CN106778609A - A kind of electric power construction field personnel uniform wears recognition methods - Google Patents
A kind of electric power construction field personnel uniform wears recognition methods Download PDFInfo
- Publication number
- CN106778609A CN106778609A CN201611160441.3A CN201611160441A CN106778609A CN 106778609 A CN106778609 A CN 106778609A CN 201611160441 A CN201611160441 A CN 201611160441A CN 106778609 A CN106778609 A CN 106778609A
- Authority
- CN
- China
- Prior art keywords
- personnel
- detection
- detection module
- picture
- safety cap
- 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
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
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
- G06V40/173—Classification, e.g. identification face re-identification, e.g. recognising unknown faces across different face tracks
Abstract
Recognition methods is worn the invention discloses a kind of electric power construction field personnel uniform, administrative staff's detection is entered first, head, the detection of metastomium and personnel towards detection are carried out when personnel are detected, and detect whether safe wearing cap；Personnel's safe wearing cap, and be oriented front and back, then carry out safety cap text detection in head zone；As personnel towards side, then armband detection is carried out with the juncture area of trunk in head；Finally according to safety cap and the testing result of armband, judge whether into personnel be staff；When non-working person is detected as, alarm is carried out.Present invention employs the method for HOG feature combination SVM classifiers, carry out texture analysis to safety cap word and armband has the remarkable advantage such as rotational invariance and gray scale consistency using LBP operators, and the robustness under illumination condition is good.
Description
Technical field
The present invention relates to one kind uniform recognition methods, particularly a kind of electric power construction field personnel uniform dress identification side
Method.
Background technology
Due to the high risk of power system facility, therefore in regions such as power distribution rooms, not only need to electric power facility
Maintenance maintenance process carries out the video monitoring of whole process, and whether need the personnel that detection enters is professional staff.And
Armband is subdued as electric power and indicated, can determine it by the way that whether testing staff wears the text information on armband combination safety cap
Whether it is staff.If arranging professional to be checked, substantial amounts of manpower and materials can be consumed, it is therefore desirable to one kind identification
The video monitoring and detection method of uniform.
Chinese patent application（Notification number：CN104408459A）It is to be proposed to make in a kind of electric apparatus monitoring by Liu Furong etc.
Image-recognizing method, its step is：Realize regularly patrolling and examining by fixed infrared camera, gained image is beamed back
Control room is analyzed.Feature first to power equipment is extracted, and is carried out image preprocessing and is split, merges, it is determined that electric
The fault zone of gas equipment, carries out the detection identification of target area failure.
Chinese patent application（Notification number：CN201869307U）It is that a kind of image for power system is proposed by Li Lixin
Monitoring system, the frequency image monitoring system is by video camera, picture control main frame, interchanger, protocol converter, optical transmitter and receiver and distribution prison
Control mechanism.Video camera is connected by video line with picture control main frame, and picture control main frame is connected by netting twine with interchanger,
Interchanger is connected by netting twine with protocol converter, and protocol converter is connected by 2M lines with optical transmitter and receiver, and optical transmitter and receiver passes through
Optical cable is connected with distribution optical transmitter and receiver, and distribution optical transmitter and receiver is connected by 2M lines with distribution protocol converter, and distribution protocol converter leads to
Netting twine is crossed to be connected with distributed monitoring machine.When the system works, the data of the main frame in multiple monitoring place are believed by interchanger
Breath is sent to protocol converter, and 2M information is converted that information into by protocol converter, using the transmission characteristic of optical fiber by image
Monitoring information transmission is realized being monitored the power supply unit in same place in different location to relatively remote.
Chinese patent application（Notification number：CN104933819A）It is to propose one kind based on recognition of face by yellow, Wang Xinying
With the alarm and alarm method of landmark identification, autotracking video camera collection face image data, and the face figure that will be gathered
As data is activation to the first computer；First computer is to the face in the face image data and designated person's data bank of collection
View data carries out aspect ratio pair, and the result according to aspect ratio pair judges whether the face image data of collection is designated person's
Face image data, and will determine that result is sent to second computer；It is on the clothes of Mark Detection device detection designated person
It is no to there is setting mark, and the result of Mark Detection is sent to second computer；Knot of the second computer according to aspect ratio pair
The output control warning device of fruit and Mark Detection sends alarm.
Therefore detection identification technology at present in domestic and international power transmission facility video monitoring system is mainly for tamper-proof, anti-
Robber, Condition Detection etc., mostly using sensor by the way of detection control circuit is combined.In frequency image monitoring system,
Hardware requirement is higher, system complex, and does not provide complete including personnel, safety cap, direction, safety cap word, armband etc.
The detection scheme in face.In terms of uniform wears identification, also not for the detection and knowledge of staff's armband and safety cap word
Other method.
The content of the invention
Recognition methods is worn it is an object of the invention to a kind of electric power construction field personnel uniform, solution cannot be detected at present
The problem of identification staff's armband and safety cap word.
A kind of electric power construction field personnel uniform wears recognition methods, and it is concretely comprised the following steps：
The first step builds power construction personnel uniform identifying system
Monitor video read in module, personnel's detection module, face detection module, towards detection module, trunk detection module, safety
Cap detection module, safety cap text detection module and armband detection module are operated in power construction personnel uniform identifying system.
Monitor video reads in module to be used to read the data that camera acquisition is arrived, and personnel's detection module is used to detect video camera
Whether there are personnel to enter in the scene for photographing, face detection module is used to detect the face part for entering personnel, towards detection
Module is used for direction of the testing staff relative to video camera, and trunk detection module is used for the torso portion of testing staff, safety cap
Detection module be used for detect enter personnel whether safe wearing cap, safety cap text detection module be used for detect enter personnel peace
Word on full cap, armband detection module is used to detect the armband for entering personnel.
Second step monitor video reads in module and reads the data that camera acquisition is arrived
Monitor video reads in module Real-time Collection field data, and the personnel detection module of being transferred to carries out subsequent treatment.
Whether there are personnel to enter in the scene that 3rd step personnel detection module detection video camera is photographed
Sample is the experimental data for classifier training, and the positive sample for personnel's detection is the picture comprising personnel targets,
Various forms and dressing including personnel；Negative sample includes any image in addition to positive sample, positive sample and negative sample
Size is unified.The negative sample used in personnel's detection is the background picture in operation field scene.
