CN110008831A - A kind of Intellectualized monitoring emerging system based on computer vision analysis - Google Patents
A kind of Intellectualized monitoring emerging system based on computer vision analysis Download PDFInfo
- Publication number
- CN110008831A CN110008831A CN201910134732.2A CN201910134732A CN110008831A CN 110008831 A CN110008831 A CN 110008831A CN 201910134732 A CN201910134732 A CN 201910134732A CN 110008831 A CN110008831 A CN 110008831A
- Authority
- CN
- China
- Prior art keywords
- edge
- pixel
- human body
- moving object
- frame
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of Intellectualized monitoring emerging system based on computer vision analysis, including video monitoring system, face identification system, video structure analyzing system, Multisensor video fusion system.Video monitoring system is responsible for acquiring video information and plant area's important area temperature information;Face identification system is responsible for identifying plant area worker;Video structure analyzing system can determine whether whether safe wearing cap, vehicle enter unauthorized area, whether, employee consistent with its working region carries package to employee work clothes to employee;Multisensor video fusion system is responsible for being managed collectively the various information of video monitoring system, face identification system and video structure analyzing system.The system can be monitored and analyze to plant area in real time, find plant area's abnormal conditions and early warning in time, to provide safeguard for plant area's normal operation.
Description
Technical field
The present invention relates to factory's running state monitoring systems, and in particular to a kind of intelligence based on computer vision analysis
Monitoring fusion system belongs to security monitoring field.
Background technique
In the production run of factory, safety is placed on primary always.However, being deposited in the daily production of factory
In many security risks, such as key area temperature is excessively high, the non-safe wearing cap of employee enters working region, vehicle drives into it
Object is taken into forbidding the important area for taking object, employee to go to the wrong way working region etc. in the region of lack of competence entrance, employee, seriously threatens
The normal operation of factory.
Summary of the invention
To solve the above problems, the present invention provides a kind of, the Intellectualized monitoring based on computer vision analysis merges system
System, including video monitoring system, face identification system, video structure analyzing system, Multisensor video fusion system.
The video monitoring system is made of front end camera, thermal infrared imager and video monitoring platform;Front end camera
It is responsible for acquisition video information;Thermal infrared imager is responsible for acquiring plant area's important area temperature information;Video monitoring platform receiving front-end
The video information of camera acquisition and plant area's important area temperature information of thermal infrared imager acquisition, are shown, simultaneously will
Video information sends video structure analyzing system to, and sends video information and plant area's important area temperature information to video
Emerging system.
The face identification system is made of face snap device, face facial recognition modules and worker's database;People
Face captures employee's image of device acquisition disengaging each key area of plant area, and sends face facial recognition modules to;Worker's number
The identity information and its face feature of each employee are stored according to library;Face facial recognition modules receive the member of face snap device acquisition
Work image carries out recognition of face, determines its identity information by the comparison with worker's database, and recognition result is transmitted
To Multisensor video fusion system.
Employee's face feature extraction process is as follows in the face identification system: the face figure of the acquisition each employee of plant area
Picture, image size are M × N, it is assumed that share K employees, then K facial images can be obtained;Grey level is carried out to K facial images
Property conversion process and 3 × 3 median filter process, obtain K face grayscale images;By every one-row pixels of every face grayscale image
Gray value be connected, then every face grayscale image may make up D=M × N-dimensional row vector, for i-th face grayscale image,
Remember its row vector x constitutedi, and calculate average faceThe K row vector that K face grayscale images are constituted into
Row arrangement constitutes the matrix of K × D dimension, and carries out Karhunen-Loeve transformation to the matrix, obtains eigenface space: w=(u1, u2, u3...,
up), in formula: p be setting dimensionality reduction after dimension;The face feature for calculating each employee, for j-th of employee, face feature
ΩjCalculation formula are as follows: Ωj=wT(xj-Ψ)。
It is as follows to carry out face recognition process for face facial recognition modules in the face identification system: acquisition face snap device
The facial image for the disengaging each key area of plant area captured carries out grey linear transformation to facial image and obtains face gray scale
Figure, the gray value of the every one-row pixels of face grayscale image is connected, row vector y is obtained;Calculate the face feature Ω of the facial imagej
=wT(xj- Ψ), in formula: w is the eigenface space acquired in claim 2: w=(u1, u2, u3..., up), Ψ wants for right
Seek the average face acquired in 2;Calculate ΩyWith the Euclidean distance of the face feature of worker's database purchase, select and ΩyEurope
For formula apart from nearest face feature, the corresponding worker of the face feature is face recognition result.
The video structure analyzing system by safety cap identification module, Car license recognition module, work clothes identification module with
And object identification module composition is taken, the video information of video monitoring system acquisition is received, safety cap identification module judgement person is passed through
Work whether safe wearing cap, whether enter by Car license recognition judgment module vehicle unauthorized area, identified by work clothes
Module judge employee work clothes it is whether consistent with its working region, judge whether employee carries package by taking object identification module,
And send analysis result to Multisensor video fusion system.
The safety cap identification module judges that whether the process of employee's safe wearing cap as follows:
The video information for reading the acquisition of detection zone camera in real time, extracts former frame and present frame, to former frame and works as
Previous frame carries out gray processing processing, obtains the grayscale image of former frame and present frame;Using frame differential method to former frame and present frame
Grayscale image handled, obtain moving object figure;It constructs one or more moving object estimation rectangle frames and extracts moving object
All independent parts mutually in body figure, it is assumed that have K independent parts mutually in figure, then K moving object estimation can be obtained
Rectangle frame;The image for extracting each moving object estimation rectangle frame corresponding part in the grayscale image and cromogram of present frame, obtains
Cromogram is estimated to K moving object estimation grayscale images and K moving objects;K moving objects are estimated using Canny operator
Grayscale image carries out edge detection one by one, obtains K moving object edge graphs, and a fortune is constructed in every moving object edge graph
Animal body rectangle frame includes all edges in the figure, obtains K moving object rectangle frame;According to human body Aspect Ratio, confirmation fortune
Human body rectangle frame in animal body rectangle frame, it is assumed that judge there be k human body rectangle frame, extract each human body rectangle frame and transporting
Animal body estimates the image of corresponding part in cromogram, obtains k human body cromograms;For each human body rectangle frame, according to
The head of people in upper half of human body feature location human body rectangle frame estimates region according to head ratio positioning security cap, obtains k
Safety cap estimates region;The image of each safety cap estimation region corresponding part in its corresponding human body cromogram is extracted,
K safety cap region color figures are obtained, are that foundation one by one judges safety cap region color figure with color;Specific steps are such as
Under:
Step 1, former frame and present frame are extracted, gray processing is carried out to former frame and present frame picture using weighted mean method
Processing, the i.e. R to each pixel, tri- components of G, B distribute different weights, specific formula is as follows: f (x, y)=0.3R (x,
Y)+0.59G (x, y)+0.11B (x, y), in formula, R (x, y), G (x, y), B (x, y) respectively refer to the R for the pixel that coordinate is (x, y),
G, B component, f (x, y) refer to the gray value after the pixel conversion that coordinate is (x, y);
Step 2, it is handled using grayscale image of the frame differential method to former frame and present frame, obtains moving object figure,
Process is as follows: the gray value of two field pictures respective pixel being subtracted each other, and takes its absolute value, obtains difference image Dn(x, y)
=| fn(x, y)-fn-1(x, y) |, wherein fn(x, y) refers to gray value of the coordinate for the pixel of (x, y), f in present framen-1(x, y) refers to
Coordinate is the gray value of the pixel of (x, y) in former frame;Threshold value T is setaIf Dn(x, y) > T, Rn=255, if Dn(x,
Y) < T, Rn=0, obtain moving object figure R;R is binary image, and wherein white area represents moving object region;
Step 3, it constructs K moving object estimation rectangle frame and extracts all independent parts mutually in moving object figure,
And the location information of each moving object estimation rectangle frame is obtained, it is stored in matrix L O, detailed process is as follows:
Step1: reference axis is established as origin in the lower left corner of the moving object figure R obtained using step 2, and moving object figure R exists
First quartile;If straight line x=h, h initial value is 0;
Step2: straight line x=h is moved to right into a unit, judges the pixel value of each pixel on straight line x=h;
Step3: if the pixel value of each pixel is 0 on straight line x=h, Step2 is returned to;If deposited on straight line x=h
In the pixel that pixel value is not 0, remember that h at this time is xmin;
Step4: straight line x=h is moved to right into a unit, judges the pixel value of each pixel on straight line x=h;
Step5: if there are the pixel that pixel value is non-zero on straight line x=h, remember that uppermost non-zero pixel is corresponding on straight line
Ordinate value and the corresponding ordinate value of bottom non-zero pixel, and be stored in array Y, return to Step4;If straight line x=
The pixel value of the upper each pixel of h is 0, remembers that h at this time is xmax;
Step6: being compared the size of the element in array Y, obtains the maximum value y in YmaxWith minimum value ymin;With
Straight line y=ymaxAs top edge, straight line y=yminAs lower edge, straight line x=xminAs left edge, straight line x=xmaxAs
Region rectangle frame is estimated in right hand edge tectonic movement;Array Y is reset, and remembers LOk=(ymax, ymin, xmin, xmax, k) and it is stored in matrix
In LO, wherein k initial value is 1, and k value adds 1 after each tectonic movement estimation region rectangle frame;
Step7: repeating Step2 to Step6, until straight line x=h terminates when being moved to binaryzation picture rightmost;
Step 4, each moving object estimation rectangle frame corresponding part in the grayscale image and cromogram of present frame is extracted
Image obtains K moving object estimation grayscale image G1..., GKCromogram P is estimated with K moving objects1..., PK;It extracted
Journey is as follows: estimating rectangle frame for i-th of moving object, its corresponding location information is read from the matrix L O that step 3 obtains
LOi=(ymax, ymin, xmin, xmax, i), confirm that square is estimated in the moving object using top edge, lower edge, left edge and right hand edge
The specific location of shape frame is to extract;
Step 5, grayscale image G is estimated to moving object one by one using Canny operator1To GKEdge detection is carried out, obtains K
Edge pixels all in every edge graph are compared by edge graph with the pixel of corresponding position in moving object figure, delete side
With the unequal edge pixel of pixel pixel value of corresponding position in moving object figure in edge pixel, final moving object is obtained
Edge graph;It includes all edges in the figure that a moving object rectangle frame is constructed in every moving object edge graph, obtains K
Moving object rectangle frame M1..., MK;
Tectonic movement object rectangle frame process is as follows: firstly, traversing from top to bottom to the moving object edge graph, obtaining
Obtain the leftmost edge pixel X of every a linele={ xl1, xl2, xl3, xl4, xl5..., xlmAnd rightmost edge pixel cross
Coordinate Xre={ xr1, xr2, xr3, xr4, xr5..., xrm, wherein xliRepresent the abscissa of the leftmost edge pixel of the i-th row, xri
The abscissa of the edge pixel of the i-th row rightmost is represented, m represents the moving object edge graph line number;Then, from left to right to this
Moving object edge graph is traversed, and the uppermost edge pixel Y of each column is obtainedhe={ yh1, yh2, yh3, yh4, yh5...,
yhnAnd nethermost edge pixel ordinate Yle={ yl1, yl2, yl3, yl4, yl5..., yln, wherein yhiRepresent the i-th column
The ordinate of uppermost edge pixel, yliThe ordinate of the nethermost edge pixel of the i-th column is represented, n represents the moving object
Edge graph columns;Finally, with XleIn least member as left edge, XreMiddle greatest member is as right hand edge, YheIn maximum
Element is as top edge, YleIn least member as lower edge, tectonic movement object rectangle frame;
Step 6, according to human body proportion to moving object rectangle frame M1To MKJudged, obtain human body rectangle frame therein,
Assuming that judging there be k human body rectangle frame, each human body rectangle frame is extracted in its corresponding moving object estimation cromogram
The image of corresponding part obtains k human body cromograms;
Judge whether it is that human body rectangle frame process is as follows: the note a length of L of moving object rectangle frame rectangle, width W, λ=W/L,
Threshold alpha, β are set;If α < λ < β, is regarded as human body rectangle frame, is otherwise considered as other objects;
Step 7, for each human body rectangle frame, according to the head of people in upper half of human body feature location human body rectangle frame
Region is estimated according to head ratio positioning security cap by portion, obtains k safety cap estimation region;
Head position fixing process is as follows: firstly, cutting to human body rectangle frame, head zone is obtained, it is specific as follows: to take people
Body rectangle frame top edge is as head zone top edge, human body rectangle frame top edge and lower edge 1/3 close to the position of top edge
As head zone lower edge, left edge and right hand edge of the left edge and right hand edge of human body rectangle frame as head zone;Its
It is secondary, the abscissa that head zone obtains every a line leftmost edge point and rightmost edges point, X are traversed from top to bottomld={ xl1, xl2,
xl3, xl4, xl5..., xlgAnd Xrd={ xr1, xr2, xr3, xr4, xr5..., xrg, wherein xliIt is horizontal to represent the i-th row Far Left marginal point
Coordinate, xriThe i-th row rightmost marginal point coordinate is represented, g represents head zone line number;Then, Far Left edge is calculated to most right
The distance D=X at side edgerd-Xld, obtain D={ D1, D2, D3, D4, D5..., Dg, calculate element latter and previous item in D
Poor DΔ={ D2-D1, D3-D2, D4-D3, D5-D4..., Dg-Dg-1};Finally find out DΔFrom the 11st element into a last element
The corresponding line number p of maximum element, is updated to pth row for head zone lower edge;
Construction safety cap rectangle frame process is as follows: finding out X respectivelyldIn from xlpTo xlgThe smallest element xmlAnd XrdIn from
xrpTo xrgGreatest member xmr, with x=xmlAs accurate head zone left edge, with x=xmrAs right hand edge, with header area
Domain top edge and lower edge obtain accurate head zone as accurate head top edge and lower edge;Then, to accurate header area
The further cutting in domain is specific as follows to obtain safety cap region: on safety cap region, left and right edge and accurate head zone one
Cause, using accurate head zone top edge and lower edge 1/3 close to top edge position as safety cap region lower edge;
Step 8, safety cap rectangle frame is extracted in the image of human body cromogram corresponding position, obtains a safety cap region color
Figure is that foundation one by one judges safety cap region color figure with color;
Its deterministic process is as follows: safety cap region color figure is switched into HSV image, due to general factory safety cap have it is red
Color, blue, white and yellow, so taking red, blue, white and yellow as judgment criteria, the corresponding color of these four colors
Adjust H, saturation degree S, the value range of lightness V as follows:
It is red: H:0-10 and 156-180;S:43-255;V:46-255
Blue: H:100-124;S:43-255;V:46-255
White: H:0-180;S:0-30;V:221-255
Yellow: H:26-34;S:43-255;V:46-255
Classify to the color of each pixel, process is as follows: if the H of pixel to be sorted, S, V value is all satisfied certain face
The value range of color then judges that the color of the pixel belongs to the color;After classification, red, blue, white and yellow are calculated
Pixel account for the ratios of whole pixels, obtain red ratio Tr, Blue Scale Tb, white proportion TwWith yellow ratio Ty;Threshold is set
Value T1If Tr、Tb、Tw、TyIt is middle that there are a values more than T1, then determine safe wearing cap;If Tr、Tb、Tw、TyRespectively less than
T1, then judge non-safe wearing cap.
It is as follows whether the Car license recognition module judgment module vehicle enters unauthorized area process:
Present frame is extracted, gray processing is carried out to present frame using mean value method and gray scale stretching is handled, obtains present frame ash
Degree figure;Edge extracting is carried out to present frame gray figure using Canny operator, obtains edge graph;Edge graph progress four times are expanded,
Four corrosion and median filtering are to eliminate noise;The straight line in edge graph is determined using Hough transform, it is flat up and down according to license plate
Row, left and right are in parallel and the ratio characteristic of license plate length and width determines license plate position, construct Car license recognition region;To vehicle
The image of board identification region carries out vertical direction projection, is split according to the characteristics of peak cluster, obtains each character picture;Make
Each character picture is identified with BP neural network, each character is obtained, to identify license plate number;Identify license plate number it
Afterwards, it searches database and determines the corresponding authorized region of the license plate, be compared with preceding camera region is worked as, if camera
Region is not belonging to license plate authorized region, then is warned.
The work clothes identification module judges employee work clothes, and whether process consistent with its working region is as follows:
Read the collected video information of detection zone camera in real time, extract present frame and former frames, to present frame and
Former frames carry out gray processing and handle to obtain the grayscale image of present frame and former frames;Using gauss hybrid models to the gray scale of former frames
Figure carries out modeling trained update, obtains final mask;Judged using each pixel of the model to present frame gray figure,
Obtain moving object figure;It constructs one or more moving object rectangle frames and extracts all independent portions mutually in moving object figure
Point, it is assumed that there are k independent parts mutually in figure, then k moving object rectangle frame can be obtained;According to human body Aspect Ratio, really
Recognize the human body rectangle frame in object rectangle frame, it is assumed that judgement has l human body rectangle frame, extracts each human body rectangle frame current
The image of frame color image corresponding position obtains l human body cromograms;For each Zhang Renti rectangle frame, according to human body proportion
It determines upper body position, region is estimated using upper body position as work clothes, it is colored in corresponding human body to extract work clothes estimation region
The image of figure corresponding part obtains l work clothes cromograms;Judge each work clothes cromogram have according to color
Body is as follows: work clothes cromogram being switched to HSV image, is classified to each pixel, if the H of pixel, S, V value is all satisfied
The value range of certain color is classified as the color, and the pixel for calculating each color accounts for the ratio of whole pixels, and threshold is arranged
The work clothes is determined as the color, inquires database, obtain by value T=0.6 if the ratio of certain color is more than threshold value T
When the corresponding working region work clothes color in the position of preceding camera, compares with the color of judgement, if inconsistent, warned
Show;If the ratio of all colours is no more than threshold value T, it is determined as that employee does not dress work clothes, is warned.
It is described to take object identification module to judge whether employee carries encapsulation process as follows:
The collected video information of detection zone camera is read in real time, extracts present frame and former frames, ash is carried out to it
Degreeization processing, obtains present frame and former frame grayscale images;Modeling training is carried out more using grayscale image of the vibe algorithm to former frames
Newly, final mask is obtained;Judged using each pixel of the model to present frame gray figure, obtains moving region, structure
Building human body bounding rectangles includes all moving regions;Object, which is constructed, according to human body proportion estimates rectangle;In object estimation rectangle
The symmetry of object is that standard judges whether to take object;Specific step is as follows:
Step 1, the collected video information of camera is read, former frames and present frame are extracted;Gray scale is carried out to each frame
Change processing, carry out gray processing to image using weighted mean method: each pixel is weighted, to R, tri- components of G, B
Distribute different weights, in f (x, y)=0.3R (x, y)+0.59G (x, y)+0.11B (x, y) formula, x, y refer to that pixel is sat
Mark;
Step 2, it carries out modeling trained update using grayscale image of the vibe algorithm to former frames, obtains final mask;
Step 3, judged using each pixel of the final mask to present frame gray figure, obtain moving region, structure
Building human body bounding rectangles includes all moving regions;
Step 4, according to human body proportion formed object estimate rectangle, it is specific as follows: with the top edge of human body bounding rectangles with
The 1/5 of lower edge close to top edge position as object estimation rectangle top edge, with human body bounding rectangles top edge under
2/5 lower edge close to the position of lower edge as object estimation rectangle between edge, with the left edge of human body bounding rectangles and the right side
It is respectively the left edge and right hand edge of object estimation rectangle between edge, forms object and estimate rectangle;
Step 5, judge whether to take object as standard using the symmetry in object estimation region, process is as follows: square is estimated with object
The shape lower left corner is that origin establishes coordinate system, and object estimates rectangle in first quartile, takes the object portion of object estimation rectangle top edge
The midpoint m divided, obtains its abscissa xm, with x=xmFor symmetry axis, the object area S1 of symmetry axis left area, calculating pair are calculated
Claim the object area S2 on the right of axis, calculates symmetrical degreeThreshold value th and tl are set, if tl < β < th, is judged as not
Object is carried, is otherwise judged as carrying object.
The Multisensor video fusion system, fusion video monitoring system, face identification system and video structure analyzing system, shape
At user oriented Video Applications system, realized in a software video check, video alarm prompt, video playback, alarm
A series of functions such as record, user management, Role Management, rights management, the convenient unified management to Video Applications.
The beneficial effects of the present invention are: the working condition of plant area can be monitored in real time and be analyzed, once discovery person
Work cap not safe to carry, employee take object enter forbid taking object area, vehicle drives into its unauthorized area, employee goes to the wrong way workspace
Domain, the excessively high dangerous situation of key area temperature can carry out warning and checking immediately in time.
Detailed description of the invention
Fig. 1 is the intelligent monitoring emerging system structure chart proposed by the invention based on computer vision analysis.
Specific embodiment
Intelligent monitoring emerging system structure based on computer vision analysis proposed by the invention is as shown in Figure 1, video
Monitoring system is made of front end camera, thermal infrared imager and video monitoring platform;Front end camera is responsible for acquiring video information;
Thermal infrared imager is responsible for acquiring plant area's important area temperature information;The video letter of video monitoring platform receiving front-end camera acquisition
Breath and plant area's important area temperature information of thermal infrared imager acquisition, are shown, while sending video information to video
Structured analysis system, and send video information and plant area's important area temperature information to Multisensor video fusion system.
Face identification system is made of face snap device, face facial recognition modules and worker's database;Face is grabbed
Employee's image of device acquisition disengaging each key area of plant area is clapped, and sends face facial recognition modules to;Worker's database
Store the identity information and its face feature of each employee;Face facial recognition modules receive employee's figure of face snap device acquisition
Picture carries out recognition of face, determines its identity information by the comparison with worker's database, and send recognition result to view
Frequency emerging system.
Video structure analyzing system is by safety cap identification module, Car license recognition module, work clothes identification module and takes
Object identification module is constituted, and is received the video information of video monitoring system acquisition, is judged that employee is by safety cap identification module
Whether the no cap of safe wearing enters unauthorized area by Car license recognition judgment module vehicle, by work clothes identification module
Judge whether employee work clothes are consistent with its working region, judge whether employee carries package by taking object identification module, and will
Analysis result sends Multisensor video fusion system to.
Multisensor video fusion system, including video alarm cue module, video playback module, alarm logging module, user management
Module, role management module, authority management module.
