel from a human body part, pixel coordinates of the pixel in the human body part are acquired.
The following are embodiments of the apparatus disclosed herein for performing the above-described method embodiments. For details which are not disclosed in the device embodiments, reference is made to the method embodiments of the present application.
The second embodiment of the present application discloses a device for monitoring behavior of power grid workers in real time, referring to the schematic structural diagram of fig. 2, the device is used for executing a method for monitoring behavior of power grid workers in real time, which is disclosed in the first embodiment of the present application, and the device includes:
and the video acquisition module 10 is used for acquiring a real-time monitoring video of the power grid.
And the video input module 20 is used for inputting the real-time monitoring video of the power grid to a pre-established human body posture 3D model. The training set used by the human body posture 3D model comprises a historical monitoring image data set, wherein the historical monitoring image data set comprises a historical monitoring image and a plurality of human body part labeling information on the historical monitoring image.
And the key point information obtaining module 30 is configured to obtain key point information of a human body part output by the human body posture 3D model, where the key point information of the human body part includes key point information of a head and key point information of a foot.
And the first distance judgment module 40 is used for judging whether the distance between the head of the worker and the power grid equipment is smaller than a preset distance or not according to the head key point information, and if so, giving an alarm.
And the second distance judgment module 50 is configured to judge whether the distance between the head and the foot of the worker in the vertical direction is smaller than a preset distance according to the head key point information and the foot key point information, and if the distance is smaller than the preset distance, judge that the worker has a risk of falling to the ground, and send an alarm at the moment, where the vertical direction is perpendicular to the horizontal direction of the ground.
Further, the apparatus further comprises a data set creating module for creating a historical monitoring image data set in advance, the data set creating module comprising:
and the labeling information classification unit is used for classifying the labeling information of the plurality of human body parts on the historical monitoring image according to different human body parts, and the classification result comprises head labeling information and foot labeling information.
And the corresponding relation establishing unit is used for corresponding the coordinates of the plurality of human body part labeling information with the coordinates of the 3D basic model and establishing the corresponding relation between the human body part labeling and the 3D basic model.
Further, the device also comprises a model building module for building a human body posture 3D model in advance, wherein the model building module comprises:
a historical monitoring image data set obtaining unit, configured to obtain the historical monitoring image data set.
And the characteristic map acquisition unit is used for extracting the characteristics of the historical monitoring images in the historical monitoring image data set through a characteristic pyramid network to acquire a characteristic map.
And the image dimension unifying unit is used for unifying the dimensions of the feature maps.
And the pixel coordinate acquisition unit is used for carrying out classification regression on the feature map with unified dimensionality and acquiring all pixel coordinates belonging to human body parts.
And the key point coordinate acquisition unit is used for acquiring the key point coordinates of the human body part according to the classification result based on the pixel coordinates of all the human body parts. The human body part key point coordinates comprise coordinates of a head key point and coordinates of a foot key point.
And the coordinate conversion mapping unit is used for converting the human body part key point coordinates into space coordinates, and mapping the human body part key points on the 3D basic model according to the corresponding relation between the pre-established human body part label and the 3D basic model.
And the model acquisition unit is used for acquiring a human body posture 3D model by mapping the human body part key points of all historical monitoring images in the historical monitoring image data set on the 3D basic model.
Further, the pixel coordinate acquiring unit includes:
and the characteristic image judging subunit is used for judging whether a certain pixel in the characteristic image with unified dimensions is from the background or a human body part.
And the pixel coordinate acquisition subunit is used for acquiring the pixel coordinate of the pixel in the human body part from the pixel of the human body part.
The present application has been described in detail with reference to the specific embodiments and examples, but these descriptions should not be construed as limiting the present application. Those skilled in the art will appreciate that various equivalent substitutions, modifications or improvements may be made to the presently disclosed embodiments and implementations thereof without departing from the spirit and scope of the present disclosure, and these fall within the scope of the present disclosure. The protection scope of this application is subject to the appended claims.
Claims (8)
1. A real-time monitoring method for the behavior of power grid workers is characterized by comprising the following steps:
acquiring a real-time monitoring video of a power grid;
inputting the power grid real-time monitoring video into a pre-established human body posture 3D model; the training set used by the human body posture 3D model comprises a historical monitoring image data set, wherein the historical monitoring image data set comprises a historical monitoring image and a plurality of human body part labeling information on the historical monitoring image;
acquiring human body part key point information output by the human body posture 3D model, wherein the human body part key point information comprises head key point information and foot key point information;
judging whether the distance between the head of the worker and the power grid equipment is smaller than a preset distance or not according to the head key point information, and if so, giving an alarm;
and judging whether the distance between the head and the feet of the worker in the vertical direction is smaller than a preset distance or not according to the head key point information and the foot key point information, if so, judging that the worker has a falling risk, and giving an alarm at the moment, wherein the vertical direction is the direction vertical to the ground.
