CN115908839A - Live working safety early warning method based on posture and wire recognition - Google Patents
Live working safety early warning method based on posture and wire recognition Download PDFInfo
- Publication number
- CN115908839A CN115908839A CN202210543595.XA CN202210543595A CN115908839A CN 115908839 A CN115908839 A CN 115908839A CN 202210543595 A CN202210543595 A CN 202210543595A CN 115908839 A CN115908839 A CN 115908839A
- Authority
- CN
- China
- Prior art keywords
- human body
- image
- early warning
- pixel
- live
- 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
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to a live working safety early warning method based on gesture and wire recognition, which comprises the steps of shooting pictures through a binocular structure optical camera, respectively processing the pictures, building a deep learning model of a convolutional neural network to recognize human body gestures, recognizing key joints of a human body, marking the key joints in a video picture, connecting the key joints in human body-shaped gestures to describe the positions of limbs and five sense organs of the human body, extracting and recognizing live wires by using an algorithm, obtaining two-dimensional coordinates of the human body gestures and key points of the wires, obtaining image depth, then building a three-dimensional coordinate model of the wires and the human body, then predicting the human body gestures in advance through the model, and calculating the distance between the human body and the wires to realize safety early warning. In the safety supervision of live-wire work, the labor cost is greatly reduced, the error rate of manually judging the distance is reduced, and the accuracy and the reliability of the safety early warning of the live-wire work are improved.
Description
Technical Field
The invention relates to a live working safety early warning method for an electric power system, in particular to a live working safety early warning method based on posture and wire recognition.
Background
To ensure the safe operation of the power grid, the construction and maintenance of the power facilities must be emphasized. When a power failure accident occurs, workers are required to arrive at the accident site immediately to carry out treatment and repair work. Electric power constructors work on the site of electric power construction and maintenance throughout the year, closely contact high-voltage lines, and once negligence is careless, electric shock accidents occur easily, and personal safety is seriously threatened.
In order to guarantee personal safety of electric power workers and reduce the incidence rate of electric shock accidents, the electric power industry combines practical experience and related research, establishes detailed safety management regulations and forcibly carries out safety education on the electric power workers. The minimum safe distance between a person and a live wire is also regulated when the person works on the live tower by the electric power safety working regulation. Although a series of specifications are established in combination with actual operation in the power industry, safety accidents can be reduced to a certain extent by the aid of the regulations, in the live-line overhaul operation process of the power distribution network, the situations of carelessness, misoperation and the like of operators still occur, and unnecessary accidents are caused.
Although the visual technology is also adopted for safety early warning in the live working process of the power system, the visual technology is not well combined with the actual working condition, and the following problems mainly exist:
(1) The commonly used early warning equipment in the market at present is an infrared distance measuring instrument, and the distance between a human body and a high-voltage wire is measured by utilizing infrared rays.
(2) In actual working conditions, the cylindrical surface of the electrified lead is slender and smooth, the identification of the reflected structured light can be influenced, the depth distance measuring precision based on the structured light is difficult to guarantee, and the applicability of the binocular camera-based triangular distance measuring method under the conditions of shielding and moving objects can be reduced. During live working, the movement of workers on the surface of the wire can also affect the distance measurement precision of the binocular camera.
(3) The two-dimensional mathematical expression of the wire in the actual working condition can not be well fitted through the existing image recognition electrified wire, and the recognition process of the electrified wire can be influenced by a mixed wire, a coincident wire and noise points. The charged conductor identification is applied to complex scenes such as split conductors, crossed conductors, insulators and the like, the effect is not ideal, troubles are caused to a common video identification conductor method, and the problems cannot be processed by a common identification algorithm.
(4) General distribution network operation belongs to high altitude construction, and the ground safety supervisory personnel who equip leads to the validity of safety supervision to obtain reliable assurance owing to factors such as visual angle is limited, and the distance is far away.