Personnel's detection module carries out HOG feature extractions：
HOG is gradient orientation histogram feature, is a kind of descriptor of local features, calculates regional area gradient direction straight
Side's figure composition target signature describes the edge of target.HOG feature extraction and calculation processes are：Gaussian smoothing is carried out to input picture,
Multiple cell factories are divided the image into, the gradient orientation histogram of each cell factory is counted, then by multiple cell factory groups
Into a block, contrast normalization is carried out in block, the set of histograms of multiple blocks is finally synthesized HOG features.
Personnel's detection module carries out SVM classifier training：
Obtain after HOG features, be trained using SVM classifier.SVM classifier is based on structural risk minimization theory in spy
Levy construction optimum segmentation hyperplane in space so that grader obtains global optimization.During training, input vector is normalized, made
The learning outcome of grader is stored with model.
Personnel's detection module carries out target detection：
Target detection identification process sets the yardstick and the dimension gain factor of minimum first using multiscale target detection, right
Altimetric image to be checked carries out the scaling of corresponding scale, and traversing graph picture is repeated several times, and exports testing result.
The detection of 4th step face detection module enters the face part of personnel
Positive sample for Face datection is the picture comprising face, including different personnel, different shooting direction；Negative sample includes
Any image in addition to positive sample, the negative sample size for Face datection is more than or equal to positive sample.Adopted in Face datection
Negative sample picture is the background picture in operation field scene.
Face detection module carries out Harr feature calculations：
Harr features are the rectangular characteristics of piece image, and rectangular characteristic of different shapes reflects different characteristics of image.In face
In the same position of non-face picture, with the pixel in white portion and value size that the pixel of black region obtains is subtracted not
Together, so as to distinguish face and non-face.
Face detection module carries out Adaboost classifier trainings and Face datection：
Adaboost is a kind of algorithm of multinomial level, and multiple strong classifiers are exactly linked together and grasped by cascade classifier
Make, and each strong classifier is made up of the weighting of several Weak Classifiers.
5th step safety cap detection module testing staff whether safe wearing cap
Positive sample for safety cap detection is the picture comprising safety cap, including different shooting direction；Negative sample include except
Any image outside positive sample, positive sample and negative sample size are unified.Because the detection of safety cap is in head given zone
Domain is carried out, therefore the negative sample picture for using is the head image without safe wearing cap.
Safety cap detection module carries out HOG feature extractions first, after obtaining HOG features, then is carried out using SVM classifier
Training, finally carries out target detection using multiscale target detection method.
The torso portion of the 6th step trunk detection module testing staff
Positive sample for trunk detection is the picture comprising personnel's trunk, including different personnel, different directions, different dressings；
Negative sample image includes any image in addition to positive sample, positive and negative sample-size size unification.Used in trunk detection
Negative sample picture be operation field scene background image and monitoring image in remove personnel's torso portion other images.
Trunk detection module carries out HOG feature extractions first, after obtaining HOG features, then is instructed using SVM classifier
Practice, target detection is finally carried out using multiscale target detection method.
Direction of 7th step towards detection module testing staff relative to camera
When front classifier training is carried out, for being just to the picture of camera, bag comprising personnel towards the positive sample of detection
Include personnel targets, head, trunk；Negative sample is not include any image of positive sample, positive and negative sample-size size unification.Front
The negative sample picture used in detection be operation field scene background image and monitoring image in remove personnel just in face of taking the photograph
As other images of head.
When side classifier training is carried out, for being the figure comprising personnel side to camera towards the positive sample of detection
Piece, including personnel targets, head, trunk；Negative sample is not include any image of positive sample, positive and negative sample-size size system
One.In the detection of side the negative sample picture that uses be operation field scene background image and monitoring image in remove personnel side
In face of other images of camera.
When back side classifier training is carried out, for being the figure comprising personnel back to camera towards the positive sample of detection
Piece, including personnel targets, head, trunk；Negative sample is not include any image of positive sample, positive and negative sample-size size system
One.In the detection of the back side negative sample picture that uses be operation field scene background image and monitoring image in remove personnel and carry on the back
In face of other images of camera.
HOG feature extractions are carried out first towards detection module, after obtaining HOG features, then are instructed using SVM classifier
Practice, target detection is finally carried out using multiscale target detection method.
8th step armband detection module is used for the armband of testing staff
When personnel are oriented side, armband detection module carries out armband detection.Positive sample for armband detection is comprising armband
Picture, including different colours, different shooting direction；Negative sample includes any image in addition to positive sample, armband detection
The middle negative sample picture for using removes other images of armband for personnel's torso portion.
Armband word is noteworthy characterized by texture, and armband detection module carries out texture analysis using LBP features.LBP is office
Portion's binary pattern, original LBP operator definitions be in the window of 3*3, the pixel value with window center as threshold value, respectively with phase
The pixel value of 8 adjacent pixels is compared, when the pixel value of certain surrounding pixel is more than central point pixel value, Current ambient
The value of pixel is marked as 1, is otherwise 0.8 points so in 3*3 fields produce the unsigned number of 8bit, i.e. window
LBP values, reflect the texture information in region.It is 0111100 that 3*3 windows produce 8 bits, and corresponding decimal number is
The LBP values of 124, the i.e. window are 124.
The LBP operators that the present invention is used are LBP invariable rotary patterns.LBP operator treatment is carried out to piece image, at each
Pixel all obtains a LBP coding, but this coding characteristic is closely related with positional information, it is impossible to is directly used in and sentences
Do not analyze.Therefore a width picture is divided into some sub-regions, in the statistics Nogata of the built-in vertical LBP features of every sub-regions
Figure, is then normalized to statistic histogram, and the statistic histogram of the every sub-regions that will finally obtain is attached
As a characteristic vector, as LBP texture feature vectors of view picture figure, for carrying out Classification and Identification.
Obtain after LBP features, armband detection module is trained using SVM classifier first, then using multiple dimensioned
Object detection method carries out target detection.
The word that the detection of 9th step safety cap text detection module enters on the safety cap of personnel
When judgement personnel are oriented front or the back side, safety cap text detection is carried out.For the positive sample of safety cap text detection
Originally it is the picture comprising safety cap word, including different shooting direction；Negative sample includes any image in addition to positive sample.
The negative sample picture used in safety cap text detection is the safety cap picture without word.
Safety cap word is noteworthy characterized by texture, and safety cap text detection module carries out texture analysis using LBP features.
Obtain after LBP features, safety cap text detection module is trained using SVM classifier first, then using multiple dimensioned mesh
Mark detection method carries out target detection.
