CN113807240A - Intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition - Google Patents
Intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition Download PDFInfo
- Publication number
- CN113807240A CN113807240A CN202111082263.8A CN202111082263A CN113807240A CN 113807240 A CN113807240 A CN 113807240A CN 202111082263 A CN202111082263 A CN 202111082263A CN 113807240 A CN113807240 A CN 113807240A
- Authority
- CN
- China
- Prior art keywords
- face recognition
- face
- dressing
- transformer substation
- uncooperative
- 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
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000012544 monitoring process Methods 0.000 title claims abstract description 31
- 238000001514 detection method Methods 0.000 claims abstract description 107
- 238000012549 training Methods 0.000 claims abstract description 46
- 238000012360 testing method Methods 0.000 claims abstract description 17
- 230000008569 process Effects 0.000 claims description 15
- 238000002372 labelling Methods 0.000 claims description 10
- 230000001681 protective effect Effects 0.000 claims description 8
- 238000012795 verification Methods 0.000 claims description 7
- 230000008859 change Effects 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 6
- 238000007689 inspection Methods 0.000 claims description 5
- 230000004927 fusion Effects 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 238000000844 transformation Methods 0.000 claims description 3
- 230000001960 triggered effect Effects 0.000 claims description 3
- 210000003128 head Anatomy 0.000 description 22
- 238000012423 maintenance Methods 0.000 description 7
- 238000004519 manufacturing process Methods 0.000 description 7
- 210000000887 face Anatomy 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000003203 everyday effect Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000009432 framing Methods 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000009435 building construction Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000002354 daily effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Economics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Human Resources & Organizations (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Bioinformatics & Computational Biology (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a transformer substation personnel dressing intelligent monitoring method based on uncooperative human face recognition, which comprises the following steps of: acquiring video data to obtain a real-time data set; marking the collected image; converting the format of the labeled data into a format required by a yolo4 target detection algorithm, and dividing a data set based on the labeled data; setting up a dressing wearing detection model and a face recognition model; training a dressing wearing detection model and a face recognition model, and correcting parameters; and carrying out dressing detection test on real-time working personnel, detecting face information in real time for the working personnel who fail to dress correctly, and sending an alarm by an alarm service system. The method monitors the personnel in the operation area by developing a detection algorithm for wearing the safety helmet of the transformer substation based on AI image recognition, and immediately alarms if the personnel are detected not to wear the safety helmet, so that a supervisor is reminded to carry out field management, the management and control efficiency of the operation area is prompted, and the safety of the personnel is guaranteed.
Description
Technical Field
The invention relates to the technical field of computer vision monitoring methods, in particular to a transformer substation personnel dressing intelligent monitoring method based on uncooperative human face recognition.
Background
The safety helmet has a protection effect on the head of a human body when being damaged by external force, safety production is a very key part for industries such as electric power and buildings, and the safety helmet must be worn when entering a construction site or an important area of a transformer substation according to relevant regulations of the transformer substation and building construction safety inspection standards. At present, a plurality of cameras are installed on most transformer substation operation sites, and data of the cameras are transmitted to a background for monitoring. However, the video information transmitted back to the monitoring center is generally viewed by human eyes, which is time-consuming and labor-consuming, and easily fatigues human eyes, easily causes careless omission, and also has a wrong judgment condition. The manual inspection is relied on, the supervision difficulty is large, the execution is not in place, and great potential safety hazards exist.
Disclosure of Invention
The technical problem to be solved by the invention is to provide an intelligent transformer substation personnel dressing monitoring method based on uncooperative human face recognition, so that the problems in the background technology are solved, the wearing detection speed and precision of a safety helmet are improved, and the management and control efficiency of operating personnel in a transformer substation park is improved.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows.
The intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition comprises the following steps:
s1, collecting video data to obtain a real-time data set;
s2, labeling the collected image, storing the video data in frames, and labeling and storing the data with pedestrians;
s3, converting the format of the labeled data into a format required by a yolo4 target detection algorithm, and classifying the dressing data and dividing the data set based on the labeled data;
s4, building a dressing wearing detection model and a face recognition model;
s5, training the wearing detection model and the face recognition model, and correcting parameters;
and S6, after the model training is finished, carrying out real-time dressing and wearing detection tests on operators, starting a face recognition module for the operators who cannot dress the garments according to the correct standard, detecting face information in real time, pushing the face information to a corresponding alarm service system, and sending an alarm by the alarm service system.
Further optimize technical scheme, utilize high definition digtal camera or the robot that patrols and examines in the transformer substation to carry out video data's collection to monitoring area and safe operation area.
In step S2, framing the dressing information of the operator in each acquired image by using a rectangular frame, and labeling data according to target detection by using a label tool, which is a label tool of the Labelme image.
Further optimizing the technical scheme, the specific targets framed by the rectangular frame comprise a safety helmet, protective gloves, work clothes and a human head; after a Labelme image marking tool is adopted to mark a target, an xml file is correspondingly generated, the coordinate information and marking information of the target are recorded in the xml file, and the marking types include head, safety helmet, protective gloves, hands, ordinary clothes and safety work clothes.
Further optimizing the technical solution, the step S3 further includes a data expansion step: and (3) carrying out some transformations of brightness, contrast, saturation and hue on the marked picture, rotating the marked picture by a certain angle, and training and expanding the marked picture by using a new Mosaic method.
According to the technical scheme, the wearing and wearing detection model adopts a yolo4 architecture, a backbone network of the wearing and wearing detection model is CSPDarknet53, SPP serves as an additional module of the Neck, PANET serves as a feature fusion module of the Neck, and Yolov3 serves as a Head.
The technical scheme is further optimized, the face recognition model is built based on an ArcFace network, a face tracking algorithm based on Kalman filtering and Hungarian matching is newly added, corresponding id identification information is distributed to the detected face, and face detection and recognition are not carried out after the id is recognized.
Further optimizing the technical scheme, the training process of the dressing wearing detection model comprises the following steps: defining parameters of a dressing and wearing detection model, training by using images in a test sample, and outputting a training log; and calculating the change of the accuracy by using the test sample, and adjusting the network parameters according to the change of the accuracy.
Further optimizing the technical scheme, the parameters of the dressing wearing detection model comprise: category total, rectangle frame size, learning rate, weight decay rate.
Further optimizing the technical scheme, the training of the face recognition model comprises face detection, face alignment and face recognition.
Further optimizing the technical scheme, the process of training the face recognition model comprises the following steps:
pretreatment: after the key points of the face are detected, a cut aligned face is obtained through similarity transformation;
training by adopting a face classifier;
and (3) testing: and extracting embedded features from the output of the face classifier, calculating the cosine distance of the two input features, and then performing face verification and face recognition.
In step S6, the face detection in detecting the face information is triggered to start.
Due to the adoption of the technical scheme, the technical progress of the invention is as follows.
The method monitors the personnel in the operation area by developing a detection algorithm for wearing the safety helmet of the transformer substation based on AI image recognition, and immediately alarms if the personnel are detected not to wear the safety helmet, so that a supervisor is reminded to carry out field management, the management and control efficiency of the operation area is prompted, and the safety of the personnel is guaranteed.
The invention relates to a transformer substation safety helmet wearing detection method based on AI image recognition, which can be well applied to transformer substation safety production, transformer substation park management and the like, processes picture features through a convolutional neural network, extracts human faces and safety wearing features simultaneously, fully considers scene diversity and complexity of target size and form, constructs a target detection algorithm based on yolo4 by using an AI image recognition algorithm, obtains higher accuracy and speed through a large amount of sample data and an extended data training model, can detect safety helmets, gloves, work clothes and human heads in real time, and provides reliable basis for transformer substation safety wearing recognition. The method has high real-time performance, integrates the methods of face detection, face recognition and dressing safety wearing detection, and improves the accuracy of detection.
The method is different from the previous yolo series detection algorithm, optimizes the backbone network, can detect the wearing and wearing image data of various scales, and is extremely accurate in detection of the small target far scene of the transformer substation scene. The invention combines the human head detection frame, the human face tracking ID and the human face identification for verification, accurately positions individuals which do not fit the wearing standard, and improves the detection accuracy and the supervision efficiency. The invention has high real-time performance, integrates human head detection and safety helmet detection, filters non-human head data and improves the detection accuracy. For the personnel who do not wear the safety helmet, the face recognition process is started to confirm the identity of the personnel, the wearing detailed information is put in a warehouse, the personnel can be remotely shout to remind, and the photos are also input into a database for being inquired by management personnel.
On the basis of the wearing and wearing detection of the safety dressing of the traditional operating personnel, the face recognition flow is added, so that potential safety hazards existing in the work of transformer substation safety operation and maintenance management, production maintenance and the like can be further eliminated, and a reward and punishment system can be sequentially established and implemented to the individual. The personnel who do not need artifical off-line to wear the regulation violating behavior in real time supervise and oral early warning, directly discern specific personnel and the warning of shouting of long-range. Meanwhile, the management system can directly inquire the information of the personnel wearing the clothing in an irregular way every day, establish a sound management system and improve the monitoring efficiency.
Drawings
FIG. 1 is a flow chart of the operation of the present invention;
FIG. 2 is a flow chart of monitoring a dressing wear detection model of the present invention;
FIG. 3 is a flow chart of monitoring of the face recognition model of the present invention.
Detailed Description
The invention will be described in further detail below with reference to the figures and specific examples.
The intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition is shown in combination with fig. 1 to 3 and comprises the following steps:
s1, video image acquisition: and acquiring video data of the monitored area and the safe operation area by using a high-definition camera in the transformer substation or a field video acquisition tool such as an inspection robot and the like to obtain a real-time data set.
In the step, the dressing safety wearing data acquisition of the operating personnel is from site videos and pictures of the operating personnel of the transformer substation in a real environment, and the acquired dressing data information is not limited to scale, illumination, style, color and the like. Therefore, training sets and test sets are enriched, and the generalization capability of the detection model is improved.
S2, image annotation: and manually marking the collected image, storing the video data in frames, and marking and storing the data with pedestrians. And manually screening and marking the acquired images to realize head marking and marking of safety helmets, protective gloves and work clothes.
In step S2, framing the dressing information of the operator in each image by a rectangular frame, wherein the targets comprise a safety helmet, protective gloves, work clothes and a human head; and the marking tool adopts a Labelme image marking tool and carries out data marking according to target detection.
After a Labelme image marking tool is adopted to mark a target, an xml file is correspondingly generated, the coordinate information and marking information of the target are recorded in the xml file, and the marking types include head, safety helmet, protective gloves, hands, ordinary clothes and safety work clothes. The type of helmet worn is hat and without helmet is person. And in the labeling process, data cleaning is manually carried out, and pictures with fuzzy targets and difficult labeling are deleted.
S3, data set construction: uniformly converting the xml format of the marked data into a txt format required by a yolo4 target detection algorithm, classifying the dressing data, dividing the data set, and constructing a training set and a verification set based on the marked data.
In step S3, clothing classification and data set division are performed based on the clothing and wearing label data, and the data set is divided into a training set and a test set.
Step S3 further includes a data expansion step: and (3) carrying out some transformations of brightness, contrast, saturation and hue on the marked picture, rotating the marked picture by a certain angle, and training and expanding the marked picture by using a new Mosaic method. More specifically, data expansion is a new data enhancement method for mixing 4 training pictures into one, so that the context information of the image can be enriched, targets outside the context can be detected, and the robustness of the model is enhanced.
S4, model construction: and (5) setting up a dressing and wearing detection model and a face recognition model. The target detection network based on deep learning is built, a yolo4 model is adopted in a core framework, and a target detection model for wearing detection and face detection of wearing helmets, working gloves, protective clothing and the like is built. And constructing a human face tracking and human face recognition model.
The wearing and wearing detection model adopts a yolo4 architecture, a backbone network of the wearing and wearing detection model is CSPDarknet53, SPP serves as an additional module of the Neck, PANET serves as a feature fusion module of the Neck, and Yolov3 serves as a Head.
A face recognition model is built based on an ArcFace network, a face tracking algorithm based on Kalman filtering and Hungary matching is newly added, corresponding id identification information is distributed to a detected face, face detection and recognition are not carried out after the id is recognized, system resource consumption is saved, and face detection and recognition rate is greatly improved. .
S5, model training and prediction: and training the dressing and wearing detection model and the face recognition model, and correcting parameters. And real-time safe dressing, wearing and detecting and face detecting are carried out on the trained model.
The training process of dressing and wearing the detection model comprises the following steps: defining parameters of a dressing and wearing detection model, training by using images in a test sample, and outputting a training log; and calculating the change of the accuracy by using the test sample so as to adjust the network parameters according to the change of the accuracy and obtain the network model meeting the requirement.
Parameters defined by a neural network model for dressing, wearing and human head detection comprise: category total, rectangle frame size, learning rate, weight decay rate.
And adjusting the neural network hyper-parameters in training to enable the loss function of the network to tend to converge in the iterative process, and finally forming a target detection model for detecting the safety helmet. Wherein, the hyper-parameters are specifically set as: the training step is 500500, using a step-size decay learning rate strategy with an initial learning rate of 0.01, multiplying by a factor of 0.1 at steps 400000 and 450000, respectively, momentum decay of 0.949, weight decay of 0.0005, and search learning rate of 0.00261, all architectures using one GPU to perform multi-scale training with a batch size of 64, while the small batch size of 8 or 4 depends on the architecture and GPU memory constraints. Note that the training class is set to 2.
In the training process, 500500 times of training are performed in total, the accuracy of the model is tested once every 1000 times of training, and the model accuracy is steadily improved along with the increase of the training times. The accuracy of the final detection model is 99.9%. The test samples are used to evaluate the network accuracy. After training is finished, carrying out safety helmet wearing detection in real time based on videos, obtaining pictures of a monitoring area, rapidly identifying personnel who do not wear the safety helmet by using a detection model, and carrying out related service alarm.
The training of the face recognition model comprises face detection, face alignment and face recognition. The process of training the face recognition model comprises the following steps:
A. preprocessing (face alignment): detecting the key points of the face by MTCNN, and obtaining a cut aligned face through similarity transformation;
B. training (face classifier): ResNet50+ ArcFace loss.
C. And (3) testing: 512-dimensional embedded features are extracted from the output of the FC1 layer of the face classifier, the cosine distance of the two input features is calculated, and then face verification and face recognition are carried out.
D. Training in the actual code is divided into a resnet model + arc head + softmax loss. The resnet model outputs the features; adding an angle interval between the features and the weights by the arc head, and then outputting a prediction tag to solve ACC by using the output tag; softmax loss evaluates the predicted tag and actual error. After the model training is finished, through testing, the recognition rate of the public data set is respectively as follows: 99.83% on LFW and 98.02% on YTF
S6, real-time monitoring: after the model training is finished, a dressing wearing detection process is started, real-time worker dressing wearing detection tests are carried out, and whether dressing wearing of workers in the monitoring range is standard or not is monitored in real time on line. For the operating personnel who are not wearing clothes correctly and normally, the face recognition module is started, face information is detected in real time, personnel identity recognition is carried out, detailed wearing information is recorded and stored in a warehouse or a screenshot, the face information is pushed to a corresponding warning service system, meanwhile, the detailed wearing information of the personnel is recorded and stored, a service warning is started, and the method plays an important role in safety production management and operation and maintenance of the transformer substation.
The face detection is not continuously started, but is started in a triggering mode, and once the face detection is compared with the local face library, the face recognition program of corresponding non-standard wearing personnel is stopped. Therefore, the consumption of system resources can be saved, and the identification speed is improved. Based on the face tracking ID, the person who is normally and safely worn does not need to start a face recognition process.
The face information of workers in each area in the station is input into the local face library, the face information is set as a white list, if temporary constructors enter the station, a temporary white list library is established, and related faces are input instantly when entering the station, so that whether the working time of the workers is standard and safe to wear and dress is effectively monitored after entering the station.
The invention relates to a transformer substation safety helmet wearing detection method based on AI image recognition, which can be well applied to transformer substation safety production, transformer substation park management and the like, processes picture features through a convolutional neural network, extracts human faces and safety wearing features simultaneously, fully considers scene diversity and complexity of target size and form, constructs a target detection algorithm based on yolo4 by using an AI image recognition algorithm, obtains higher accuracy and speed through a large amount of sample data and an extended data training model, can detect safety helmets, gloves, work clothes and human heads in real time, and provides reliable basis for transformer substation safety wearing recognition. The method has high real-time performance, integrates the methods of face detection, face recognition and dressing safety wearing detection, and improves the accuracy of detection.
Compared with the existing method for detecting the wearing of the safety helmet, the method mainly has the following differences:
1. and (5) innovation of an application scene. The invention is suitable for work such as substation management and daily operation and inspection in a scene. The method is very effective for application scenes such as the access of important monitoring places of the transformer substation, the overhaul of equipment in the transformer substation, the project construction and the like. The detection comprises the detection of wearing of the safety helmet and the detection of the non-fit face.
2. The helmet detects an improvement in the network. The traditional safety helmet detection algorithm is mainly based on a traditional target detection algorithm, or is a deep learning network built based on an SSD or an early yolo series target detection algorithm. The method is customized and developed aiming at the application scenes, is suitable for safety helmet detection algorithms of working scenes such as operation and maintenance management and overhaul of the transformer substation, and combines the latest Yolo4 rapid target detection algorithm to detect the safety helmet and the human head. The target detection network adopts the latest yolo4 architecture, the backbone network of the target detection network is changed from Darknet53 to CSPDarknet53, SPP is used as an additional module of the Neck, PANet is used as a feature fusion module of the Neck, and Yolov3 is used as a Head. The network structure is improved to a certain extent, so that the detection speed and efficiency are greatly improved. Aiming at the same data set researched by the invention, the identification accuracy is improved to 99.95% from 99.8%, and the detection speed is improved to 33fps from 25 fps.
3. And detecting the non-fit type face. A safety helmet target detection algorithm based on Yolo4 is combined with an international leading edge face detection algorithm ArcFace to detect faces, and two deep learning algorithms are combined to assist the transformer substation safety helmet and face detection application scene to fall to the ground. For personnel who do not wear the safety helmet, the face detection is started, face information is detected in real time, and the face information is recorded and stored or captured and pushed to a corresponding warning service system, so that the safety production management and operation and maintenance of the transformer substation are played an important role.
The face detection is not continuously started but is triggered to be started, and once the face detection is compared with the local face library, the face recognition program of the corresponding person who does not wear the safety helmet is stopped. Therefore, the consumption of system resources can be saved, and the identification speed is improved; the person who wears the safety helmet does not need to start the face recognition process.
The face information of workers in each area in the station is recorded in the local face library, the face information is set as a white list, if a temporary constructor enters the station, a temporary white list library is established, and related faces are recorded instantly when the constructor enters the station, so that whether the working time of the constructor is worn by a safety helmet is effectively monitored after the constructor enters the station.
On the basis of the wearing and wearing detection of the safety dressing of the traditional operating personnel, the face recognition flow is added, so that potential safety hazards existing in the work of transformer substation safety operation and maintenance management, production maintenance and the like can be further eliminated, and a reward and punishment system can be sequentially established and implemented to the individual. The personnel who do not need artifical off-line to wear the regulation violating behavior in real time supervise and oral early warning, directly discern specific personnel and the warning of shouting of long-range. Meanwhile, the management system can directly inquire the information of the personnel wearing the clothing in an irregular way every day, establish a sound management system and improve the monitoring efficiency.
The method is different from the previous yolo series detection algorithm, optimizes the backbone network, can detect the wearing and wearing image data of various scales, and is extremely accurate in detection of the small target far scene of the transformer substation scene. The invention combines the human head detection frame, the human face tracking ID and the human face identification for verification, accurately positions individuals which do not fit the wearing standard, and improves the detection accuracy and the supervision efficiency. The invention has high real-time performance, integrates human head detection and safety helmet detection, filters non-human head data and improves the detection accuracy. For the personnel who do not wear the safety helmet, the face recognition process is started to confirm the identity of the personnel, the wearing detailed information is put in a warehouse, the personnel can be remotely shout to remind, and the photos are also input into a database for being inquired by management personnel. The invention combines the head detection frame and the safety helmet detection frame for verification, thereby improving the detection accuracy.
Claims (12)
1. The intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition is characterized by comprising the following steps of:
s1, collecting video data to obtain a real-time data set;
s2, labeling the collected image, storing the video data in frames, and labeling and storing the data with pedestrians;
s3, converting the format of the labeled data into a format required by a yolo4 target detection algorithm, and classifying the dressing data and dividing the data set based on the labeled data;
s4, building a dressing wearing detection model and a face recognition model;
s5, training the wearing detection model and the face recognition model, and correcting parameters;
and S6, after the model training is finished, carrying out real-time dressing and wearing detection tests on operators, starting a face recognition module for the operators who cannot dress the garments according to the correct standard, detecting face information in real time, pushing the face information to a corresponding alarm service system, and sending an alarm by the alarm service system.
2. The intelligent transformer substation personnel dressing monitoring method based on uncooperative human face recognition according to claim 1, characterized in that a high-definition camera or an inspection robot in a transformer substation is used for collecting video data of a monitoring area and a safe operation area.
3. The intelligent substation worker dressing monitoring method based on uncooperative human face recognition according to claim 1, wherein in step S2, worker dressing information in each acquired image is framed with a rectangular frame, and a Labelme image labeling tool is adopted as a labeling tool for data labeling according to target detection.
4. The intelligent transformer substation personnel dressing monitoring method based on uncooperative human face recognition as recited in claim 3, wherein the objects framed with rectangular frames specifically include safety helmets, protective gloves, work clothes, heads of people; after a Labelme image marking tool is adopted to mark a target, an xml file is correspondingly generated, the coordinate information and marking information of the target are recorded in the xml file, and the marking types include head, safety helmet, protective gloves, hands, ordinary clothes and safety work clothes.
5. The transformer substation personnel dressing intelligent monitoring method based on uncooperative face recognition of claim 1, wherein the step S3 further comprises the data expansion step: and (3) carrying out some transformations of brightness, contrast, saturation and hue on the marked picture, rotating the marked picture by a certain angle, and training and expanding the marked picture by using a new Mosaic method.
6. The intelligent substation personnel dressing monitoring method based on uncooperative face recognition of claim 1, wherein the dressing and wearing detection model adopts yolo4 architecture, the backbone network of the dressing and wearing detection model is CSPDarknet53, SPP serves as an additional module of Neck, PANET serves as a feature fusion module of Neck, and Yolov3 serves as Head.
7. The intelligent transformer substation personnel dressing monitoring method based on uncoordinated face recognition is characterized in that a face recognition model is built based on an ArcFace network, a face tracking algorithm based on Kalman filtering and Hungarian matching is newly added, corresponding id identification information is allocated to a detected face, and after the id is recognized, face detection and recognition are not carried out.
8. The intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition according to claim 1, wherein the training process of the dressing and wearing detection model comprises the following steps: defining parameters of a dressing and wearing detection model, training by using images in a test sample, and outputting a training log; and calculating the change of the accuracy by using the test sample, and adjusting the network parameters according to the change of the accuracy.
9. The intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition of claim 8, wherein the parameters of the dressing and wearing detection model comprise: category total, rectangle frame size, learning rate, weight decay rate.
10. The intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition of claim 1, wherein the face recognition model training comprises face detection, face alignment and face recognition.
11. The intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition according to claim 10, wherein the face recognition model training process comprises:
A. pretreatment: after the key points of the face are detected, a cut aligned face is obtained through similarity transformation;
B. training by adopting a face classifier;
C. and (3) testing: and extracting embedded features from the output of the face classifier, calculating the cosine distance of the two input features, and then performing face verification and face recognition.
12. The intelligent transformer substation personnel dressing monitoring method based on uncooperative human face recognition of claim 1, wherein in step S6, human face detection during human face information detection is triggered to start.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111082263.8A CN113807240A (en) | 2021-09-15 | 2021-09-15 | Intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111082263.8A CN113807240A (en) | 2021-09-15 | 2021-09-15 | Intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113807240A true CN113807240A (en) | 2021-12-17 |
Family
ID=78895435
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111082263.8A Pending CN113807240A (en) | 2021-09-15 | 2021-09-15 | Intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113807240A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114445748A (en) * | 2022-01-28 | 2022-05-06 | 深圳市中云慧通科技有限公司 | Video human body feature detection and linkage alarm method and storage medium |
CN115393905A (en) * | 2022-11-01 | 2022-11-25 | 合肥中科类脑智能技术有限公司 | Helmet wearing detection method based on attitude correction |
CN117315575A (en) * | 2023-09-21 | 2023-12-29 | 山东高速济南发展有限公司 | Intelligent cloud helmet with safety protection clothing wearing detection and vital sign monitoring functions |
CN117830686A (en) * | 2023-11-17 | 2024-04-05 | 西南交通大学 | Intelligent management and control platform system and method for wearing personal protective equipment on intelligent construction site |
CN118038589A (en) * | 2024-04-11 | 2024-05-14 | 山东科技大学 | Face recognition system for transformer substation personnel |
CN118097198A (en) * | 2024-04-29 | 2024-05-28 | 三峡高科信息技术有限责任公司 | Automatic dressing compliance management and control system and method based on artificial intelligence |
CN117830686B (en) * | 2023-11-17 | 2024-07-02 | 西南交通大学 | Intelligent management and control platform system and method for wearing personal protective equipment on intelligent construction site |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170124848A1 (en) * | 2015-11-02 | 2017-05-04 | Leauto Intelligent Technology (Beijing) Co. Ltd. | Image-based remote observation and alarm device and method for in-car moving objects |
US20180349703A1 (en) * | 2018-07-27 | 2018-12-06 | Yogesh Rathod | Display virtual objects in the event of receiving of augmented reality scanning or photo of real world object from particular location or within geofence and recognition of real world object |
CN109635697A (en) * | 2018-12-04 | 2019-04-16 | 国网浙江省电力有限公司电力科学研究院 | Electric operating personnel safety dressing detection method based on YOLOv3 target detection |
CN111383429A (en) * | 2020-03-04 | 2020-07-07 | 西安咏圣达电子科技有限公司 | Method, system, device and storage medium for detecting dress of workers in construction site |
CN111401314A (en) * | 2020-04-10 | 2020-07-10 | 上海东普信息科技有限公司 | Dressing information detection method, device, equipment and storage medium |
CN111860318A (en) * | 2020-07-20 | 2020-10-30 | 杭州品茗安控信息技术股份有限公司 | Construction site pedestrian loitering detection method, device, equipment and storage medium |
CN112464884A (en) * | 2020-12-11 | 2021-03-09 | 武汉工程大学 | ADAS infrared night vision method and system |
CN112584094A (en) * | 2020-11-23 | 2021-03-30 | 国网浙江省电力有限公司临海市供电公司 | Transformer substation safety supervision system based on intelligent sensing |
CN112613375A (en) * | 2020-12-16 | 2021-04-06 | 中国人寿财产保险股份有限公司 | Tire damage detection and identification method and device |
-
2021
- 2021-09-15 CN CN202111082263.8A patent/CN113807240A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170124848A1 (en) * | 2015-11-02 | 2017-05-04 | Leauto Intelligent Technology (Beijing) Co. Ltd. | Image-based remote observation and alarm device and method for in-car moving objects |
US20180349703A1 (en) * | 2018-07-27 | 2018-12-06 | Yogesh Rathod | Display virtual objects in the event of receiving of augmented reality scanning or photo of real world object from particular location or within geofence and recognition of real world object |
CN109635697A (en) * | 2018-12-04 | 2019-04-16 | 国网浙江省电力有限公司电力科学研究院 | Electric operating personnel safety dressing detection method based on YOLOv3 target detection |
CN111383429A (en) * | 2020-03-04 | 2020-07-07 | 西安咏圣达电子科技有限公司 | Method, system, device and storage medium for detecting dress of workers in construction site |
CN111401314A (en) * | 2020-04-10 | 2020-07-10 | 上海东普信息科技有限公司 | Dressing information detection method, device, equipment and storage medium |
CN111860318A (en) * | 2020-07-20 | 2020-10-30 | 杭州品茗安控信息技术股份有限公司 | Construction site pedestrian loitering detection method, device, equipment and storage medium |
CN112584094A (en) * | 2020-11-23 | 2021-03-30 | 国网浙江省电力有限公司临海市供电公司 | Transformer substation safety supervision system based on intelligent sensing |
CN112464884A (en) * | 2020-12-11 | 2021-03-09 | 武汉工程大学 | ADAS infrared night vision method and system |
CN112613375A (en) * | 2020-12-16 | 2021-04-06 | 中国人寿财产保险股份有限公司 | Tire damage detection and identification method and device |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114445748A (en) * | 2022-01-28 | 2022-05-06 | 深圳市中云慧通科技有限公司 | Video human body feature detection and linkage alarm method and storage medium |
CN115393905A (en) * | 2022-11-01 | 2022-11-25 | 合肥中科类脑智能技术有限公司 | Helmet wearing detection method based on attitude correction |
CN117315575A (en) * | 2023-09-21 | 2023-12-29 | 山东高速济南发展有限公司 | Intelligent cloud helmet with safety protection clothing wearing detection and vital sign monitoring functions |
CN117315575B (en) * | 2023-09-21 | 2024-02-20 | 山东高速济南发展有限公司 | Intelligent cloud helmet with safety protection clothing wearing detection and vital sign monitoring functions |
CN117830686A (en) * | 2023-11-17 | 2024-04-05 | 西南交通大学 | Intelligent management and control platform system and method for wearing personal protective equipment on intelligent construction site |
CN117830686B (en) * | 2023-11-17 | 2024-07-02 | 西南交通大学 | Intelligent management and control platform system and method for wearing personal protective equipment on intelligent construction site |
CN118038589A (en) * | 2024-04-11 | 2024-05-14 | 山东科技大学 | Face recognition system for transformer substation personnel |
CN118097198A (en) * | 2024-04-29 | 2024-05-28 | 三峡高科信息技术有限责任公司 | Automatic dressing compliance management and control system and method based on artificial intelligence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110738127B (en) | Helmet identification method based on unsupervised deep learning neural network algorithm | |
CN110826538B (en) | Abnormal off-duty identification system for electric power business hall | |
CN113807240A (en) | Intelligent transformer substation personnel dressing monitoring method based on uncooperative face recognition | |
CN111598040B (en) | Construction worker identity recognition and safety helmet wearing detection method and system | |
CN111191586B (en) | Method and system for inspecting wearing condition of safety helmet of personnel in construction site | |
CN112396658B (en) | Indoor personnel positioning method and system based on video | |
CN112785798A (en) | Behavior analysis method for construction project constructors of electric power substation engineering | |
CN112766050B (en) | Dressing and operation checking method, computer device and storage medium | |
CN113903081A (en) | Visual identification artificial intelligence alarm method and device for images of hydraulic power plant | |
CN113516076A (en) | Improved lightweight YOLO v4 safety protection detection method based on attention mechanism | |
CN110543866A (en) | Safety management system and method for capital construction engineering constructors | |
CN113743256A (en) | Construction site safety intelligent early warning method and device | |
CN112613449A (en) | Safety helmet wearing detection and identification method and system based on video face image | |
CN114926778A (en) | Safety helmet and personnel identity recognition system under production environment | |
CN111177468A (en) | Laboratory personnel unsafe behavior safety inspection method based on machine vision | |
CN115223249A (en) | Quick analysis and identification method for unsafe behaviors of underground personnel based on machine vision | |
CN116580349A (en) | Method for detecting wearing of safety helmet based on yolov7 building site | |
CN111723725A (en) | Multi-dimensional analysis system based on video AI | |
CN117079203A (en) | Road construction monitoring method, system and storage medium | |
CN115063921B (en) | Building site intelligent gate system and building method | |
CN114429677A (en) | Coal mine scene operation behavior safety identification and assessment method and system | |
CN116030405A (en) | Pedestrian target safety protection tool wearing monitoring method based on artificial intelligence | |
CN110969127A (en) | Vision-based safety helmet identification method | |
CN117830686B (en) | Intelligent management and control platform system and method for wearing personal protective equipment on intelligent construction site | |
Chen et al. | A novel system for automated proper use identification of personal protective equipment in decommissioning site of fukushima daiichi nuclear power station |
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 |