(5) In the live-line work process, the existing video manual supervision safety early warning system technology cannot guarantee the reliability and the real-time performance of judgment, only provides another visual angle for ground safety supervision personnel to judge the safety of live-line maintenance personnel, and still needs the safety supervision personnel to judge the safety distance between the personnel and the lead according to the experience of the safety supervision personnel.
Because the existing visual safety early warning technology has the problems, an automatic early warning device for live working needs to be designed, and the accident rate of the live working is technically reduced.
Disclosure of Invention
The invention aims to provide a live working safety early warning method based on gesture and wire recognition, which can efficiently and accurately recognize human body gestures and live wires, establishes a three-dimensional mathematical model of people and the live wires in real time and in advance, tracks the distance between the human body and the wires and gives early warning, aiming at the defects of the existing live working safety detection and early warning method of an electric power system.
The technical solution of the invention is as follows:
a live working safety early warning method based on gesture and wire recognition comprises the following steps:
first step, color image acquisition:
and (3) shooting a color picture by using a binocular structure optical camera, transmitting the video stream of the color binocular picture to a computer end program in real time for processing, and reading data of the color binocular picture. Calibrating parameters of a video camera to obtain internal and external parameters of the camera, wherein the internal parameters comprise a focal length, an imaging origin and a distortion coefficient, the external parameters comprise a rotation matrix and a translation vector, distortion elimination and line alignment are respectively carried out on a left view and a right view to obtain a color image which correctly reflects position information, and the image which can correctly reflect the position information is displayed.
And secondly, extracting live conductors in the image:
1. preprocessing of the image:
carrying out graying processing on the acquired image by using a built-in program, reading an input image in an RGB format, and processing a color image into a black, white and gray image; then, through gray value screening, setting the gray threshold value asθ gre In a gray scale image 0 andθ gre the black pixel point value between the two is set as the pixel point value of 255, the pixel values of other pixel points are all set as 0 values, the pixel values of other pixel points become black in the image, and the gray threshold value can be adjusted according to the actual working condition; and then, carrying out mean value filtering on the image pixels by adopting a Gaussian filtering method to obtain a gray level image.
2. And (3) extracting a straight line:
adopting a double-parameter Canny edge extraction algorithm, solving a pixel gradient map of the gray map according to a Sobel edge difference operator, and setting parametersθ canny1 And withθ canny2 Whereinθ canny2 ＞θ canny1 Gradient value of pixel greater thanθ canny2 The pixel points are identified as edge pixel points, and the gradient value of the pixel points is less than that of the edge pixel pointsθ canny1 The pixel points are identified as the pixel points which are not the edge, and the intermediate parameters are setθ canny1 And withθ canny2 And identifying the pixels between the pixels to be possible edge points, judging the possible edge points, setting the pixels as edge pixels if the pixels are adjacent to the edge pixels, and otherwise, setting the pixels as pixels which are not the edge. After the edge of the image is obtained, identifying a straight line set in the image by adopting Hough transformation, and setting a straight line voting threshold valuethresholdLength of interval from maximum straight linemaxLineGapAnd minimum line lengthminLineLengthNumerous sets of lines are available that may be live wires.
3. Screening of the conducting wires:
and classifying and screening the linear set to obtain a two-dimensional mathematical expression of the charged wire.
Thirdly, processing the image by a convolutional neural network:
1. human body posture detection model:
collecting N images with four limbs or five sense organs of a human body, and carrying out gray scale processing on the images to convert the images into gray scale images; and then calibrating the information of the four limbs and the five sense organs of the human body in the image, taking each gray image, the human body posture connecting line in the gray image and the information of the four limbs and the five sense organs as input, taking the input as a data set of a convolutional neural network, and performing machine learning and building a human body recognition model, so that the human body recognition model can accurately recognize the human body posture information in the image which is not in the data set, and further outputting the two-dimensional coordinates of the key joints in the human body posture.
2. Outputting a human body posture image:
on the basis of the first step, inputting the processed video stream picture into a human body posture detection model, outputting two-dimensional coordinates of key joints and five sense organs of the human body posture, wherein the key joint points comprise two eyes, two ears, the outermost edge point of the nose, two arms, two shoulders, two elbows, two wrists, left and right thighs, and two side end points of the left and right shanks, connecting all joint points in the image according to the shape conforming to the human body posture based on the human body posture model, and displaying the joint points in each frame of image of the color video stream by using thick straight lines.
Fourthly, predicting the posture of the human body:
and calculating displacement, speed, acceleration and other information of each joint point according to the position relation of the corresponding joint in the human body posture image output in the third step and the adjacent frames of images, predicting and correcting the position information of the joint at the next moment by linear extrapolation, and outputting the corrected position information at the next moment by adopting a Kalman filtering algorithm.
Step five, obtaining the depth distance:
the structured light binocular camera is used for capturing structural change of structured light by projecting infrared rays with a certain structure to an object, converting the change of the structured light into depth information through an operation unit integrated in the camera, transmitting the depth information to a computer end through an RTSP (real time streaming protocol) real-time streaming transmission protocol for processing, and obtaining the real distance from a camera optical center to a corresponding pixel in the real world, wherein the unit is millimeter; the binocular camera uses the parallax relation between two images, and performs auxiliary calculation on some points which cannot be accurately measured by structured light in the images through a limit-constrained stereo matching algorithm and a triangular distance measurement principle to obtain the depth value of each pixel point of the images.
Sixthly, calculating the distance between the lead and the human body:
based on the two-dimensional mathematical expression of the conductor in the second step 3 and the two-dimensional position coordinate of the human body predicted in the fourth step, the corresponding depth distance can be searched in the fifth step, a three-dimensional coordinate system is introduced according to the depth distance value, a three-dimensional space coordinate system of the human body and the conductor and the direction vector of the straight line where the high-voltage conductor is located are establishedn 1 =(x 1 ，y 1 ，z 1 ) And the direction vector of the straight line of the limb line segmentn 2 =(x 2 ，y 2 ，z 2 ). Respectively searching a point A and a point B on the two straight lines to obtain vectors
And there is a quantity>
The projection in the direction of the common perpendicular line, i.e. the distance between the two non-coplanar straight lines, is the shortest distance.
WhereindIn order to be the length of the projection,mis a common perpendicular vector obtained by cross-multiplying two vectors.
Step seven, safety early warning:
and if the shortest distance in the sixth step is smaller than the preset safety distance, the computer software sends out a warning signal, and meanwhile, an early warning annunciator worn by a worker sends out harsh buzzing sound to realize safety early warning.
The invention has the technical effects that: the invention captures the action behavior of the live working staff and the three-dimensional coordinates of the wire in real time by adopting the video camera, predicts the safety according to the action of the staff, judges the distance between the wire and the staff in real time and in advance, and determines whether to send out safety early warning or not by comparing the distance with the safety distance.
The invention combines the binocular camera and the structured light depth acquisition, is suitable for measuring the gesture depth distance between the lead and the human body, ensures the accuracy of depth ranging, and reduces the difficulty of depth value acquisition limited by the visual angle. The charged conductor two-dimensional expression in the image is effectively extracted through three-time screening of the conductor, the human posture is automatically recognized through a convolutional neural network deep learning method, the distance between the human body and the conductor is calculated in real time, and the human posture is predicted. In the live-wire work safety supervision, the labor cost is greatly reduced, the error rate of manually judging the distance is reduced, and the accuracy and the reliability of the live-wire work safety early warning are improved.
Drawings
FIG. 1 is a schematic diagram of a hot-line work safety early warning system according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of video stream color image preprocessing according to an embodiment of the present invention;
FIG. 3 is a flow chart of a method for screening and classifying identification lines of live conductors according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of human body posture detection according to an embodiment of the present invention;
fig. 5 is a schematic diagram of a lead and a three-dimensional human body frame model according to an embodiment of the invention.
Detailed Description
The following describes embodiments of the present invention in further detail with reference to the accompanying drawings.
Fig. 1 is a schematic diagram of a hot-line work safety early warning system based on posture and wire recognition.