2. The method for monitoring the behavior of the staff in the power grid in real time as claimed in claim 1, wherein the method further comprises pre-establishing a historical monitoring image data set by the steps of:
classifying a plurality of human body part marking information on the historical monitoring image according to different human body parts, wherein the classification result comprises head marking information and foot marking information;
and corresponding the coordinates of the plurality of human body part labeling information with the coordinates of the 3D basic model, and establishing the corresponding relation between the human body part labeling and the 3D basic model.
3. The method for monitoring the behavior of the power grid staff in real time according to claim 2, wherein the method further comprises the step of pre-establishing a human posture 3D model through the following steps:
acquiring the historical monitoring image data set;
extracting the characteristics of the historical monitoring images in the historical monitoring image data set through a characteristic pyramid network to obtain a characteristic diagram;
carrying out dimension unification on the feature graphs;
carrying out classification regression on the feature map with unified dimension to obtain all pixel coordinates belonging to human body parts;
acquiring the coordinates of key points of the human body part according to the classification result based on the pixel coordinates of all the human body parts; the human body part key point coordinates comprise coordinates of a head key point and coordinates of a foot key point;
converting the coordinates of the key points of the human body parts into space coordinates, and mapping the key points of the human body parts on a 3D basic model according to the corresponding relation between the pre-established human body part labels and the 3D basic model;
and mapping key points of human body parts of all historical monitoring images in the historical monitoring image data set on the 3D basic model to obtain a human body posture 3D model.
4. The method for monitoring the behavior of the power grid staff in real time according to claim 3, wherein the step of performing classification regression on the feature map with unified dimensions to obtain all pixel coordinates belonging to human body parts comprises the following steps:
judging whether a certain pixel in the feature map with unified dimension is from a background or a human body part;
for a pixel from a human body part, pixel coordinates of the pixel in the human body part are acquired.
5. A real-time monitoring device for the behavior of a power grid worker, which is applied to the real-time monitoring method for the behavior of the power grid worker as claimed in any one of claims 1 to 4, the device comprising:
the video acquisition module is used for acquiring a real-time monitoring video of the power grid;
the video input module is used for inputting the real-time monitoring video of the power grid to a pre-established human body posture 3D model; the training set used by the human body posture 3D model comprises a historical monitoring image data set, wherein the historical monitoring image data set comprises a historical monitoring image and a plurality of human body part labeling information on the historical monitoring image;
the key point information acquisition module is used for acquiring key point information of a human body part output by the human body posture 3D model, and the key point information of the human body part comprises head key point information and foot key point information;
the first distance judgment module is used for judging whether the distance between the head of a worker and the power grid equipment is smaller than a preset distance or not according to the head key point information, and if so, giving an alarm;
and the second distance judgment module is used for judging whether the distance between the head and the foot of the worker in the vertical direction is smaller than a preset distance according to the head key point information and the foot key point information, if so, judging that the worker has a falling risk, and sending an alarm at the moment, wherein the vertical direction is vertical to the horizontal direction of the ground.
6. The device for monitoring the behavior of the staff in the power grid in real time as claimed in claim 5, wherein the device further comprises a data set establishing module for establishing a historical monitoring image data set in advance, and the data set establishing module comprises:
the labeling information classifying unit is used for classifying a plurality of human body part labeling information on the historical monitoring image according to different human body parts, and the classification result comprises head labeling information and foot labeling information;
and the corresponding relation establishing unit is used for corresponding the coordinates of the plurality of human body part labeling information with the coordinates of the 3D basic model and establishing the corresponding relation between the human body part labeling and the 3D basic model.
7. The device for monitoring the behavior of the power grid staff in real time as claimed in claim 6, wherein the device further comprises a model building module for building a human body posture 3D model in advance, the model building module comprises:
a historical monitoring image data set acquisition unit for acquiring the historical monitoring image data set;
the characteristic map acquisition unit is used for extracting the characteristics of the historical monitoring images in the historical monitoring image data set through a characteristic pyramid network to acquire a characteristic map;
the image dimension unifying unit is used for unifying the dimensions of the feature maps;
the pixel coordinate acquisition unit is used for carrying out classification regression on the feature map with unified dimensionality and acquiring pixel coordinates of all parts belonging to the human body;
a key point coordinate obtaining unit, configured to obtain, based on the pixel coordinates of all the human body parts, key point coordinates of the human body parts according to the classification result; the human body part key point coordinates comprise coordinates of a head key point and coordinates of a foot key point;
the coordinate conversion mapping unit is used for converting the human body part key point coordinates into space coordinates, and mapping the human body part key points on the 3D basic model according to the corresponding relation between the pre-established human body part label and the 3D basic model;
and the model acquisition unit is used for acquiring a human body posture 3D model by mapping the human body part key points of all historical monitoring images in the historical monitoring image data set on the 3D basic model.