The safety cap identification module judges that whether the process of employee's safe wearing cap as follows:
Step 1, former frame and present frame are extracted, gray processing is carried out to former frame and present frame picture using weighted mean method
Processing, the i.e. R to each pixel, tri- components of G, B distribute different weights, specific formula is as follows: f (x, y)=0.3R (x,
Y)+0.59G (x, y)+0.11B (x, y), in formula, R (x, y), G (x, y), B (x, y) respectively refer to the R for the pixel that coordinate is (x, y),
G, B component, f (x, y) refer to the gray value after the pixel conversion that coordinate is (x, y);
Step 2, it is handled using grayscale image of the frame differential method to former frame and present frame, obtains moving object figure,
Process is as follows: the gray value of two field pictures respective pixel being subtracted each other, and takes its absolute value, obtains difference image Dn(x, y)
=| fn(x, y)-fn-1(x, y) |, wherein fn(x, y) refers to gray value of the coordinate for the pixel of (x, y), f in present framen-1(x, y) refers to
Coordinate is the gray value of the pixel of (x, y) in former frame;Threshold value T is seta, this example takes TaValue is the 1/10 of current frame pixel, i.e.,
Ta=fn(x, y)/10;If Dn(x, y) > T, Rn=255, if Dn(x, y) < T, Rn=0, obtain moving object figure R;R is
Binary image, wherein white area represents moving object region;
Step 3, it constructs k moving object estimation rectangle frame and extracts all independent parts mutually in moving object figure,
And the location information of each moving object estimation rectangle frame is obtained, it is stored in matrix L O, detailed process is as follows:
Step1: reference axis is established as origin in the lower left corner of the moving object figure R obtained using step 2, and moving object figure R exists
First quartile;If straight line x=h, h initial value is 0;
Step2: straight line x=h is moved to right into a unit, judges the pixel value of each pixel on straight line x=h;
Step3: if the pixel value of each pixel is 0 on straight line x=h, Step2 is returned to;If deposited on straight line x=h
In the pixel that pixel value is not 0, remember that h at this time is xmin;
Step4: straight line x=h is moved to right into a unit, judges the pixel value of each pixel on straight line x=h;
Step5: if there are the pixel that pixel value is non-zero on straight line x=h, remember that uppermost non-zero pixel is corresponding on straight line
Ordinate value and the corresponding ordinate value of bottom non-zero pixel, and be stored in array Y, return to Step4;If straight line x=
The pixel value of the upper each pixel of h is 0, remembers that h at this time is xmax;
Step6: being compared the size of the element in array Y, obtains the maximum value y in YmaxWith minimum value ymin;With
Straight line y=ymaxAs top edge, straight line y=yminAs lower edge, straight line x=xminAs left edge, straight line x=xmaxAs
Region rectangle frame is estimated in right hand edge tectonic movement;Array Y is reset, and remembers LOk=(ymax, ymin, xmin, xmax, k) and it is stored in matrix
In LO, wherein k initial value is 1, and k value adds 1 after each tectonic movement estimation region rectangle frame;
Step7: repeating Step2 to Step6, until straight line x=h terminates when being moved to binaryzation picture rightmost;
Step 4, each moving object estimation rectangle frame corresponding part in the grayscale image and cromogram of present frame is extracted
Image obtains K moving object estimation grayscale image G1..., GKCromogram P is estimated with K moving objects1..., PK;Extraction process
It is as follows: rectangle frame being estimated for i-th of moving object, its corresponding location information LO is read from the matrix L O that step 3 obtainsi
=(ymax, ymin, xmin, xmax, i), confirm that rectangle is estimated in the moving object using top edge, lower edge, left edge and right hand edge
The specific location of frame is to extract;
Step 5, grayscale image G is estimated to moving object one by one using Canny operator1To GKEdge detection is carried out, obtains K
Edge pixels all in every edge graph are compared by edge graph with the pixel of corresponding position in moving object figure, delete side
With the unequal edge pixel of pixel pixel value of corresponding position in moving object figure in edge pixel, final moving object is obtained
Edge graph;It includes all edges in the figure that a moving object rectangle frame is constructed in every moving object edge graph, obtains K
Moving object rectangle frame M1..., MK;
Tectonic movement object rectangle frame process is as follows: firstly, traversing from top to bottom to the moving object edge graph, obtaining
Obtain the leftmost edge pixel X of every a linele={ xl1, xl2, xl3, xl4, xl5..., xlmAnd rightmost edge pixel cross
Coordinate Xre={ xr1, xr2, xr3, xr4, xr5..., xrm, wherein xliRepresent the abscissa of the leftmost edge pixel of the i-th row, xri
The abscissa of the edge pixel of the i-th row rightmost is represented, m represents the moving object edge graph line number;Then, from left to right to this
Moving object edge graph is traversed, and the uppermost edge pixel Y of each column is obtainedhe={ yh1, yh2, yh3, yh4, yh5...,
yhnAnd nethermost edge pixel ordinate Yle={ yl1, yl2, yl3, yl4, yl5..., yln, wherein yhiRepresent the i-th column most
The ordinate of edge pixel above, yliThe ordinate of the nethermost edge pixel of the i-th column is represented, n represents the moving object side
Edge figure columns;Finally, with XleIn least member as left edge, XreMiddle greatest member is as right hand edge, YheIn greastest element
Element is used as top edge, YleIn least member as lower edge, tectonic movement object rectangle frame;
Step 6, according to human body proportion to moving object rectangle frame M1To MKJudged, obtain human body rectangle frame therein,
Assuming that judging there be k human body rectangle frame, each human body rectangle frame is extracted in its corresponding moving object estimation cromogram
The image of corresponding part obtains k human body cromograms;
Judge whether it is that human body rectangle frame process is as follows: the note a length of L of moving object rectangle frame rectangle, width W, λ=W/L,
Threshold alpha, β are set, and α, β take 3 and 5 respectively in this example;If α < λ < β, is regarded as human body rectangle frame, is otherwise considered as it
His object;
Step 7, for each human body rectangle frame, according to the head of people in upper half of human body feature location human body rectangle frame
Region is estimated according to head ratio positioning security cap by portion, obtains k safety cap estimation region;
Head position fixing process is as follows: firstly, cutting to human body rectangle frame, head zone is obtained, it is specific as follows: to take people
Body rectangle frame top edge is as head zone top edge, human body rectangle frame top edge and lower edge 1/3 close to the position of top edge
As head zone lower edge, left edge and right hand edge of the left edge and right hand edge of human body rectangle frame as head zone;Its
It is secondary, the abscissa that head zone obtains every a line leftmost edge point and rightmost edges point, X are traversed from top to bottomld={ xl1, xl2,
xl3, xl4, xl5..., xlgAnd Xrd={ xr1, xr2, xr3, xr4, xr5..., xrg, wherein xliIt is horizontal to represent the i-th row Far Left marginal point
Coordinate, xriThe i-th row rightmost marginal point coordinate is represented, g represents head zone line number;Then, Far Left edge is calculated to most right
The distance D=X at side edgerd-Xld, obtain D={ D1, D2, D3, D4, D5..., Dg, calculate element latter and previous item in D
Poor DΔ={ D2-D1, D3-D2, D4-D3, D5-D4..., Dg-Dg-1};Finally find out DΔFrom the 11st element into a last element
The corresponding line number p of maximum element, is updated to pth row for head zone lower edge;
Construction safety cap rectangle frame process is as follows: finding out X respectivelyldIn from xlpTo xlgThe smallest element xmlAnd XrdIn from
xrpTo xrgGreatest member xmr, with x=xmlAs accurate head zone left edge, with x=xmrAs right hand edge, with header area
Domain top edge and lower edge obtain accurate head zone as accurate head top edge and lower edge;Then, to accurate header area
The further cutting in domain is specific as follows to obtain safety cap region: on safety cap region, left and right edge and accurate head zone one
Cause, using accurate head zone top edge and lower edge 1/3 close to top edge position as safety cap region lower edge;
Step 8, safety cap rectangle frame is extracted in the image of human body cromogram corresponding position, obtains a safety cap region color
Figure is that foundation one by one judges safety cap region color figure with color;
Its deterministic process is as follows: safety cap region color figure is switched into HSV image, due to general factory safety cap have it is red
Color, blue, white and yellow, so taking red, blue, white and yellow as judgment criteria, the corresponding color of these four colors
Adjust H, saturation degree S, the value range of lightness V as follows:
It is red: H:0-10 and 156-180;S:43-255;V:46-255
Blue: H:100-124;S:43-255;V:46-255
White: H:0-180;S:0-30;V:221-255
Yellow: H:26-34;S:43-255;V:46-255
Classify to the color of each pixel, process is as follows: if the H of pixel to be sorted, S, V value is all satisfied certain face
The value range of color then judges that the color of the pixel belongs to the color;After classification, red, blue, white and yellow are calculated
Pixel account for the ratios of whole pixels, obtain red ratio Tr, Blue Scale Tb, white proportion TwWith yellow ratio Ty;Threshold is set
Value T1, this example T1Take 0.5;If Tr、Tb、Tw、TyIt is middle that there are a values more than T1, then determine safe wearing cap;If Tr、
Tb、Tw、TyRespectively less than T1, then judge non-safe wearing cap.
It is described to take object identification module to judge whether employee carries encapsulation process as follows:
Step 1, the collected video information of camera is read, former frames and present frame are extracted;Gray scale is carried out to each frame
Change processing, carry out gray processing to image using weighted mean method: each pixel is weighted, to R, tri- components of G, B
Distribute different weights, in f (x, y)=0.3R (x, y)+0.59G (x, y)+0.11B (x, y) formula, x, y refer to that pixel is sat
Mark;
Step 2, it carries out modeling trained update using grayscale image of the vibe algorithm to former frames, obtains final mask;
Step 3, judged using each pixel of the final mask to present frame gray figure, obtain moving region, structure
Building human body bounding rectangles includes all moving regions;
Step 4, according to human body proportion formed object estimate rectangle, it is specific as follows: with the top edge of human body bounding rectangles with
The 1/5 of lower edge close to top edge position as object estimation rectangle top edge, with human body bounding rectangles top edge under
2/5 lower edge close to the position of lower edge as object estimation rectangle between edge, with the left edge of human body bounding rectangles and the right side
It is respectively the left edge and right hand edge of object estimation rectangle between edge, forms object and estimate rectangle;
Step 5, judge whether to take object as standard using the symmetry in object estimation region, process is as follows: square is estimated with object
The shape lower left corner is that origin establishes coordinate system, and object estimates rectangle in first quartile, takes the object portion of object estimation rectangle top edge
The midpoint m divided, obtains its abscissa xm, with x=xmFor symmetry axis, the object area S1 of symmetry axis left area, calculating pair are calculated
Claim the object area S2 on the right of axis, calculates symmetrical degreeThreshold value th and tl are set, this example th value 0.75, tl takes
Value 1.33;If tl < β < th, is judged as and does not carry object, be otherwise judged as carrying object.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
Anyone skilled in the art in the technical scope disclosed by the present invention, the change or replacement that can be guessed, all
It is covered by the protection scope of the present invention.
Claims (7)
1. a kind of Intellectualized monitoring emerging system based on computer vision analysis, which is characterized in that including video monitoring system,
Face identification system, video structure analyzing system, Multisensor video fusion system;Wherein,
The video monitoring system is made of front end camera, thermal infrared imager and video monitoring platform;Front end camera is responsible for
Acquire video information;Thermal infrared imager is responsible for acquiring plant area's important area temperature information;The camera shooting of video monitoring platform receiving front-end
The video information of head acquisition and plant area's important area temperature information of thermal infrared imager acquisition, are shown, while by video
Information sends video structure analyzing system to, and sends video information and plant area's important area temperature information to video fusion
System;
The face identification system is made of face snap device, face facial recognition modules and worker's database;Face is grabbed
Employee's image of device acquisition disengaging each key area of plant area is clapped, and sends face facial recognition modules to;Worker's database
Store the identity information and its face feature of each employee;Face facial recognition modules receive employee's figure of face snap device acquisition
Picture carries out recognition of face, determines its identity information by the comparison with worker's database, and send recognition result to view
Frequency emerging system;
The video structure analyzing system is by safety cap identification module, Car license recognition module, work clothes identification module and takes
Object identification module is constituted, and is received the video information of video monitoring system acquisition, is judged that employee is by safety cap identification module
Whether the no cap of safe wearing enters unauthorized area by Car license recognition judgment module vehicle, by work clothes identification module
Judge whether employee work clothes are consistent with its working region, judge whether employee carries package by taking object identification module, and will
Analysis result sends Multisensor video fusion system to;
The Multisensor video fusion system, including video alarm cue module, video playback module, alarm logging module, user management
Module, role management module, authority management module.
2. a kind of Intellectualized monitoring emerging system based on computer vision analysis as described in claim 1, which is characterized in that
The face feature of each employee of worker's database purchase, calculating process are as follows in the face identification system:
The facial image of each employee of plant area is acquired, image size is M × N, it is assumed that share K employees, then K faces can be obtained
Image;Grey linear transformation processing and 3 × 3 median filter process are carried out to K facial images, obtain K face grayscale images;
The gray value of every one-row pixels of every face grayscale image is connected, then every face grayscale image may make up a D=M × N-dimensional
Row vector, for i-th face grayscale image, remember its constitute row vector xi, and calculate average faceBy K
It opens the K row vector that face grayscale image is constituted to be arranged, constitutes the matrix of K × D dimension, and Karhunen-Loeve transformation is carried out to the matrix, obtain
To eigenface space: w=(u1, u2, u3..., up), in formula: p be setting dimensionality reduction after dimension;The face for calculating each employee is special
Sign, for j-th of employee, face feature ΩjCalculation formula are as follows: Ωj=wT(xj-Ψ)。
3. a kind of Intellectualized monitoring emerging system based on computer vision analysis as described in claim 1, which is characterized in that
The process of face facial recognition modules progress recognition of face is as follows in the face identification system:
It is linear to carry out gray scale to facial image for the facial image for disengaging each key area of plant area that acquisition face snap device is captured
Transformation obtains face grayscale image, and the gray value of the every one-row pixels of face grayscale image is connected, row vector y is obtained;Calculate the face
The face feature Ω of imagej=wT(xj- Ψ), in formula: w is the eigenface space acquired in claim 2: w=(u1, u2,
u3..., up), Ψ is the average face acquired in claim 2;Calculate ΩyWith the face feature of worker's database purchase
Euclidean distance, is selected and ΩyThe nearest face feature of Euclidean distance, the corresponding worker of the face feature is recognition of face
As a result.
4. a kind of Intellectualized monitoring emerging system based on computer vision analysis as described in claim 1, which is characterized in that
Safety cap identification module judges that whether the process of employee's safe wearing cap as follows in the video structure analyzing system:
The video information for reading the acquisition of detection zone camera in real time, extracts former frame and present frame, to former frame and present frame
Gray processing processing is carried out, the grayscale image of former frame and present frame is obtained;Using frame differential method to the ash of former frame and present frame
Degree figure is handled, and moving object figure is obtained;It constructs one or more moving object estimation rectangle frames and extracts moving object figure
In all independent parts mutually, it is assumed that have the independent parts mutually K in figure, then K moving object can be obtained and estimate rectangle
Frame;The image for extracting each moving object estimation rectangle frame corresponding part in the grayscale image and cromogram of present frame, obtains K
Grayscale image is estimated in moving object and cromogram is estimated in K moving objects;Gray scale is estimated to K moving objects using Canny operator
Figure carries out edge detection one by one, obtains K moving object edge graphs, a moving object is constructed in every moving object edge graph
Body rectangle frame includes all edges in the figure, obtains K moving object rectangle frame;According to human body Aspect Ratio, moving object is confirmed
Human body rectangle frame in body rectangle frame, it is assumed that judge there be k human body rectangle frame, extract each human body rectangle frame in moving object
Body estimates the image of corresponding part in cromogram, obtains k human body cromograms;For each human body rectangle frame, according to human body
Region is estimated according to head ratio positioning security cap in the head of people in upper part of the body feature location human body rectangle frame, obtains k safety
Cap estimates region;The image for extracting each safety cap estimation region corresponding part in its corresponding human body cromogram, obtains k
Safety cap region color figure is opened, is that foundation one by one judges safety cap region color figure with color;Specific step is as follows:
Step 1, former frame and present frame are extracted, former frame and present frame picture are carried out at gray processing using weighted mean method