In this embodiment, the distance between the lead and the posture of the human body is calculated, and the safety of the live working personnel is evaluated on line, as shown in fig. 1, the live working safety early warning method based on posture and lead recognition of the present invention includes the following steps:
s1, obtaining a human body posture model based on a convolutional neural network
Establishing a convolutional neural network, collecting 10000 images with four limbs or five sense organs of a human body, and performing gray processing on the images to convert the images into gray images; and then calibrating the information of the four limbs and the five sense organs of the human body in the image, taking each gray image and the human body posture connecting line in the gray image and the information of the four limbs and the five sense organs as input as a data set of a convolutional neural network, and carrying out machine learning and building a human body recognition model.
S2, transmission and display of video stream
The visual picture part is firstly sent to a raspberry group for data operation and processing to obtain pictures collected by a camera, then a depth map collected by a structured light camera and color pictures collected by a binocular camera are transmitted to a computer end through Ethernet by using RTSP transmission, and then the pictures subjected to human body posture recognition and wire recognition are displayed in real time by the programmed software.
S3, identifying straight lines and judging electrified leads
S3.1, graying of image
Performing graying processing on the acquired image by using a built-in program, reading an input image in an RGB format as shown in FIG. 2, and processing a color image into a black, white and gray image; then, through the gray value screening, the gray threshold value is set toθ gre Here, theIs provided withθ gre =30; 0 and 0 in the gray scale mapθ gre The black pixel point value between the two is set to be 255, the pixel point value is represented as white in the image, the pixel values of other pixel points are all set to be 0, the pixel value is represented as black in the image, and the gray threshold value can be adjusted according to the actual working condition; and then, carrying out mean value filtering on the image pixels by adopting a Gaussian filtering method to obtain a gray level image.
S3.2, edge extraction algorithm
And (3) solving a pixel gradient map of the middle gray level map according to a Sobel edge difference operator by adopting a double-parameter Canny edge extraction algorithm. Calculating the difference between the horizontal and vertical directions
And &>
. The gradient mode and direction can then be calculated as follows:
setting parameters according to a Canny edge extraction algorithmθ canny1 Andθ canny2 whereinθ canny2 ＞θ canny1 Gradient value of pixel greater thanθ canny2 The pixel points are identified as edge pixel points, and the gradient value of the pixel points is less than that of the edge pixel pointsθ canny1 Identifying the pixel points to be pixel points which are not edges, and determining the parameters between the pixel pointsθ canny1 Andθ canny2 and identifying the pixel points between the two pixels to be possible edge points, judging the possible edge points, if the pixel point is adjacent to the edge pixel point, setting the pixel point as the edge pixel point, otherwise, setting the pixel point as a pixel point which is not the edge. Herein, theθ canny1 Is set to be 50 in the above-mentioned order,θ canny2 set to 150.
S3.3, straight line recognition algorithm
After the image edge is obtained, identifying a straight line set in the image by adopting Hough transformation, and setting a straight line voting threshold valuethresholdLength of interval from maximum straight linemaxLineGapAnd minimum line lengthminLineLengthHere, a voting threshold value may be setthreshold=100, minimum line lengthminLineLength=30, maximum line intervalmaxLineGap=30。
S3.3, straight line screening
When the line set is classified, screening for three times to obtain a two-dimensional mathematical expression of the charged wire, as shown in fig. 3, when the line is screened for the first time, obtaining slope angles of all detected lines, obtaining an average slope angle of the slope angles and then screening and removing lines deviating from the average slope angle in all slope angles; most of the miscellaneous wires which do not belong to the electrified conducting wire are excluded, and the electrified conducting wire and a small number of miscellaneous wires with the same slope as the electrified conducting wire are left; the second screening classifies the straight lines with approximately same slope and approximately same intercept into a straight line group, and outputs the straight line of the path of the maximum path between two end points in the straight line group as a pre-selected electrified lead; outputting straight lines with the length meeting the requirement as actual charged conducting wires by the third screening, establishing a two-dimensional mathematical expression of the actual charged conducting wires by using a least square method, and outputting the straight lines as the actual charged conducting wires if the length of the straight lines screened for the third time exceeds a preset threshold; and if all the screening straight lines do not meet the threshold requirement, outputting the longest three straight lines.
S4, recognizing human body postures
Inputting the processed video stream picture into a human body posture detection model, as shown in fig. 4, outputting two-dimensional coordinates of key joints and five sense organs of the human body posture, wherein the key joint points comprise two eyes, two ears, the outermost edge point of the nose, and end points of two arms, two shoulders, two elbows, two wrists, left and right thighs, and left and right shanks, connecting all joint points in the image according to the shape conforming to the human body posture based on the human body posture model, and displaying the joint points in each frame image of the color video stream by using thick straight lines.
S5, establishing a three-dimensional model of the human body posture and the conducting wire
The image after the recognition processing of the human body posture model and the charged wire is used for establishing a three-dimensional model by searching the corresponding depth value in the depth map according to the two-dimensional coordinates of the key joint points of the human body and the two-dimensional coordinates of the two ends of the wire, as shown in figure 5.
S6, predicting human body posture
Let the current coordinate of a certain joint point P be: (x，y，z) The predicted coordinates arex_predict，y_predict，z_ predict). The relationship is as follows,
meanwhile, the time for processing one frame of image by the system is within 0.16s, and the coordinate in the previous frame is considered to be (x_old，y_ old，z_old) At this time, it can be considered that the speed and direction of the P-point motion are unchanged between two adjacent frames:
the predicted position of point P in the next frame can be obtained simultaneously:
according to the position relation of the corresponding joints in the adjacent frames of images of the output human body posture image, calculating the information of displacement, speed, acceleration and the like of each joint point, predicting the position information of the joint at the next moment by linear extrapolation, predicting and correcting the position information by adopting a Kalman filtering algorithm, and outputting the corrected position information at the next moment.
S7, calculating the real-time distance
Direction vector of straight line of high-voltage wiren 1 =(x 1 ，y 1 ，z 1 ) And the line of limb segmentDirection vectorn 2 =(x 2 ，y 2 ，z 2 ). Respectively searching a point A and a point B on two straight lines to obtain vectors
Then there is a direction amount>
The projection in the direction of the common perpendicular line, namely the distance between the two non-coplanar straight lines, is also the shortest distance.
WhereindIn order to be the length of the projection,mis a common perpendicular vector obtained by cross-multiplying two vectors.
S8, setting dangerous distances of different voltage levels
The voltage class of online input live working is set to the safety distances of different safety classes of 10kV or less, 20/35kV, 66/110kV, 220kV and 500kV in a software program according to the regulations made by the national power grid company electric safety work rules and the line part. The voltage class is set to be the dangerous distance from low to high: 0.7m, 1.0m, 1.5m, 3.0m, 5.0m.
S9, safety early warning
If the shortest distance is smaller than the preset safety distance, the computer software sends out a warning signal, and the computer is connected with an early warning annunciator worn on the body of a worker through the Ethernet.
The early warning annunciator comprises a single chip microcomputer system, a Wifi module, a led indicating lamp, a loudspeaker and a buzzer. The single chip microcomputer system adopts STM32F104 series, the wiFi module be used for the communication of computer end and early warning signal ware, the single chip microcomputer receives computer end alarm signal after, sends high-low signal control led pilot lamp, speaker and buzzer open and close, and then makes the early warning, and then sends early warning signal.
The live working safety early warning method based on gesture and wire recognition, which is recorded by the invention, has the following advantages:
(1) The binocular camera acquires the video stream of the color image and the depth map video stream returned by the structured light camera, and the video stream is transmitted to a computer end for processing through an RTSP video stream transmission protocol and Ethernet packaging, so that the reliability and the real-time property of information transmission are improved.
(2) The Canny and Hough algorithm based on double thresholds obtains a two-dimensional coordinate of the line through three times of recognition, so that the effective filtering of the non-conducting wire and noise points of the conducting wire is enhanced, and a two-dimensional mathematical expression of the live conducting wire is more accurately output.
(3) And based on the three-dimensional mathematical model, the human body posture is predicted in one step by using a Kalman filtering algorithm, and the predicted value is corrected according to the returned human body posture coordinate at the prediction moment, so that the accurate prediction of the future coordinate of the human body posture is realized.
(4) The minimum distance between a certain vector and a certain wire in the human body posture is obtained by calculating the distance between the human body and the wire and using the vector algorithm, whether the live-line overhaul personnel is in a safe state or not is judged according to the live-line operation voltage level input by the computer end, if the live-line operation voltage level is in the unsafe state, the computer end sends an alarm signal to the early warning annunciator to send a strong signal, so that the safety early warning system is simple and efficient and is also suitable for live-line operation environments under different working conditions.
Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can modify and substitute the detailed embodiments of the present invention, and it is intended that all the inventions and inventions utilizing the concepts of the present invention are protected.
Claims (6)
1. A live working safety early warning method based on gesture and wire recognition is characterized in that: it comprises the following steps:
first step, color image acquisition:
shooting a color picture by using a binocular structure optical camera, transmitting a video stream of the color binocular picture to a computer end program in real time for processing, and reading data of the color binocular picture; calibrating parameters of a camera to obtain internal and external parameters of the camera, wherein the internal parameters comprise a focal length, an imaging origin and a distortion coefficient, the external parameters comprise a rotation matrix and a translation vector, distortion elimination and line alignment are respectively carried out on a left view and a right view to obtain a color image which correctly reflects position information, and the image which can correctly reflect the position information is displayed;
second step, extraction of live lines in the image:
(1) And image preprocessing:
carrying out graying processing on the acquired image by using a built-in program, reading an input image in an RGB format, and processing a color image into a black, white and gray image; then, through gray value screening, setting the gray threshold value asθ gre 0 in the gray scale mapθ gre The black pixel point value between the two is set as the pixel point value of 255, the pixel values of other pixel points are all set as 0 values, the pixel values of other pixel points become black in the image, and the gray threshold value can be adjusted according to the actual working condition; then, carrying out mean filtering on the image pixels by adopting a Gaussian filtering method to obtain a gray level image;
(2) And extracting a straight line:
adopting a Canny edge extraction algorithm with double parameters, solving a pixel gradient map of the gray map according to a Sobel edge difference operator, and setting parametersθ canny1 Andθ canny2 whereinθ canny2 ＞θ canny1 Gradient value of pixel greater thanθ canny2 The pixel points are identified as edge pixel points, and the gradient value of the pixel points is less than that of the edge pixel pointsθ canny1 The pixel points are identified as the pixel points which are not the edge, and the intermediate parameters are setθ canny1 Andθ canny2 identifying the pixels between the two pixels to be possible edge points, judging the possible edge points, if the pixel is adjacent to the edge pixel, setting the pixel as the edge pixel, otherwise, setting the pixel as a pixel which is not the edge; after the edge of the image is obtained, identifying a straight line set in the image by adopting Hough transformation, and setting a straight line voting threshold valuethresholdLength of interval from maximum straight linemaxLineGapAnd minimum line lengthminLineLengthNumerous straight wires are available which may be live wiresCollecting lines;
(3) Screening the lead:
classifying and screening the linear set to obtain a two-dimensional mathematical expression of the electrified lead;
thirdly, processing the image by a convolutional neural network:
(1) And a human body posture detection model:
collecting N images with four limbs or five sense organs of a human body, and carrying out gray scale processing on the images to convert the images into gray scale images; then, information of limbs and the five sense organs of the human body is calibrated in the image, each gray image, a human body posture connecting line in the gray image and the information of the limbs and the five sense organs are used as input and serve as a data set of a convolutional neural network, machine learning and building of a human body recognition model are carried out, the human body recognition model can accurately recognize human body posture information in the image which is not in the data set, and therefore two-dimensional coordinates of key joints in the human body posture are output;
(2) Outputting the human body posture image:
inputting the processed video stream picture into a human body posture detection model on the basis of the first step, outputting two-dimensional coordinates of key joints and five sense organs of the human body posture, wherein the key joint points comprise two eyes, two ears, outermost edge points of a nose, two arms, two shoulders, two elbows, two wrists, left and right thighs, and two side end points of the left and right shanks, connecting all joint points in the image according to the shape conforming to the human body posture on the basis of the human body posture model, and displaying the joint points in each frame of image of the color video stream by using thick straight lines;
fourthly, predicting the posture of the human body:
calculating information such as displacement, speed, acceleration and the like of each joint point according to the human body posture image output in the third step and the position relation of the corresponding joint in the adjacent frames of images, predicting and correcting the position information of the joint at the next moment by linear extrapolation, and outputting the corrected position information at the next moment by adopting a Kalman filtering algorithm;
step five, obtaining the depth distance:
the structured light binocular camera is used for capturing structural change of structured light by projecting infrared rays with a certain structure to an object, converting the change of the structured light into depth information through an operation unit integrated in the camera, transmitting the depth information to a computer end through an RTSP (real time streaming protocol) real-time streaming transmission protocol for processing, and obtaining the real distance from a camera optical center to a corresponding pixel in the real world, wherein the unit is millimeter; the binocular camera performs auxiliary calculation on some points which cannot be accurately measured by structured light in the image by using a parallax relation between two images through a three-dimensional matching algorithm of limit constraint and a triangular distance measurement principle to obtain a depth value of each pixel point of the image;
sixthly, calculating the distance between the lead and the human body:
based on the two-dimensional mathematical expression of the conductor in the second step 3 and the two-dimensional position coordinate of the human body predicted in the fourth step, the corresponding depth distance can be searched in the fifth step, a three-dimensional coordinate system is introduced according to the depth distance value, a three-dimensional space coordinate system of the human body and the conductor and the direction vector of the straight line where the high-voltage conductor is located are establishedn 1 =(x 1 ，y 1 ，z 1 ) And the direction vector of the straight line of the limb line segmentn 2 =(x 2 ，y 2 ，z 2 ) (ii) a Respectively searching a point A and a point B on the two straight lines to obtain vectors
And there is a quantity>
The projection in the direction of the common vertical line, namely the distance between the two different-surface straight lines, is the shortest distance;
whereindIn order to be the length of the projection,ma common perpendicular vector obtained by cross multiplication of the two vectors;
step seven, safety early warning:
if the shortest distance in the sixth step is smaller than the preset safety distance, the computer software sends out a warning signal, and meanwhile, an early warning annunciator worn by a worker sends out a harsh buzzer, so that safety early warning is realized.
2. The live-wire work safety early warning method based on posture and wire recognition of claim 1, characterized in that when the straight line set is classified, three times of screening are performed to obtain a two-dimensional mathematical expression of the live wire, when the straight line is screened for the first time, slope angles are obtained for all detected straight lines, an average slope angle of a sum of the slope angles is obtained, and then the straight lines deviating from the average slope angle to be too large in all the slope angles are screened and removed; most of the miscellaneous wires which do not belong to the electrified conducting wire are excluded, and the electrified conducting wire and a small number of miscellaneous wires with the same slope as the electrified conducting wire are left; the second screening classifies the straight lines with approximately same slope and approximately same intercept into a straight line group, and outputs the straight line of the path of the maximum path between two end points in the straight line group as a pre-selected electrified lead; and (3) performing third screening by taking the straight line with the length meeting the requirement as an actual electrified conducting wire, and establishing a two-dimensional mathematical expression of the actual electrified conducting wire by using a least square method.
3. The live-wire work safety early warning method based on posture and wire recognition of claim 2, characterized in that if the length of the straight line screened for the third time exceeds the preset threshold value, the straight line is output as the actual live wire; and if all the screening straight lines do not meet the threshold requirement, outputting the longest three straight lines.
4. The live working safety pre-warning method based on posture and wire recognition as claimed in claim 1, wherein in the safety pre-warning step, the voltage class of the live working is inputted on line, the safety distances of different safety classes of 20/35kV, 66/110kV, 220kV and 500kV, which are 10kV and below, are set inside the software program, and the voltage class is set to the dangerous distance from low to high: 0.7m, 1.0m, 1.5m, 3.0m, 5.0m.
5. The live working safety early warning method based on gesture and wire recognition as claimed in claim 1, characterized in that the early warning annunciator comprises a single chip microcomputer system, a Wifi module and led indicator light, a speaker, a buzzer.
6. The live-wire work safety early warning method based on posture and wire recognition as claimed in claim 5, characterized in that the single chip microcomputer system adopts STM32F104 series, the WiFi module is used for communication between the computer end and the early warning annunciator, and after receiving the alarm signal from the computer end, the single chip microcomputer sends high and low signals to control the on and off of the led indicator light, the loudspeaker and the buzzer, so as to make early warning and further send out the early warning signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210543595.XA CN115908839A (en) | 2022-05-19 | 2022-05-19 | Live working safety early warning method based on posture and wire recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210543595.XA CN115908839A (en) | 2022-05-19 | 2022-05-19 | Live working safety early warning method based on posture and wire recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115908839A true CN115908839A (en) | 2023-04-04 |
Family
ID=86488626
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210543595.XA Pending CN115908839A (en) | 2022-05-19 | 2022-05-19 | Live working safety early warning method based on posture and wire recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115908839A (en) |
-
2022
- 2022-05-19 CN CN202210543595.XA patent/CN115908839A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113313005B (en) | Power transmission conductor on-line monitoring method and system based on target identification and reconstruction | |
| CN110263686A (en) | A kind of construction site safety of image cap detection method based on deep learning | |
| CN112396658B (en) | Indoor personnel positioning method and system based on video | |
| CN109271844B (en) | Electrical cabinet electrical symbol recognition method based on OpenCV | |
| CN108537154A (en) | Transmission line of electricity Bird's Nest recognition methods based on HOG features and machine learning | |
| CN109117764A (en) | Using the method for color threshold method identification target object region electrical symbol in power monitoring | |
| CN109409395A (en) | Using the method for template matching method identification target object region electrical symbol in power monitoring | |
| CN109271872A (en) | A kind of judgement of high voltage isolator division state and trouble-shooter and method | |
| CN110348380A (en) | A kind of visualization grounding switch state video recognition system and method | |
| CN111523386B (en) | High-speed railway platform door monitoring and protecting method and system based on machine vision | |
| CN117079397B (en) | Wild human and animal safety early warning method based on video monitoring | |
| CN116866520B (en) | AI-based monorail crane safe operation real-time monitoring management system | |
| CN113283290A (en) | Power distribution room monitoring method based on 3D vision and RGB characteristics | |
| CN112532927A (en) | Intelligent safety management and control system for construction site | |
| CN115908839A (en) | Live working safety early warning method based on posture and wire recognition | |
| CN115171006B (en) | Detection method for automatically identifying person entering electric power dangerous area based on deep learning | |
| CN116862712A (en) | Electric power construction potential safety risk detection method and system based on thunder fusion | |
| CN116152945A (en) | Under-mine inspection system and method based on AR technology | |
| CN103198326B (en) | A kind of image classification method of transmission line of electricity helicopter routing inspection | |
| CN114330477B (en) | Power equipment defect detection system and method based on mixed reality equipment | |
| CN109063634A (en) | Using the method for hough-circle transform identification target object region electrical symbol in power monitoring | |
| Rong et al. | A joint faster RCNN and stereovision algorithm for vegetation encroachment detection in power line corridors | |
| CN111881863B (en) | Regional group abnormal behavior detection method | |
| CN114325573A (en) | Method for rapidly detecting identity and position information of operation and maintenance personnel of transformer substation | |
| CN113076825A (en) | Transformer substation worker climbing safety monitoring method |
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 |