Personnel's detection, safety cap and armband identifying processing are completed in the server, and the result for obtaining is by man-machine interactive platform
Command centre shows that staff is checked by command centre and controls whole monitoring system.
The present invention can be in video monitoring information identify into the personnel in monitor area, and by personnel
Armband and safety cap word be identified, judge whether it is staff.If being detected as non-working person, alarmed
Prompting.The method that HOG feature combination SVM classifiers are employed in the present invention；Wherein HOG features can to image geometry and light
Deformation keeps good consistency, and the trickle limb action of pedestrian can be ignored without influence Detection results.Face is examined
Survey using the method for Harr feature combination Adaboost graders, Harr features are the rectangular characteristic of piece image, Adaboost
It is a kind of algorithm of multinomial level, substantially increases detection speed and accuracy rate.Texture point is carried out to safety cap word and armband
Analysis has the remarkable advantage such as rotational invariance and gray scale consistency using LBP operators, and the robustness under illumination condition is good；Calculate
Speed is fast, because it can be obtained by being compared operation in small neighbourhood so that the analysis chart under complicated real-time conditions
As being possibly realized；Because LBP operators are a kind of methods of printenv, its distribution need not be carried out in advance in application process
Assuming that.
Brief description of the drawings
A kind of electric power construction field personnel uniforms of Fig. 1 wear the power construction personnel uniform identifying system described in recognition methods
The structural representation of embodiment one；
The power construction personnel uniform identifying system that a kind of electric power construction field personnel uniforms of Fig. 2 wear described in recognition methods is implemented
The structural representation of example two；
A kind of electric power construction field personnel uniforms of Fig. 3 wear the power construction personnel uniform identifying system stream described in recognition methods
Cheng Tu；
A kind of electric power construction field personnel uniforms of Fig. 4 wear the HOG characteristic extraction procedure flow charts described in recognition methods；
A kind of electric power construction field personnel uniforms of Fig. 5 wear the classifier training process described in recognition methods；
A kind of electric power construction field personnel uniforms of Fig. 6 wear the Harr feature sets described in recognition methods；
A kind of electric power construction field personnel uniforms of Fig. 7 wear the Adaboost sorter models described in recognition methods；
A kind of electric power construction field personnel uniforms of Fig. 8 wear the LBP operators described in recognition methods；
A kind of electric power construction field personnel uniforms of Fig. 9 wear the LBP characteristic extraction procedures described in recognition methods.
Specific embodiment
The first step builds power construction personnel uniform identifying system
Embodiment one is as shown in Figure 1：Power construction personnel uniform identifying system includes：Video camera 1, information transfer net 3, long-range prison
Control server 2.Video camera 1 is connected with remote monitoring server 2 by information transfer net 3, and monitor video reads in module 28, personnel
Detection module 21, face detection module 22, towards detection module 23, trunk detection module 24, safety cap detection module 25, safety
Cap text detection module 26 and armband detection module 27 are operated on remote monitoring server 2.The live number of the collection of video camera 1
According to, remote monitoring server 2 is sent to by information transfer net 3, control information is returned to video camera by remote monitoring server 2
1。
Embodiment two is as shown in Figure 2：Power construction personnel uniform identifying system operates in the DSP of intelligent network camera 4
In processor 41, identifying system includes：Monitor video reads in module 28, personnel's detection module 21, face detection module 22, direction
Detection module 23, trunk detection module 24, safety cap detection module 25, safety cap text detection module 26 and armband detection module
27。
Monitor video reads in module 28 to be used to read the data that video camera 1 is collected, and personnel's detection module 21 is used to detect
Whether there are personnel to enter in the scene that video camera 1 is photographed, face detection module 22 is used to detect the face part for entering personnel,
It is used for direction of the testing staff relative to video camera 1 towards detection module 23, trunk detection module 24 is used for the body of testing staff
Stem portion, safety cap detection module 25 be used to detect enter personnel whether safe wearing cap, safety cap text detection module 26 uses
Word on the safety cap that detection enters personnel, armband detection module 27 is used to detect the armband for entering personnel.
Second step monitor video reads in module 28 and reads the data that video camera 1 is collected
Monitor video reads in the Real-time Collection field data of module 28, and being transferred to personnel's detection module 21 carries out subsequent treatment.
Whether there are personnel to enter in the scene that the detection video camera 1 of 3rd step personnel detection module 21 is photographed
Sample is the experimental data for classifier training, and the positive sample for personnel's detection is the picture comprising personnel targets,
Various forms and dressing including personnel；Negative sample includes any image in addition to positive sample, positive sample and negative sample
Size is unified.The negative sample used in personnel's detection is the background picture in operation field scene.
Personnel's detection module 21 carries out HOG feature extractions：
HOG is gradient orientation histogram feature, is a kind of descriptor of local features, calculates regional area gradient direction straight
Side's figure composition target signature describes the edge of target.HOG feature extraction and calculation processes are as shown in Figure 4：First, input picture is entered
Row Gaussian smoothing.Then, each pixel to image calculates Grad, obtains the gradient vector in x directions and y directions；Per 8*8
Individual pixel constitutes a cell factory, calculates the gradient orientation histogram of cell factory.Then, 2*2 cell factory is constituted
One block, contrast normalization is carried out in block by the gradient orientation histogram of each cell factory.Finally, because of a training window
Mouth size is 64*128, and it is scanned by 8*8 pixel, the combination of eigenvectors of each block is risen and is used as picture in its entirety
HOG characteristic vectors.Personnel's detection module 21 carries out SVM classifier training：
Obtain after HOG features, be trained using SVM classifier：
SVM classifier is based on structural risk minimization theory construction optimum segmentation hyperplane in feature space so that classify
To global optimization.The idiographic flow of training is as shown in Figure 5.For different graders, using appropriately sized training window,
And set the various parameters such as stopping criterion for iteration, grader species, SVM types, relaxation factor, the ginseng for using in the present invention
Number has：Classifier type is C_SVC, linear kernel function, relaxation factor C=0.01.
During training, positive negative sample is unified into size respectively, positive sample HOG eigenmatrixes, label 1 is special with negative sample HOG
Matrix, the input of label -1 SVM classifier is levied to be trained.The XML file the inside obtained after the completion of Linear SVM training, there is three
Parameter：Support vector arrays, alpha arrays, rho floating numbers.Alpha matrixes are multiplied with support vector,
In one element rho of last addition of the column vector, grader learning outcome is obtained.
Train the grader learning outcome for obtaining may be undesirable for the first time, then with the learning outcome of first time to all
Negative sample detected, the target that will be detected unify size preservation, add negative sample in form new negative sample.Using just
Sample and new negative sample re -training, can obtain relatively good learning outcome.
Finally, target detection is carried out using multiscale target detection method：
Input image to be identified, multiscale target detection is carried out using corresponding grader learning outcome to the image.Many chis
Degree target detection the step of be：The yardstick and the dimension gain factor of minimum are set first, and treating detection image carries out corresponding scale
Scaling, be repeated several times traversing graph picture, situation of the penalty factor to same target by repeated detection out is set and is suppressed, most
After export testing result.
The detection of 4th step face detection module 22 enters the face part of personnel
Positive sample for Face datection is the picture comprising face, including different personnel, different shooting direction；Negative sample includes
Any image in addition to positive sample, the negative sample size for Face datection is more than or equal to positive sample.Adopted in Face datection
Negative sample picture is the background picture in operation field scene.
Face detection module 22 carries out Harr feature calculations：
Harr features are the rectangular characteristics of piece image.In same position of the face with non-face picture, same square is used
The Haar features that shape operator is obtained have bigger difference, so as to distinguish face and non-face.
Haar feature sets as shown in Figure 6 calculate the Haar characteristic values of positive negative sample respectively, determine the optimal of Weak Classifier
Threshold value.Weak Classifier is weighted and is combined into a strong classifier, and multiple strong classifiers are cascaded up.Strong classifier pair
The differentiation degree of accuracy of negative sample is very high, once finding that target negative sample just selects to abandon, only positive sample can just be sent to next
Individual strong classifier is checked again, substantially increases the accuracy rate and detection speed of detection.
Face detection module 22 carries out Adaboost classifier trainings and Face datection：
Adaboost is a kind of iterative algorithm.When initial, the weight of all training samples is equal, is trained under this sample distribution
One Weak Classifier.N-th（N=1,2,3 ... T, T is iterations）In secondary iteration, the weight of sample is by (n-1)th iteration
Result depending on.Weight is adjusted after each iteration, the sample for being classified mistake will obtain weight higher, obtain a new sample
This distribution.To this, new sample distribution is trained, and a Weak Classifier is obtained again.Finally, these Weak Classifiers are pressed
Get up according to certain weighted superposition, just obtain final strong classifier.As shown in Figure 7.
The testing staff of 5th step safety cap detection module 25 whether safe wearing cap
Positive sample for safety cap detection is the picture comprising safety cap, including different shooting direction；Negative sample include except
Any image outside positive sample, positive sample and negative sample size are unified.Because the detection of safety cap is in head given zone
Domain is carried out, therefore the negative sample picture for using is the head image without safe wearing cap.
Safety cap detection module 25 carries out HOG feature extractions first, after obtaining HOG features, then is entered using SVM classifier
Row training, finally carries out target detection using multiscale target detection method.
The torso portion of the testing staff of the 6th step trunk detection module 24
Positive sample for trunk detection is the picture comprising personnel's trunk, including different personnel, different directions, different dressings；
Negative sample image includes any image in addition to positive sample, positive and negative sample-size size unification.Used in trunk detection
Negative sample picture be operation field scene background image and monitoring image in remove personnel's torso portion other images.
Trunk detection module 24 carries out HOG feature extractions first, after obtaining HOG features, then is carried out using SVM classifier
Training, finally carries out target detection using multiscale target detection method.
Direction of 7th step towards the testing staff of detection module 23 relative to camera
When front classifier training is carried out, for being just to the picture of camera, bag comprising personnel towards the positive sample of detection
Include personnel targets, head, trunk；Negative sample is not include any image of positive sample, positive and negative sample-size size unification.Front
The negative sample picture used in detection be operation field scene background image and monitoring image in remove personnel just in face of taking the photograph
As other images of head.
When side classifier training is carried out, for being the figure comprising personnel side to camera towards the positive sample of detection
Piece, including personnel targets, head, trunk；Negative sample is not include any image of positive sample, positive and negative sample-size size system
One.In the detection of side the negative sample picture that uses be operation field scene background image and monitoring image in remove personnel side
In face of other images of camera.
When back side classifier training is carried out, for being the figure comprising personnel back to camera towards the positive sample of detection
Piece, including personnel targets, head, trunk；Negative sample is not include any image of positive sample, positive and negative sample-size size system
One.In the detection of the back side negative sample picture that uses be operation field scene background image and monitoring image in remove personnel and carry on the back
In face of other images of camera.
HOG feature extractions are carried out first towards detection module 23, after obtaining HOG features, then are carried out using SVM classifier
Training, finally carries out target detection using multiscale target detection method.
8th step armband detection module 27 is used for the armband of testing staff
When personnel are oriented side, armband detection module 27 carries out armband detection.Positive sample for armband detection is comprising sleeve
The picture of chapter, including different colours, different shooting direction；Negative sample includes any image in addition to positive sample, armband inspection
The negative sample picture used in survey removes other images of armband for personnel's torso portion.
Armband is noteworthy characterized by texture, and LBP features can be very good to carry out texture analysis.LBP is local binary patterns,
Original LBP operator definitions be in the window of 3*3, the pixel value with window center as threshold value, respectively with 8 adjacent pixels
Pixel value be compared, when certain surrounding pixel pixel value be more than central point pixel value when, the value of Current ambient pixel
1 is marked as, is otherwise 0.8 points so in 3*3 fields produce the unsigned number of 8bit, i.e. the LBP values of window, reflect
The texture information in region.As shown in figure 8, it is 0111100 that the 3*3 windows produce 8 bits, corresponding decimal number is
The LBP values of 124, the i.e. window are 124.
The LBP operators that the present invention is used are LBP invariable rotary patterns.LBP operator treatment is carried out to piece image, at each
Pixel all obtains a LBP coding, but this coding characteristic is closely related with positional information, it is impossible to is directly used in and sentences
Do not analyze.Therefore the LBP feature extraction flows that the present invention is used are as shown in Figure 9.Calculating process be roughly divided into it is following some：
1. detection window is divided into 16 × 16 zonule first；
2. for a pixel in each zonule, 8 adjacent gray values of pixel are compared with it, if the pixel
Point gray value is more than center pixel value, then be marked as 1, is otherwise 0.So, 8 bits can be produced in 3*3 neighborhoods, is turned
It is changed to the LBP values that decimal number obtains the window center pixel；
3. the histogram of each zonule is then calculated, i.e. each numeral（That is LBP values of each pixel）The frequency of appearance；So
The histogram is normalized afterwards.
The LBP characteristic vectors of the statistic histogram connection composition view picture figure of each zonule that 4. will finally obtain.
Obtain after LBP features, armband detection module 27 is trained using SVM classifier first, then using many chis
Degree object detection method carries out target detection.
The word that the detection of 9th step safety cap text detection module 26 enters on the safety cap of personnel
When judgement personnel are oriented front or the back side, safety cap text detection is carried out.For the positive sample of safety cap text detection
Originally it is the picture comprising safety cap word, including different shooting direction；Negative sample includes any image in addition to positive sample.
The negative sample picture used in safety cap text detection is the safety cap picture without word.
Safety cap word is noteworthy characterized by texture, and safety cap text detection module 26 carries out texture point using LBP features
Analysis.Obtain after LBP features, safety cap text detection module 26 is trained using SVM classifier first, then using many chis
Degree object detection method carries out target detection.
Personnel's detection, safety cap and armband identifying processing are completed in the server, and the result for obtaining is by man-machine interactive platform
Command centre shows that staff is checked by command centre and controls whole monitoring system.
Claims (3)
1. a kind of electric power construction field personnel uniform wears recognition methods, it is characterised in that concretely comprise the following steps：
The first step builds power construction personnel uniform identifying system
Monitor video reads in module（28）, personnel's detection module（21）, face detection module（22）, towards detection module（23）、
Trunk detection module（24）, safety cap detection module（25）, safety cap text detection module（26）With armband detection module（27）
Operate in power construction personnel uniform identifying system in；
Monitor video reads in module（28）For reading video camera（1）The data for collecting, personnel's detection module（21）For examining
Survey video camera（1）Whether there are personnel to enter in the scene for photographing, face detection module（22）Enter the people of personnel for detecting
Face part, towards detection module（23）For testing staff relative to video camera（1）Direction, trunk detection module（24）For
The torso portion of testing staff, safety cap detection module（25）For detect enter personnel whether safe wearing cap, safety cap text
Word detection module（26）For detecting the word on the safety cap for entering personnel, armband detection module（27）Enter people for detecting
The armband of member；
Second step monitor video reads in module（28）Read video camera（1）The data for collecting
Monitor video reads in module（28）Real-time Collection field data, is transferred to personnel's detection module（21）Carry out subsequent treatment；
3rd step personnel's detection module（21）Detection video camera（1）Whether there are personnel to enter in the scene for photographing
Sample is the experimental data for classifier training, and the positive sample for personnel's detection is the picture comprising personnel targets,
Various forms and dressing including personnel；Negative sample includes any image in addition to positive sample, positive sample and negative sample
Size is unified；The negative sample used in personnel's detection is the background picture in operation field scene；
Personnel's detection module（21）Carry out HOG feature extractions：
HOG is gradient orientation histogram feature, is a kind of descriptor of local features, calculates regional area gradient direction straight
Side's figure composition target signature describes the edge of target；HOG feature extraction and calculation processes are：Gaussian smoothing is carried out to input picture,
Multiple cell factories are divided the image into, the gradient orientation histogram of each cell factory is counted, then by multiple cell factory groups
Into a block, contrast normalization is carried out in block, the set of histograms of multiple blocks is finally synthesized HOG features；
Personnel's detection module（21）Carry out SVM classifier training：
Obtain after HOG features, be trained using SVM classifier；SVM classifier is based on structural risk minimization theory in spy
Levy construction optimum segmentation hyperplane in space so that grader obtains global optimization；During training, input vector is normalized, made
The learning outcome of grader is stored with model；
Personnel's detection module（21）Carry out target detection：
Target detection identification process sets the yardstick and the dimension gain factor of minimum first using multiscale target detection, right
Altimetric image to be checked carries out the scaling of corresponding scale, and traversing graph picture is repeated several times, and exports testing result；
4th step face detection module（22）Detection enters the face part of personnel
Positive sample for Face datection is the picture comprising face, including different personnel, different shooting direction；Negative sample includes
Any image in addition to positive sample, the negative sample size for Face datection is more than or equal to positive sample；Adopted in Face datection
Negative sample picture is the background picture in operation field scene；
Face detection module（22）Carry out Harr feature calculations：
Harr features are the rectangular characteristics of piece image, and rectangular characteristic of different shapes reflects different characteristics of image；In face
In the same position of non-face picture, with the pixel in white portion and value size that the pixel of black region obtains is subtracted not
Together, so as to distinguish face and non-face；
Face detection module（22）Carry out Adaboost classifier trainings and Face datection：
Adaboost is a kind of algorithm of multinomial level, and multiple strong classifiers are exactly linked together and grasped by cascade classifier
Make, and each strong classifier is made up of the weighting of several Weak Classifiers；
5th step safety cap detection module（25）Testing staff whether safe wearing cap
Positive sample for safety cap detection is the picture comprising safety cap, including different shooting direction；Negative sample include except
Any image outside positive sample, positive sample and negative sample size are unified；Because the detection of safety cap is in head given zone
Domain is carried out, therefore the negative sample picture for using is the head image without safe wearing cap；
Safety cap detection module（25）HOG feature extractions are carried out first, after obtaining HOG features, then are carried out using SVM classifier
Training, finally carries out target detection using multiscale target detection method；
6th step trunk detection module（24）The torso portion of testing staff
Positive sample for trunk detection is the picture comprising personnel's trunk, including different personnel, different directions, different dressings；
Negative sample image includes any image in addition to positive sample, positive and negative sample-size size unification；Used in trunk detection
Negative sample picture be operation field scene background image and monitoring image in remove personnel's torso portion other images；
Trunk detection module（24）HOG feature extractions are carried out first, after obtaining HOG features, then are instructed using SVM classifier
Practice, target detection is finally carried out using multiscale target detection method；
7th step is towards detection module（23）Direction of the testing staff relative to camera
When front classifier training is carried out, for being just to the picture of camera, bag comprising personnel towards the positive sample of detection
Include personnel targets, head, trunk；Negative sample is not include any image of positive sample, positive and negative sample-size size unification；Front
The negative sample picture used in detection be operation field scene background image and monitoring image in remove personnel just in face of taking the photograph
As other images of head；
When side classifier training is carried out, for being the picture comprising personnel side to camera, bag towards the positive sample of detection
Include personnel targets, head, trunk；Negative sample is not include any image of positive sample, positive and negative sample-size size unification；Side
The negative sample picture used in detection be operation field scene background image and monitoring image in remove personnel side to taking the photograph
As other images of head；
When back side classifier training is carried out, for being the picture comprising personnel back to camera, bag towards the positive sample of detection
Include personnel targets, head, trunk；Negative sample is not include any image of positive sample, positive and negative sample-size size unification；The back side
The negative sample picture used in detection be operation field scene background image and monitoring image in remove the personnel back side to taking the photograph
As other images of head；
Towards detection module（23）HOG feature extractions are carried out first, after obtaining HOG features, then are instructed using SVM classifier
Practice, target detection is finally carried out using multiscale target detection method；
8th step armband detection module（27）For the armband of testing staff
When personnel are oriented side, armband detection module（27）Carry out armband detection；For armband detection positive sample be comprising
The picture of armband, including different colours, different shooting direction；Negative sample includes any image in addition to positive sample, armband
The negative sample picture used in detection removes other images of armband for personnel's torso portion；
Armband word is noteworthy characterized by texture, armband detection module（27）Texture analysis is carried out using LBP features；LBP is office
Portion's binary pattern, original LBP operator definitions be in the window of 3*3, the pixel value with window center as threshold value, respectively with phase
The pixel value of 8 adjacent pixels is compared, when the pixel value of certain surrounding pixel is more than central point pixel value, Current ambient
The value of pixel is marked as 1, is otherwise 0；8 points so in 3*3 fields produce the unsigned number of 8bit, i.e. window
LBP values, reflect the texture information in region；It is 0111100 that 3*3 windows produce 8 bits, and corresponding decimal number is
The LBP values of 124, the i.e. window are 124；
The LBP operators that the present invention is used are LBP invariable rotary patterns；LBP operator treatment is carried out to piece image, in each pixel
Point all obtains a LBP coding, but this coding characteristic is closely related with positional information, it is impossible to be directly used in differentiation point
Analysis；Therefore a width picture is divided into some sub-regions, in the statistic histogram of the built-in vertical LBP features of every sub-regions, so
Statistic histogram is normalized afterwards, the statistic histogram of the every sub-regions that will finally obtain is attached as one
The LBP texture feature vectors of individual characteristic vector, as view picture figure, for carrying out Classification and Identification；
After obtaining LBP features, armband detection module（27）It is trained using SVM classifier first, then using multiple dimensioned
Object detection method carries out target detection；
9th step safety cap text detection module（26）The word that detection enters on the safety cap of personnel
When judgement personnel are oriented front or the back side, safety cap text detection is carried out；For the positive sample of safety cap text detection
Originally it is the picture comprising safety cap word, including different shooting direction；Negative sample includes any image in addition to positive sample；
The negative sample picture used in safety cap text detection is the safety cap picture without word；
Safety cap word is noteworthy characterized by texture, safety cap text detection module（26）Texture analysis is carried out using LBP features；
After obtaining LBP features, safety cap text detection module（26）It is trained using SVM classifier first, then using many chis
Degree object detection method carries out target detection；
Personnel's detection, safety cap and armband identifying processing are completed in the server, and the result for obtaining is commanded by man-machine interactive platform
Center shows that staff is checked by command centre and controls whole monitoring system.
2. electric power construction field personnel uniform according to claim 1 wears recognition methods, it is characterised in that：Described electricity
Power workmen uniform identifying system includes：Video camera（1）, information transfer net（3）And remote monitoring server（2）；Video camera
（1）With remote monitoring server（2）By information transfer net（3）Connection, monitor video reads in module（28）, personnel's detection module
（21）, face detection module（22）, towards detection module（23）, trunk detection module（24）, safety cap detection module（25）, peace
Full cap text detection module（26）With armband detection module（27）Operate in remote monitoring server（2）On；Video camera（1）Collection
Field data, by information transfer net（3）It is sent to remote monitoring server（2）, remote monitoring server（2）Control is believed
Breath returns to video camera（1）.
3. electric power construction field personnel uniform according to claim 1 wears recognition methods, it is characterised in that：Power construction
Personnel's uniform identifying system operates in intelligent network camera（4）Dsp processor（41）In, described identifying system includes：
Monitor video reads in module（28）, personnel's detection module（21）, face detection module（22）, towards detection module（23）, trunk
Detection module（24）, safety cap detection module（25）, safety cap text detection module（26）With armband detection module（27）.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611160441.3A CN106778609A (en) | 2016-12-15 | 2016-12-15 | A kind of electric power construction field personnel uniform wears recognition methods |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611160441.3A CN106778609A (en) | 2016-12-15 | 2016-12-15 | A kind of electric power construction field personnel uniform wears recognition methods |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106778609A true CN106778609A (en) | 2017-05-31 |
Family
ID=58887590
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611160441.3A Pending CN106778609A (en) | 2016-12-15 | 2016-12-15 | A kind of electric power construction field personnel uniform wears recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106778609A (en) |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107609544A (en) * | 2017-10-27 | 2018-01-19 | 华润电力技术研究院有限公司 | A kind of detection method and device |
| CN108052900A (en) * | 2017-12-12 | 2018-05-18 | 成都睿码科技有限责任公司 | A kind of method by monitor video automatic decision dressing specification |
| CN108491830A (en) * | 2018-04-23 | 2018-09-04 | 济南浪潮高新科技投资发展有限公司 | A kind of job site personnel uniform dress knowledge method for distinguishing based on deep learning |
| CN108565963A (en) * | 2017-12-29 | 2018-09-21 | 云南玉溪中汇电力设备有限责任公司 | Power distribution network intelligent operation maintenance system |
| CN108805777A (en) * | 2018-04-27 | 2018-11-13 | 安徽继远软件有限公司 | A kind of on-site personnel security control method based on trusted identity certification |
| CN109344679A (en) * | 2018-07-25 | 2019-02-15 | 深圳云天励飞技术有限公司 | Building site monitoring method, device and readable storage medium storing program for executing based on image analysis |
| CN109359805A (en) * | 2018-08-28 | 2019-02-19 | 安徽继远软件有限公司 | A kind of substation's electric operation personnel safety management-control method |
| CN109376676A (en) * | 2018-11-01 | 2019-02-22 | 哈尔滨工业大学 | Highway engineering site operation personnel safety method for early warning based on unmanned aerial vehicle platform |
| CN109389729A (en) * | 2018-12-03 | 2019-02-26 | 广东电网有限责任公司 | A kind of more scene recognition of face monitoring systems of smart grid |
| CN109389360A (en) * | 2018-08-28 | 2019-02-26 | 安徽继远软件有限公司 | A kind of personnel safety management-control method based on multiple authentication manipulation permission |
| CN109460719A (en) * | 2018-10-24 | 2019-03-12 | 四川阿泰因机器人智能装备有限公司 | A kind of electric operating safety recognizing method |
| CN109522845A (en) * | 2018-11-19 | 2019-03-26 | 国网四川省电力公司电力科学研究院 | Distribution transformer based on intelligent robot tests safety work measure of supervision |
| CN109558783A (en) * | 2018-09-20 | 2019-04-02 | 中建科技有限公司深圳分公司 | A kind of violation detection method, system and equipment for construction site |
| CN109672863A (en) * | 2018-12-24 | 2019-04-23 | 海安常州大学高新技术研发中心 | A kind of construction personnel's safety equipment intelligent monitoring method based on image recognition |
| CN109800665A (en) * | 2018-12-28 | 2019-05-24 | 广州粤建三和软件股份有限公司 | A kind of Human bodys' response method, system and storage medium |
| CN109902595A (en) * | 2019-01-31 | 2019-06-18 | 天津大学 | Using station personnel's method for detecting abnormality of multi-cam |
| CN109919182A (en) * | 2019-01-24 | 2019-06-21 | 国网浙江省电力有限公司电力科学研究院 | A kind of terminal side electric power safety operation image-recognizing method |
| CN109977843A (en) * | 2019-03-21 | 2019-07-05 | 重庆工程职业技术学院 | A kind of colliery scene human behavior monitoring and intelligent identifying system violating the regulations |
| CN110009857A (en) * | 2019-05-14 | 2019-07-12 | 沈阳风驰软件股份有限公司 | A kind of detection of platform one-meter line and geofence control system |
| CN110458075A (en) * | 2019-08-05 | 2019-11-15 | 北京泰豪信息科技有限公司 | Detection method, storage medium, detection device and the detection system that safety cap is worn |
| CN110580427A (en) * | 2018-06-08 | 2019-12-17 | 杭州海康威视数字技术股份有限公司 | face detection method, device and equipment |
| CN110827434A (en) * | 2019-09-23 | 2020-02-21 | 重庆特斯联智慧科技股份有限公司 | Community security patrol recording system and method for grid target identification |
| CN111582209A (en) * | 2020-05-14 | 2020-08-25 | 重庆邮电大学 | Abnormal behavior supervision method for construction personnel in capital construction site |
| CN113837138A (en) * | 2021-09-30 | 2021-12-24 | 重庆紫光华山智安科技有限公司 | Dressing monitoring method, system, medium and electronic terminal |
| CN114724287A (en) * | 2021-01-06 | 2022-07-08 | 中国石油天然气股份有限公司 | On-duty system and on-duty method of oil and gas station |
| CN116110081A (en) * | 2023-04-12 | 2023-05-12 | 齐鲁工业大学(山东省科学院) | Detection method and system for wearing safety helmet based on deep learning |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103714431A (en) * | 2013-12-26 | 2014-04-09 | 苏州清研微视电子科技有限公司 | Airport people identity authentication management system based on face recognition |
| CN103745226A (en) * | 2013-12-31 | 2014-04-23 | 国家电网公司 | Dressing safety detection method for worker on working site of electric power facility |
| CN104933819A (en) * | 2015-06-17 | 2015-09-23 | 福建永易信息科技有限公司 | Alarm device based on face recognition and sign recognition and alarm method |
-
2016
- 2016-12-15 CN CN201611160441.3A patent/CN106778609A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103714431A (en) * | 2013-12-26 | 2014-04-09 | 苏州清研微视电子科技有限公司 | Airport people identity authentication management system based on face recognition |
| CN103745226A (en) * | 2013-12-31 | 2014-04-23 | 国家电网公司 | Dressing safety detection method for worker on working site of electric power facility |
| CN104933819A (en) * | 2015-06-17 | 2015-09-23 | 福建永易信息科技有限公司 | Alarm device based on face recognition and sign recognition and alarm method |
Non-Patent Citations (2)
| Title |
|---|
| 杨杰 等: "《视频目标检测和跟踪及其应用》", 31 August 2012, 上海交通大学出版社 * |
| 王一丁 等: "《数字图像处理》", 31 August 2015, 西安电子科技大学 * |
Cited By (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107609544A (en) * | 2017-10-27 | 2018-01-19 | 华润电力技术研究院有限公司 | A kind of detection method and device |
| CN108052900A (en) * | 2017-12-12 | 2018-05-18 | 成都睿码科技有限责任公司 | A kind of method by monitor video automatic decision dressing specification |
| CN108565963A (en) * | 2017-12-29 | 2018-09-21 | 云南玉溪中汇电力设备有限责任公司 | Power distribution network intelligent operation maintenance system |
| CN108565963B (en) * | 2017-12-29 | 2020-08-14 | 云南玉溪中汇电力设备有限责任公司 | Intelligent operation maintenance system for power distribution network |
| CN108491830A (en) * | 2018-04-23 | 2018-09-04 | 济南浪潮高新科技投资发展有限公司 | A kind of job site personnel uniform dress knowledge method for distinguishing based on deep learning |
| CN108805777A (en) * | 2018-04-27 | 2018-11-13 | 安徽继远软件有限公司 | A kind of on-site personnel security control method based on trusted identity certification |
| CN110580427A (en) * | 2018-06-08 | 2019-12-17 | 杭州海康威视数字技术股份有限公司 | face detection method, device and equipment |
| CN109344679A (en) * | 2018-07-25 | 2019-02-15 | 深圳云天励飞技术有限公司 | Building site monitoring method, device and readable storage medium storing program for executing based on image analysis |
| WO2020019673A1 (en) * | 2018-07-25 | 2020-01-30 | 深圳云天励飞技术有限公司 | Construction site monitoring method and device based on image analysis, and readable storage medium |
| CN109359805A (en) * | 2018-08-28 | 2019-02-19 | 安徽继远软件有限公司 | A kind of substation's electric operation personnel safety management-control method |
| CN109389360A (en) * | 2018-08-28 | 2019-02-26 | 安徽继远软件有限公司 | A kind of personnel safety management-control method based on multiple authentication manipulation permission |
| CN109558783A (en) * | 2018-09-20 | 2019-04-02 | 中建科技有限公司深圳分公司 | A kind of violation detection method, system and equipment for construction site |
| CN109460719A (en) * | 2018-10-24 | 2019-03-12 | 四川阿泰因机器人智能装备有限公司 | A kind of electric operating safety recognizing method |
| CN109376676A (en) * | 2018-11-01 | 2019-02-22 | 哈尔滨工业大学 | Highway engineering site operation personnel safety method for early warning based on unmanned aerial vehicle platform |
| CN109522845A (en) * | 2018-11-19 | 2019-03-26 | 国网四川省电力公司电力科学研究院 | Distribution transformer based on intelligent robot tests safety work measure of supervision |
| CN109389729A (en) * | 2018-12-03 | 2019-02-26 | 广东电网有限责任公司 | A kind of more scene recognition of face monitoring systems of smart grid |
| CN109672863A (en) * | 2018-12-24 | 2019-04-23 | 海安常州大学高新技术研发中心 | A kind of construction personnel's safety equipment intelligent monitoring method based on image recognition |
| CN109800665A (en) * | 2018-12-28 | 2019-05-24 | 广州粤建三和软件股份有限公司 | A kind of Human bodys' response method, system and storage medium |
| CN109919182A (en) * | 2019-01-24 | 2019-06-21 | 国网浙江省电力有限公司电力科学研究院 | A kind of terminal side electric power safety operation image-recognizing method |
| CN109919182B (en) * | 2019-01-24 | 2021-10-22 | 国网浙江省电力有限公司电力科学研究院 | Terminal side electric power safety operation image identification method |
| CN109902595B (en) * | 2019-01-31 | 2023-04-21 | 天津大学 | Transformer substation maintenance personnel abnormality detection method adopting multiple cameras |
| CN109902595A (en) * | 2019-01-31 | 2019-06-18 | 天津大学 | Using station personnel's method for detecting abnormality of multi-cam |
| CN109977843A (en) * | 2019-03-21 | 2019-07-05 | 重庆工程职业技术学院 | A kind of colliery scene human behavior monitoring and intelligent identifying system violating the regulations |
| CN109977843B (en) * | 2019-03-21 | 2021-03-30 | 重庆工程职业技术学院 | Coal mine field personnel behavior monitoring and violation intelligent identification system |
| CN110009857A (en) * | 2019-05-14 | 2019-07-12 | 沈阳风驰软件股份有限公司 | A kind of detection of platform one-meter line and geofence control system |
| CN110458075A (en) * | 2019-08-05 | 2019-11-15 | 北京泰豪信息科技有限公司 | Detection method, storage medium, detection device and the detection system that safety cap is worn |
| CN110458075B (en) * | 2019-08-05 | 2023-08-25 | 北京泰豪信息科技有限公司 | Method, storage medium, device and system for detecting wearing of safety helmet |
| CN110827434A (en) * | 2019-09-23 | 2020-02-21 | 重庆特斯联智慧科技股份有限公司 | Community security patrol recording system and method for grid target identification |
| CN111582209A (en) * | 2020-05-14 | 2020-08-25 | 重庆邮电大学 | Abnormal behavior supervision method for construction personnel in capital construction site |
| CN114724287A (en) * | 2021-01-06 | 2022-07-08 | 中国石油天然气股份有限公司 | On-duty system and on-duty method of oil and gas station |
| CN113837138A (en) * | 2021-09-30 | 2021-12-24 | 重庆紫光华山智安科技有限公司 | Dressing monitoring method, system, medium and electronic terminal |
| CN113837138B (en) * | 2021-09-30 | 2023-08-29 | 重庆紫光华山智安科技有限公司 | Dressing monitoring method, dressing monitoring system, dressing monitoring medium and electronic terminal |
| CN116110081A (en) * | 2023-04-12 | 2023-05-12 | 齐鲁工业大学(山东省科学院) | Detection method and system for wearing safety helmet based on deep learning |
| CN116110081B (en) * | 2023-04-12 | 2023-06-30 | 齐鲁工业大学(山东省科学院) | Detection method and system for wearing safety helmet based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106778609A (en) | A kind of electric power construction field personnel uniform wears recognition methods | |
| CN104166841B (en) | The quick detection recognition methods of pedestrian or vehicle is specified in a kind of video surveillance network | |
| CN103942577B (en) | Based on the personal identification method for establishing sample database and composite character certainly in video monitoring | |
| CN106128022B (en) | A kind of wisdom gold eyeball identification violent action alarm method | |
| CN105701467B (en) | A kind of more people's abnormal behaviour recognition methods based on human figure feature | |
| CN109819208A (en) | A kind of dense population security monitoring management method based on artificial intelligence dynamic monitoring | |
| CN109558810B (en) | Target person identification method based on part segmentation and fusion | |
| CN108052900A (en) | A kind of method by monitor video automatic decision dressing specification | |
| CN108288033A (en) | A kind of safety cap detection method merging multiple features based on random fern | |
| Chen et al. | Detection of safety helmet wearing based on improved faster R-CNN | |
| CN106919921B (en) | Gait recognition method and system combining subspace learning and tensor neural network | |
| CN109672863A (en) | A kind of construction personnel's safety equipment intelligent monitoring method based on image recognition | |
| CN112396658A (en) | Indoor personnel positioning method and positioning system based on video | |
| CN109460719A (en) | A kind of electric operating safety recognizing method | |
| CN110728252B (en) | Face detection method applied to regional personnel motion trail monitoring | |
| CN108268832A (en) | Electric operating monitoring method, device, storage medium and computer equipment | |
| CN110414400A (en) | A kind of construction site safety cap wearing automatic testing method and system | |
| CN111860471B (en) | Work clothes wearing identification method and system based on feature retrieval | |
| Raghukumar et al. | Comparison of machine learning algorithms for detection of medicinal plants | |
| CN112541393A (en) | Transformer substation personnel detection method and device based on deep learning | |
| CN109086657B (en) | A kind of ear detection method, system and model based on machine learning | |
| Yen et al. | Abnormal event detection using HOSF | |
| CN107316024B (en) | Perimeter alarm algorithm based on deep learning | |
| Bhatia et al. | Face detection using fuzzy logic and skin color segmentation in images | |
| CN107886060A (en) | Pedestrian's automatic detection and tracking based on video |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170531 |
|
| RJ01 | Rejection of invention patent application after publication |