8. The device for monitoring the behavior of the power grid staff in real time according to claim 7, wherein the pixel coordinate acquiring unit comprises:
the characteristic image judging subunit is used for judging whether a certain pixel in the characteristic image with unified dimensionality is from a background or a human body part;
and the pixel coordinate acquisition subunit is used for acquiring the pixel coordinate of the pixel in the human body part from the pixel of the human body part.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110036945.9A CN112766129A (en) | 2021-01-12 | 2021-01-12 | Power grid worker behavior real-time monitoring method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110036945.9A CN112766129A (en) | 2021-01-12 | 2021-01-12 | Power grid worker behavior real-time monitoring method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112766129A true CN112766129A (en) | 2021-05-07 |
Family
ID=75701644
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110036945.9A Pending CN112766129A (en) | 2021-01-12 | 2021-01-12 | Power grid worker behavior real-time monitoring method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112766129A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109919132A (en) * | 2019-03-22 | 2019-06-21 | 广东省智能制造研究所 | A kind of pedestrian's tumble recognition methods based on skeleton detection |
| CN110781765A (en) * | 2019-09-30 | 2020-02-11 | 腾讯科技（深圳）有限公司 | Human body posture recognition method, device, equipment and storage medium |
| CN111126321A (en) * | 2019-12-27 | 2020-05-08 | 广州市奔流电力科技有限公司 | Electric power safety construction protection method and device and computer equipment |
| CN111428617A (en) * | 2020-03-20 | 2020-07-17 | 广东电网有限责任公司 | Video image-based distribution network violation maintenance behavior identification method and system |
| CN111507308A (en) * | 2020-05-07 | 2020-08-07 | 广东电网有限责任公司 | Transformer substation safety monitoring system and method based on video identification technology |
| CN111640282A (en) * | 2020-05-29 | 2020-09-08 | 北京潞电电气设备有限公司 | Method, system and device for monitoring safety distance of personnel in power distribution room |
| CN111753712A (en) * | 2020-06-22 | 2020-10-09 | 中国电力科学研究院有限公司 | Method, system and equipment for monitoring safety of power production personnel |
-
2021
- 2021-01-12 CN CN202110036945.9A patent/CN112766129A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109919132A (en) * | 2019-03-22 | 2019-06-21 | 广东省智能制造研究所 | A kind of pedestrian's tumble recognition methods based on skeleton detection |
| CN110781765A (en) * | 2019-09-30 | 2020-02-11 | 腾讯科技（深圳）有限公司 | Human body posture recognition method, device, equipment and storage medium |
| CN111126321A (en) * | 2019-12-27 | 2020-05-08 | 广州市奔流电力科技有限公司 | Electric power safety construction protection method and device and computer equipment |
| CN111428617A (en) * | 2020-03-20 | 2020-07-17 | 广东电网有限责任公司 | Video image-based distribution network violation maintenance behavior identification method and system |
| CN111507308A (en) * | 2020-05-07 | 2020-08-07 | 广东电网有限责任公司 | Transformer substation safety monitoring system and method based on video identification technology |
| CN111640282A (en) * | 2020-05-29 | 2020-09-08 | 北京潞电电气设备有限公司 | Method, system and device for monitoring safety distance of personnel in power distribution room |
| CN111753712A (en) * | 2020-06-22 | 2020-10-09 | 中国电力科学研究院有限公司 | Method, system and equipment for monitoring safety of power production personnel |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP5267782B2 (en) | Security device, security system, security camera, security system operating method, security monitoring method, and program | |
| CN109460744B (en) | Video monitoring system based on deep learning | |
| CN103595974A (en) | Video geographic information system and method for urban areas | |
| CN110458794B (en) | Quality detection method and device for accessories of rail train | |
| CN110568310A (en) | Visual online monitoring system for power transmission line | |
| CN110852222A (en) | Campus corridor scene intelligent monitoring method based on target detection | |
| CN102572508A (en) | Video fault diagnosis system and method | |
| CN116883801A (en) | YOLOv8 target detection method based on attention mechanism and multi-scale feature fusion | |
| CN110458535A (en) | A kind of computer room O&M method based on AR glasses | |
| CN112766129A (en) | Power grid worker behavior real-time monitoring method and device | |
| CN102118605A (en) | Network video monitoring system and method | |
| CN114694073B (en) | Intelligent detection method, device, storage medium and equipment for wearing condition of safety belt | |
| CN116316233A (en) | Intelligent substation inspection system, method, equipment and storage medium | |
| CN113435433B (en) | Audio and video data extraction processing system based on operation site | |
| CN115482502A (en) | Abnormal behavior identification method, system and medium based on characteristic object and human body key point | |
| KR20230097854A (en) | Method and system for recognizing dangerous behavior of workers in power plant | |
| CN115147929A (en) | Construction scene monitoring method, device, equipment and storage medium | |
| CN112215211B (en) | Method for extracting chamber branch link topological relation based on CAD drawing data | |
| CN115471911A (en) | Power grid operation field violation action identification method and system | |
| CN113888407A (en) | Target detection system based on super-resolution technology | |
| CN112529881B (en) | Power control cabinet cable anomaly identification method and device | |
| WO2022173809A1 (en) | Computer vision systems and methods for detecting power line hazards from imagery | |
| CN114283442A (en) | Intelligent identification method and device for secondary wiring diagram and storage medium | |
| CN114007040A (en) | Oil-gas video safety monitoring method and device based on edge intelligent calculation | |
| CN115474441A (en) | Obstacle detection method, apparatus, device, and computer storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |