CN116168337A - System for intelligent image recognition of unworn violations of safety helmet on electric power construction site - Google Patents
System for intelligent image recognition of unworn violations of safety helmet on electric power construction site Download PDFInfo
- Publication number
- CN116168337A CN116168337A CN202211596770.8A CN202211596770A CN116168337A CN 116168337 A CN116168337 A CN 116168337A CN 202211596770 A CN202211596770 A CN 202211596770A CN 116168337 A CN116168337 A CN 116168337A
- Authority
- CN
- China
- Prior art keywords
- construction site
- image
- training
- violations
- unworn
- 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
- 238000010276 construction Methods 0.000 title claims abstract description 66
- 238000012549 training Methods 0.000 claims abstract description 67
- 238000012544 monitoring process Methods 0.000 claims abstract description 44
- 238000012545 processing Methods 0.000 claims abstract description 20
- 238000003384 imaging method Methods 0.000 claims abstract description 15
- 230000008054 signal transmission Effects 0.000 claims abstract description 11
- 230000000007 visual effect Effects 0.000 claims abstract description 8
- 238000012216 screening Methods 0.000 claims abstract description 7
- 230000005540 biological transmission Effects 0.000 claims description 18
- 238000012360 testing method Methods 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 6
- 230000006399 behavior Effects 0.000 claims description 4
- 238000013527 convolutional neural network Methods 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000007405 data analysis Methods 0.000 claims 1
- 238000001514 detection method Methods 0.000 description 15
- 230000006870 function Effects 0.000 description 9
- 238000004458 analytical method Methods 0.000 description 8
- 238000004891 communication Methods 0.000 description 8
- 238000000034 method Methods 0.000 description 7
- 238000004519 manufacturing process Methods 0.000 description 5
- 238000011161 development Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000012423 maintenance Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000003062 neural network model Methods 0.000 description 3
- 230000005856 abnormality Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000006378 damage Effects 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012552 review Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 230000001629 suppression Effects 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 230000001012 protector Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- 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
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Evolutionary Computation (AREA)
- Entrepreneurship & Innovation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- General Business, Economics & Management (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Development Economics (AREA)
- Computing Systems (AREA)
- Educational Administration (AREA)
- Artificial Intelligence (AREA)
- Game Theory and Decision Science (AREA)
- Human Computer Interaction (AREA)
- Social Psychology (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Psychiatry (AREA)
- Computational Linguistics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Alarm Systems (AREA)
Abstract
The invention provides a system for performing imaging intelligent recognition on unworn violations of a safety helmet on an electric power construction site, and belongs to the technical field of image recognition. The utility model provides a system that electric power job site was not worn the violation and is carried out imaging intelligent recognition to safety helmet, includes the video acquisition module, the video information that the video acquisition module obtained is sent to the backstage image monitoring module through signal transmission module and is carried out data processing to make the condition feedback with the result of data processing by the system alarm module, the training step of intelligent recognition model is: step 1: screening qualified pictures for forming a training set; step 2: constructing a residual network layer based on a Darknet-53 architecture; step 3: training the images in the training set; step 4: optimizing parameters; step 5: checking a training set; step 6: and identifying the field image. On the basis of visual monitoring of a construction site, a model algorithm is adopted to realize effective supervision and management on whether field constructors wear labor protection equipment.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to a system for performing imaging intelligent recognition on unworn violations of safety helmets on an electric power construction site.
Background
The electric power capital construction is an important component of the new capital construction in China. The method plays a vital role in 5G base station power supply, direct current and extra-high voltage power transmission and transformation, a new rail transit driving system, an energy charging pile, electric power Internet of things and the like.
With the deep promotion of the economic development and energy production and consumption revolution in China, higher requirements are also put forward on the safety supervision of the electric power capital construction.
The traditional safety supervision of the electric power capital construction is mostly realized by manual supervision, and the method is time-consuming and labor-consuming, has low supervision efficiency, is extremely easy to cause missed detection due to manual negligence, brings potential safety hazards, and is capable of frequently causing injuries such as high-altitude falling, object hitting, collapse and the like on the electric power line construction site. The following safety hazards mainly exist.
(1) The safety manager cannot track the site construction process in the whole course, and the site, the operation object, the operation environment, the operation personnel and the like to be observed are different, so that the safety manager cannot monitor the site in all due to limited energy even in the site.
(2) The constructor on site lacks sufficient safety consciousness, and safety consciousness and risk recognition capability are limited, and even safety measures or wrong entering intervals can be changed, so that safety accidents such as striking, falling and the like of some objects occur, and if safety equipment is not worn, the consequences are serious.
(3) By adopting video monitoring, people can monitor videos and identify the videos, the problems of monitoring omission and false identification are very easy to occur, the overall efficiency is low, the accuracy is low, and the safety management blind area is large.
For example, in the CN109784613a patent, a work log mode is adopted to implement management and control of illegal operations, but the mode does not consider the behavior of irregular wearing of protection equipment on a construction site, which often causes security accidents of incomplete protection.
Patent CN104346959A realizes the effective management and control to the violation, but this mode relies on manual identification operation, very easily takes place to monitor omission and misrecognition's problem, and overall efficiency is not high, and the rate of accuracy is also very low, and the safety control blind area is very big.
Disclosure of Invention
The invention aims to solve the technical problem of providing a system for intelligently identifying whether a worker on an electric power construction site wears a safety helmet (work clothes) or not and performing imaging on the safety helmet which is not worn against rules based on the condition of video monitoring.
In order to solve the technical problems, the invention adopts the following technical scheme:
the system for performing imaging intelligent recognition on unworn violations of the safety helmet in the electric power construction site comprises a video acquisition module, wherein video information acquired by the video acquisition module is sent into a background image monitoring module through a signal transmission module for data processing, and a system alarm module is used for performing condition feedback on the data processing result,
the background image monitoring module carries out the training steps of the intelligent recognition model:
step 1: reading the data stream of the monitoring video of the construction site, screening 3000 qualified pictures, and forming a training set by using the detected tag images of the safety helmet (working clothes);
step 2: constructing a multi-branch convolutional neural network based on a Darknet-53 architecture, and adding a residual network layer after the convolutional layer;
step 3: inputting the training set in the step 1 into the depth residual error network in the step 2, and training the safety helmet (working clothes) image in the training set;
step 4: repeating the step 3 for M times, training to obtain a depth residual error network training model, and obtaining N multiplied by M loss function values; then find out the minimum loss function value from all loss function values; then, the weight vector and the bias term corresponding to the loss function value with the minimum value are correspondingly used as the optimal weight vector and the optimal bias term of the depth residual error network training model; wherein M is more than 1;
step 5: randomly selecting a safety helmet (work clothes) image with the width of K and the height of G as a test image; then inputting the test image into a depth residual error network training model for prediction, obtaining a classification result of the test image by prediction, comparing the predicted classification result with the actual situation of the safety helmet image, and verifying whether the prediction result of the depth residual error network training model is accurate; if the depth residual error network training model test result is accurate, performing step 6; otherwise, returning to the step 4;
step 6: accessing a construction site data stream, inputting data to be predicted into the depth residual error network training model in the step 5 for prediction, and obtaining a classification result of a safety helmet (work clothes) image to be predicted by prediction; the classification result is normal wearing or non-normal wearing.
The video acquisition module is a distributed camera monitoring framework which is arranged on a construction site and fully covers a construction area, the signal transmission module adopts compatible wireless transmission equipment to realize data transmission of video information, the background image monitoring module comprises a monitoring workstation, a server and a disk array, and the system alarm module adopts an audible and visual alarm device to realize warning of illegal behaviors.
In the step 1: the time length of screening out the video information of the qualified picture is within T1 min, the qualified picture comprises at least one picture of an image acquired by any camera, the training set comprises the qualified picture, a time node corresponding to the qualified picture and a camera number, and the training set is stored in a disk array in an incremental backup mode.
The computer program of step 2 is provided in the server.
The step 3: and carrying out image recognition training on whether the safety helmet (working clothes) is provided with the differential mark or not on the qualified pictures in the training set.
And 4, adopting a repeated recognition training mode to perform optimization solving on the value vector and the bias term of the depth residual error network training model.
Step 5, verifying the model identification, if the model identification is successful, executing step 6, otherwise, re-executing step 4; step 5 is to analyze whether the qualified picture strives to wear the safety helmet (work clothes) and output the processing result,
and 5, after the verification is successful, backing up the depth residual error network training model,
and after the verification in the step 5 is successful, repeatedly executing the step 1 every time T2, and importing the qualified picture in the step 1 into the step 6 to perform violation identification.
Safety is the basis of all works, and enterprises, power grids and staff development are interviews without safety guarantee. The method ensures the safety and personal safety of the power grid, is a great social responsibility and political responsibility of the power grid company, and is a departure point and a foothold of the safety work of the company.
The safe and stable operation of the power grid is related to national folk life; personnel safety of staff, happiness of family and parents is directly related, and relationship enterprises and society are stable; maintaining a good situation of safety and stability is of great importance to the continuous healthy development of companies.
According to the design thought of the three-integration five-major system construction of the national network company and the development direction of regulation and control integration, operation and maintenance integration and maintenance specialization, the management system gradually transits to the intensive and specialized direction, and the contradiction between the increase of production activities and the lack of management staff existing in the safety supervision and management of the power grid enterprise is very prominent, such as the fact that the safety responsibility is not implemented in the production activities, the safety supervision is not in place, the safety training is not developed, and other key problems are not effectively controlled, thereby causing accidents such as personal casualties to field operation, power grid outage and equipment damage.
Therefore, the enhancement of the safety awareness of the whole staff is a great importance of the construction of the current safety supervision system, the safety quality of management staff and on-site staff at all levels in production activities is enhanced, the level of company safety management is improved, and the safety supervision of the power grid enterprise on the power production work site is realized.
At present, the following problems exist in the safety supervision and management of power grid enterprises:
the on-site point is multi-faceted and wide: the electric power system of the last two years is very prominent towards structural absences; depending on the existing human resources, no method is available for performing safety supervision full coverage on all operation sites;
the efficiency is low: the safety flight inspection mainly comprises the steps that 4 groups of supervising staff run to the site for inspection, the supervising staff run to each operation site for a long time, the working efficiency is low due to a working mechanism, and meanwhile, the risk of traffic accidents of the staff is increased;
the electric power line construction site is often damaged by high altitude falling, object striking, collapse and the like.
At present, the following modes are mainly adopted for safety management and violation correction of the power line construction site:
(1) Safety management personnel are specially arranged for regular inspection;
(2) And part of the construction area is additionally provided with a camera monitoring device, and personnel monitor whether violations exist by watching the display picture.
The following technical drawbacks are mainly present.
(1) The safety manager cannot track the site construction process in the whole course, and the site, the operation object, the operation environment, the operation personnel and the like to be observed are different, so that the safety manager cannot monitor the site in all due to limited energy even in the site.
(2) The constructor on site lacks sufficient safety consciousness, and safety consciousness and risk recognition capability are limited, and even safety measures or wrong entering intervals can be changed, so that safety accidents such as striking, falling and the like of some objects occur, and if safety equipment is not worn, the consequences are serious.
(3) By adopting video monitoring, people can monitor videos and identify the videos, the problems of monitoring omission and false identification are very easy to occur, the overall efficiency is low, the accuracy is low, and the safety management blind area is large.
Therefore, a set of intelligent video monitoring system for the electric power construction site is required to be developed, whether constructors wear safety helmets, work clothes and the like is detected, and once abnormality is found, the alarm is given out in time and pushed to safety management staff for processing.
The system can be divided into four modules of video acquisition, signal transmission, background image monitoring and system alarm. At the electric power construction site, the actual working condition is shot through an installed camera, the working picture of site constructors is collected, the image data is transmitted to a server side through a 4G signal, the image transmitted back from the front end is processed through intelligent video analysis of the server side, the processed result is displayed on an image display device, intelligent processing of the image is completed, and meanwhile the processed result is stored in a magnetic disk for further checking.
1 video acquisition module
The video acquisition module utilizes a high-resolution camera arranged on an electric power construction site to acquire construction images, and meanwhile, the camera is required to be provided with a special power supply to supply power in real time. The wireless network camera is mainly used in the power construction site because the wireless network camera is simple to install, and most of the power construction site is located in a remote place in the field, so that the installation of the monitoring camera is mainly simple to install and convenient to use. Since construction sites are mostly in the field, the power supply problem is one of the factors that must be considered when installing the camera.
In order to solve the problem that the high-low voltage conversion of the power supply is difficult in the monitoring and installation of the camera on the power construction site, the power supply is powered by adopting a mode of combining the solar panel with the storage battery, so that the power supply problem is simplified, and the method is safer and more convenient. The solar panel adopted in the power supply module is composed of a plurality of solar cells, is safe, energy-saving and environment-friendly, light in weight and easy to install, and is suitable for stable power supply for safety monitoring in the power construction site. The camera acquires high-definition images with the resolution ratio of 1920 multiplied by 1080, and the high-definition images are transmitted through the signal transmission module after the video image acquisition module.
2 signal transmission module
In the power construction safety monitoring system, the signal transmission module uses a 4G (5G compatible) communication technology for transmission, and the transmission mode is advanced in recent years and can realize remote transmission. The 4G mobile communication technology has certain advancement, and can obtain the most efficient and convenient communication mode on a plurality of network systems, platforms and wireless communication, so that the system can adapt to rapid transmission, receiving and positioning operations.
The 4G communication technology is optimized and upgraded on the 3G communication technology, high-definition transmission of images and videos can be realized, and the transmission quality can be comparable to that of computer transmission. Compared with the characteristic of slow transmission speed of 3G communication, the speed of 4G communication can reach tens of megabytes per second at the highest speed, and the transmission speed is far more than that of 3G communication. Therefore, the 4G signal coverage is performed around the power construction, so that the data transmission between the construction site and the background can be realized. The plurality of cameras are connected to the same switch, so that the electric power construction safety system can monitor a plurality of construction sites simultaneously, and subsequent image processing is facilitated.
3 background image monitoring module
The background image processing module realizes the functions of image transmission, storage, processing and the like. The module mainly comprises three components: the monitoring workstation, the server and the disk array cooperate with each other to realize intelligent processing of the background image. The image data transmitted through the 4G signals are stored by the disk array, the image data are displayed on the monitoring workstation, and workers on the power grid can regularly monitor and replay pictures at the monitoring center to watch construction conditions of different monitoring points and wearing conditions of safety protection tools of the workers.
Meanwhile, an image target detection algorithm, namely a multi-branch deep learning algorithm, is embedded in the intelligent analysis server and is used for target detection of safety helmets and workwear on the power construction site, the algorithm is installed in the intelligent analysis server in a software mode, and meanwhile, a GPU (graphic processing unit) is arranged in the intelligent analysis server and is used for improving the detection efficiency of the algorithm, so that the purpose of real-time operation is achieved.
The background image monitoring module is the core of the whole power construction monitoring system, integrates an intelligent analysis algorithm, performs target detection on an image transmitted back from the front end, and then transmits a detection result to the system alarm module as a signal.
4 system alarm module
The system alarm module adopts an audible and visual alarm device which is connected with the intelligent analysis server of the previous module, and when the safety protector is not worn in the detection result of the intelligent analysis server, the system gives an alarm to remind a safety management and control responsible person to manage on-site constructors.
The audible and visual alarm is mainly a signal device for alarming on-site personnel by utilizing sound and a light source at a place easy to generate dangerous emergency. In the power construction monitoring system, an audible and visual alarm device is connected with an intelligent image processing module, and after the detection result of the module is received, the audible and visual alarm device sends out a signal whether to alarm or not.
Compared with the prior art, the invention has the following beneficial effects:
(1) When the intelligent target recognition algorithm is designed, compared with a traditional neural network model structure, the intelligent target recognition system uses a multi-branch neural network model frame based on the Darknet-53, and by adding a residual network layer, the detection efficiency of large, medium and small targets in a complex electric power construction scene is effectively improved.
(2) The neural network detection algorithm needs to perform non-maximum suppression operation to select a sufficiently correct prediction result, and compared with the traditional algorithm, the system replaces the selection index IoU of the prediction result with GIoU, so that the detection capability of safety helmets and working clothes in complex and various environments is improved, and particularly, when a target shielding condition exists, the detection accuracy is obviously improved.
(3) Saving the cost.
Compared with the traditional safety monitoring mode, the system saves human resources and reduces personnel cost. Based on the intelligent management and control system, remote positioning and monitoring of site constructors can be realized by only a small amount of management staff, and the advantages of long distance, high efficiency, low consumption and low cost are reflected. In addition, through recording construction equipment operating condition, help the managers in time to discover the trouble and maintain, effectively promoted on-the-spot safety management level.
(4) The safety is improved.
According to the research and analysis of construction safety accidents, the safety accidents caused by human factors are obvious. Because the mobility of personnel is large and uncontrollable factors are many, the 'person-in-person' mode is unavoidable to have a omission. The system is utilized to continuously monitor the safety protection measures of constructors, and can immediately give an alarm once abnormality is found.
(5) Monitoring system with high recognition rate
By improving a multi-branch neural network model framework based on Darknet-53, a residual error unit layer is added after a traditional convolutional neural network layer, so that the detection performance of a small target is improved; when the neural network detection algorithm carries out non-maximum suppression operation, the prediction result conforming to the GIoU index is adopted, so that the detection capability of safety helmets and working clothes in complex and diverse environments is improved.
Drawings
The present invention will be described in further detail with reference to the accompanying drawings.
Fig. 1: the structure of the invention is schematically shown;
fig. 2: the invention discloses a training step schematic diagram of an intelligent recognition model.
Detailed Description
For a better understanding of the present invention, the content of the present invention will be further clarified below with reference to the examples and the accompanying drawings, but the scope of the present invention is not limited to the following examples only. In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the invention may be practiced without one or more of these details.
Example 1
As shown in fig. 1-2, the embodiment provides a system for performing imaging intelligent recognition on unworn violations of a helmet in an electric power construction site, which comprises a video acquisition module, wherein video information acquired by the video acquisition module is sent into a background image monitoring module through a signal transmission module for data processing, and a system alarm module feeds back the result of the data processing,
the background image monitoring module carries out the training steps of the intelligent recognition model:
step 1: reading the data stream of the monitoring video of the construction site, screening 3000 qualified pictures, and forming a training set by using the detected tag images of the safety helmet (working clothes);
step 2: constructing a multi-branch convolutional neural network based on a Darknet-53 architecture, and adding a residual network layer after the convolutional layer;
step 3: inputting the training set in the step 1 into the depth residual error network in the step 2, and training the safety helmet (working clothes) image in the training set;
step 4: repeating the step 3 for M times, training to obtain a depth residual error network training model, and obtaining N multiplied by M loss function values; then find out the minimum loss function value from all loss function values; then, the weight vector and the bias term corresponding to the loss function value with the minimum value are correspondingly used as the optimal weight vector and the optimal bias term of the depth residual error network training model; wherein M is more than 1;
step 5: randomly selecting a safety helmet (work clothes) image with the width of K and the height of G as a test image; then inputting the test image into a depth residual error network training model for prediction, obtaining a classification result of the test image by prediction, comparing the predicted classification result with the actual situation of the safety helmet image, and verifying whether the prediction result of the depth residual error network training model is accurate; if the depth residual error network training model test result is accurate, performing step 6; otherwise, returning to the step 4;
step 6: accessing a construction site data stream, inputting data to be predicted into the depth residual error network training model in the step 5 for prediction, and obtaining a classification result of a safety helmet (work clothes) image to be predicted by prediction; the classification result is normal wearing or non-normal wearing.
The video acquisition module is a distributed camera monitoring framework which is arranged on a construction site and fully covers a construction area, the signal transmission module adopts compatible wireless transmission equipment to realize data transmission of video information, the background image monitoring module comprises a monitoring workstation, a server and a disk array, and the system alarm module adopts an audible and visual alarm device to realize warning of illegal behaviors.
In the step 1: the time length of screening out the video information of the qualified picture is within T1 min, the qualified picture comprises at least one picture of an image acquired by any camera, the training set comprises the qualified picture, a time node corresponding to the qualified picture and a camera number, and the training set is stored in a disk array in an incremental backup mode.
The computer program of step 2 is provided in the server.
The step 3: and carrying out image recognition training on whether the safety helmet (working clothes) is provided with the differential mark or not on the qualified pictures in the training set.
And 4, adopting a repeated recognition training mode to perform optimization solving on the value vector and the bias term of the depth residual error network training model.
Step 5, verifying the model identification, if the model identification is successful, executing step 6, otherwise, re-executing step 4; step 5 is to analyze whether the qualified picture strives to wear the safety helmet (work clothes) and output the processing result,
and 5, after the verification is successful, backing up the depth residual error network training model,
and after the verification in the step 5 is successful, repeatedly executing the step 1 every time T2, and importing the qualified picture in the step 1 into the step 6 to perform violation identification.
And D, the system alarm module carries out alarm early warning according to the non-standard wearing classification result output in the step six.
And the training set formed by the qualified pictures is stored in the disk array in an incremental backup mode.
The system alarm module is arranged on one side of the distributed camera in the construction site.
Example 2
And if the video information fed back by the camera in the video acquisition module is not completely online, the background image monitoring module sends out an identification code that the equipment is not completely online, and a warning is made through the system alarm module, the display screen and the third party platform, and the number of the camera which is not online is displayed.
The related monitoring personnel can inform the maintenance work team of carrying out maintenance work according to the number of the camera which is not on line.
The universal management of the field monitoring is realized, the safety management system is perfected, and the supervision quality is further improved.
The video acquisition module adopts a compatibility protocol interface, can realize video reading of the camera operated on the construction site, adopts a sharing mode among platforms, realizes cooperative utilization of hardware facilities, optimizes the setting quantity of the cameras, and reduces economic cost.
Example 3
And step 6, linking a face recognition algorithm, carrying out frame selection on pictures which are not regularly worn by field operators, and importing the pictures into the face recognition algorithm to realize the identification of the illegal personnel.
And constructing an archive directory of the unnormalized wearing safety helmet (working clothes), and importing the identification result of the offender into the archive directory of the unnormalized wearing safety helmet (working clothes) to realize intelligent management and control of relevant information of the offender.
The frame selection mechanism realizes the frame selection of the person in the construction range in the image.
The face recognition algorithm is internally provided with head portrait information and work numbers of construction site operators.
And the face recognition algorithm performs face matching according to the frame selection result in the picture of the work piece of the work site worker which is not regularly worn with the safety helmet (work clothes), and outputs the work number of the work piece corresponding to the qualified matching into the file directory of the work piece which is not regularly worn with the safety helmet (work clothes).
Example 4
The training model in the step 4 can be repeatedly executed according to the requirement, so that the accuracy and the processing speed of image recognition are improved.
And (3) adding an artificial review mechanism to the unnormalized wearing classification picture output in the step (6).
And (4) re-importing the pictures which do not pass the review in the step (4) for re-training until a correct result is output.
Finally, it is noted that the above-mentioned embodiments are merely for illustrating the technical solution of the present invention, and that other modifications and equivalents thereof by those skilled in the art should be included in the scope of the claims of the present invention without departing from the spirit and scope of the technical solution of the present invention.
Claims (9)
1. The utility model provides a system of electric power job site carries out imaging intelligent recognition to unworn violations of safety helmet which characterized in that: comprises a video acquisition module, wherein video information acquired by the video acquisition module is sent into a background image monitoring module through a signal transmission module for data processing, and the result of the data processing is subjected to condition feedback by a system alarm module,
the background image monitoring module carries out the training steps of the intelligent recognition model:
step 1: reading the data stream of the monitoring video of the construction site, screening 3000 qualified pictures, and forming a training set by using the detected tag images of the safety helmet (working clothes);
step 2: constructing a multi-branch convolutional neural network based on a Darknet-53 architecture, and adding a residual network layer after the convolutional layer;
step 3: inputting the training set in the step 1 into the depth residual error network in the step 2, and training the safety helmet (working clothes) image in the training set;
step 4: repeating the step 3 for M times, training to obtain a depth residual error network training model, and obtaining N multiplied by M loss function values; then find out the minimum loss function value from all loss function values; then, the weight vector and the bias term corresponding to the loss function value with the minimum value are correspondingly used as the optimal weight vector and the optimal bias term of the depth residual error network training model; wherein M is more than 1;
step 5: randomly selecting a safety helmet (work clothes) image with the width of K and the height of G as a test image; then inputting the test image into a depth residual error network training model for prediction, obtaining a classification result of the test image by prediction, comparing the predicted classification result with the actual situation of the safety helmet image, and verifying whether the prediction result of the depth residual error network training model is accurate; if the depth residual error network training model test result is accurate, performing step 6; otherwise, returning to the step 4;
step 6: accessing a construction site data stream, inputting data to be predicted into the depth residual error network training model in the step 5 for prediction, and obtaining a classification result of a safety helmet (work clothes) image to be predicted by prediction; the classification result is normal wearing or non-normal wearing.
2. The system for intelligent imaging identification of unworn violations of helmets at an electrical construction site of claim 1, wherein: the video acquisition module is a distributed camera monitoring framework which is arranged on a construction site and fully covers a construction area, the signal transmission module adopts compatible wireless transmission equipment to realize data transmission of video information, the background image monitoring module comprises a monitoring workstation, a server and a disk array, and the system alarm module adopts an audible and visual alarm device to realize warning of illegal behaviors.
3. The system for intelligent imaging identification of unworn violations of helmets at an electrical construction site of claim 1, wherein: in the step 1: the time length of screening out the video information of the qualified picture is within T1 min, the qualified picture comprises at least one picture of an image acquired by any camera, the training set comprises the qualified picture, a time node corresponding to the qualified picture and a camera number, and the training set is stored in a disk array in an incremental backup mode.
4. The system for intelligent imaging identification of unworn violations of helmets at an electrical construction site of claim 2, wherein: the computer program of step 2 is provided in the server.
5. The system for intelligent imaging identification of unworn violations of helmets at an electrical construction site of claim 1, wherein: the step 3: and carrying out image recognition training on whether the safety helmet (working clothes) is provided with the differential mark or not on the qualified pictures in the training set.
6. The system for intelligent imaging identification of unworn violations of helmets at an electrical construction site of claim 1, wherein: and 4, adopting a repeated recognition training mode to perform optimization solving on the value vector and the bias term of the depth residual error network training model.
7. The system for intelligent imaging identification of unworn violations of helmets at an electrical construction site of claim 1, wherein: step 5, verifying the model identification, if the model identification is successful, executing step 6, otherwise, re-executing step 4; and 5, carrying out data analysis on whether the qualified picture strives to wear the safety helmet (work clothes) or not and outputting a processing result.
8. The system for intelligent imaging identification of unworn violations of helmets at an electrical construction site of claim 1, wherein: and 5, after the verification is successful, backing up the depth residual error network training model.
9. The system for intelligent imaging identification of unworn violations of helmets at an electrical construction site of claim 1, wherein: and after the verification in the step 5 is successful, repeatedly executing the step 1 every time T2, and importing the qualified picture in the step 1 into the step 6 to perform violation identification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211596770.8A CN116168337A (en) | 2022-12-12 | 2022-12-12 | System for intelligent image recognition of unworn violations of safety helmet on electric power construction site |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211596770.8A CN116168337A (en) | 2022-12-12 | 2022-12-12 | System for intelligent image recognition of unworn violations of safety helmet on electric power construction site |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116168337A true CN116168337A (en) | 2023-05-26 |
Family
ID=86420987
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211596770.8A Pending CN116168337A (en) | 2022-12-12 | 2022-12-12 | System for intelligent image recognition of unworn violations of safety helmet on electric power construction site |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116168337A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116883952A (en) * | 2023-09-07 | 2023-10-13 | 吉林同益光电科技有限公司 | Electric power construction site violation identification method and system based on artificial intelligence algorithm |
| CN118072356A (en) * | 2024-04-11 | 2024-05-24 | 克拉玛依市富城油气研究院有限公司 | Well site remote monitoring system and method based on Internet of things technology |
-
2022
- 2022-12-12 CN CN202211596770.8A patent/CN116168337A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116883952A (en) * | 2023-09-07 | 2023-10-13 | 吉林同益光电科技有限公司 | Electric power construction site violation identification method and system based on artificial intelligence algorithm |
| CN116883952B (en) * | 2023-09-07 | 2023-11-17 | 吉林同益光电科技有限公司 | Electric power construction site violation identification method and system based on artificial intelligence algorithm |
| CN118072356A (en) * | 2024-04-11 | 2024-05-24 | 克拉玛依市富城油气研究院有限公司 | Well site remote monitoring system and method based on Internet of things technology |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110378606B (en) | Mobile visual safety management and control system | |
| CN116168337A (en) | System for intelligent image recognition of unworn violations of safety helmet on electric power construction site | |
| CN108915771B (en) | Cloud control mine comprehensive management system | |
| CN202795643U (en) | All-weather submarine cable security monitoring and control device | |
| CN112785458A (en) | Intelligent management and maintenance system for bridge health big data | |
| CN111768123A (en) | Visual management system for power distribution network construction site | |
| CN105989682A (en) | Safety early warning and monitoring system and monitoring method for construction machinery under power transmission line | |
| CN112308510A (en) | Green and environment-friendly building construction management system and method | |
| CN112112629A (en) | Safety business management system and method in drilling operation process | |
| CN110652684A (en) | Electric fire safety integrated management system | |
| CN110121053A (en) | A kind of video monitoring method of situ of drilling well risk stratification early warning | |
| CN106251240A (en) | Power transmission network method for early warning based on big data | |
| CN108389360A (en) | A kind of fire disaster of power transmission line monitoring method and system based on video analysis | |
| CN204375138U (en) | Based on the intelligent early-warning system of people's current density recognition technology | |
| CN112702570A (en) | Security protection management system based on multi-dimensional behavior recognition | |
| CN112686511A (en) | Construction safety management system based on building information model | |
| CN116193083A (en) | Safety identification system based on artificial intelligence | |
| CN115052129A (en) | Construction site visual supervision system and supervision method thereof | |
| CN105811578B (en) | Transmission line of electricity monitor supervision platform and its Power Supply Monitoring algorithm and image warning algorithm | |
| CN115169777A (en) | Intelligent supervision system for safety operation | |
| CN116882670A (en) | Intelligent building site management system for water transport engineering based on BIM | |
| CN111510685A (en) | Information output method and device based on three-dimensional model | |
| CN117081917A (en) | Intelligent gateway based on AI (advanced technology attachment) internet of things sewage treatment operation and maintenance management | |
| CN116882729A (en) | Security management platform based on 5g wisdom power plant | |
| CN115953278A (en) | Municipal works wisdom building site management system based on BIM |
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 |