Reason, i.e. the R to each pixel, tri- components of G, B distribute different weights, specific formula is as follows: f (x, y)=0.3R (x, y)
+ 0.59G (x, y)+0.11B (x, y), in formula, R (x, y), G (x, y), B (x, y) respectively refer to R, the G for the pixel that coordinate is (x, y),
B component, f (x, y) refer to the gray value after the pixel conversion that coordinate is (x, y);
Step 2, it is handled using grayscale image of the frame differential method to former frame and present frame, obtains moving object figure, process
It is as follows: the gray value of two field pictures respective pixel being subtracted each other, and takes its absolute value, obtains difference image Dn(x, y)=| fn
(x, y)-fn-1(x, y) |, wherein fn(x, y) refers to gray value of the coordinate for the pixel of (x, y), f in present framen-1(x, y) refers to previous
Coordinate is the gray value of the pixel of (x, y) in frame;Threshold value T is setaIf Dn(x, y) > T, Rn=255, if Dn(x, y) <
T, Rn=0, obtain moving object figure R;R is binary image, and wherein white area represents moving object region;
Step 3, it constructs K moving object estimation rectangle frame and extracts all independent parts mutually in moving object figure, and obtain
The location information for taking each moving object estimation rectangle frame, is stored in matrix L O, detailed process is as follows:
Step1: reference axis is established, moving object figure R is first as origin in the lower left corner of the moving object figure R obtained using step 2
Quadrant;If straight line x=h, h initial value is 0;
Step2: straight line x=h is moved to right into a unit, judges the pixel value of each pixel on straight line x=h;
Step3: if the pixel value of each pixel is 0 on straight line x=h, Step2 is returned to;If on straight line x=h, there are pictures
Element value is not 0 pixel, remembers that h at this time is xmin;
Step4: straight line x=h is moved to right into a unit, judges the pixel value of each pixel on straight line x=h;
Step5: if there are the pixel that pixel value is non-zero on straight line x=h, remember the corresponding vertical seat of uppermost non-zero pixel on straight line
Scale value and the corresponding ordinate value of bottom non-zero pixel, and be stored in array Y, return to Step4;If on straight line x=h
The pixel value of each pixel is 0, remembers that h at this time is xmax;
Step6: being compared the size of the element in array Y, obtains the maximum value y in YmaxWith minimum value ymin;With straight line y
=ymaxAs top edge, straight line y=yminAs lower edge, straight line x=xminAs left edge, straight line x=xmaxAs the right
Region rectangle frame is estimated in edge tectonic movement;Array Y is reset, and remembers LOk=(ymax, ymin, xmin, xmax, k) and it is stored in matrix L O
In, wherein k initial value is 1, and k value adds 1 after each tectonic movement estimation region rectangle frame;
Step7: repeating Step2 to Step6, until straight line x=h terminates when being moved to binaryzation picture rightmost;
Step 4, the image of each moving object estimation rectangle frame corresponding part in the grayscale image and cromogram of present frame is extracted,
Obtain K moving object estimation grayscale image G1..., GKCromogram P is estimated with K moving objects1..., PK;Extraction process is as follows:
Rectangle frame is estimated for i-th of moving object, its corresponding location information LO is read from the matrix L O that step 3 obtainsi=
(ymax, ymin, xmin, xmax, i), confirm that rectangle frame is estimated in the moving object using top edge, lower edge, left edge and right hand edge
Specific location to extracting;
Step 5, grayscale image G is estimated to moving object one by one using Canny operator1To GKEdge detection is carried out, K edges are obtained
Figure, edge pixels all in every edge graph are compared with the pixel of corresponding position in moving object figure, delete edge picture
With the unequal edge pixel of pixel pixel value of corresponding position in moving object figure in element, final moving object edge is obtained
Figure;It includes all edges in the figure that a moving object rectangle frame is constructed in every moving object edge graph, obtains K movement
Object rectangle frame M1..., MK;
Tectonic movement object rectangle frame process is as follows: firstly, traversing from top to bottom to the moving object edge graph, obtaining every
The leftmost edge pixel X of a linele={ xl1, xl2, xl3, xl4, xl5..., xlmAnd rightmost edge pixel abscissa
Xre={ xr1, xr2, xr3, xr4, xr5..., xrm, wherein xliRepresent the abscissa of the leftmost edge pixel of the i-th row, xriIt represents
The abscissa of the edge pixel of i-th row rightmost, m represent the moving object edge graph line number;Then, from left to right to the movement
Object edge figure is traversed, and the uppermost edge pixel Y of each column is obtainedhe={ yh1, yh2, yh3, yh4, yh5..., yhn}
With the ordinate Y of nethermost edge pixelle={ yl1, yl2, yl3, yl4, yl5..., yln, wherein yhiIt is most upper to represent the i-th column
The ordinate of the edge pixel in face, yliThe ordinate of the nethermost edge pixel of the i-th column is represented, n represents the moving object edge
Figure columns;Finally, with XleIn least member as left edge, XreMiddle greatest member is as right hand edge, YheIn greatest member
As top edge, YleIn least member as lower edge, tectonic movement object rectangle frame;
Step 6, according to human body proportion to moving object rectangle frame M1To MKJudged, obtain human body rectangle frame therein, it is assumed that
Judge there be k human body rectangle frame, it is corresponding in its corresponding moving object estimation cromogram to extract each human body rectangle frame
Partial image obtains k human body cromograms;
Judge whether it is that human body rectangle frame process is as follows: the note a length of L of moving object rectangle frame rectangle, width W, λ=W/L, setting
Threshold alpha, β;If α < λ < β, is regarded as human body rectangle frame, is otherwise considered as other objects;
Step 7, for each human body rectangle frame, according to the head of people in upper half of human body feature location human body rectangle frame, root
Region is estimated according to head ratio positioning security cap, obtains k safety cap estimation region;
Head position fixing process is as follows: firstly, cutting to human body rectangle frame, head zone is obtained, it is specific as follows: to take human body square
Shape upper frame edge edge is as head zone top edge, human body rectangle frame top edge and lower edge 1/3 close to the position conduct of top edge
Head zone lower edge, left edge and right hand edge of the left edge and right hand edge of human body rectangle frame as head zone;Secondly, from
Under to upper traversal head zone obtain the abscissa of every a line leftmost edge point and rightmost edges point, Xld={ xl1, xl2, xl3,
xl4, xl5..., xlgAnd Xrd={ xr1, xr2, xr3, xr4, xr5..., xrg, wherein xliRepresent the i-th horizontal seat of row Far Left marginal point
Mark, xriThe i-th row rightmost marginal point coordinate is represented, g represents head zone line number;Then, Far Left edge is calculated to rightmost
The distance D=X at edgerd-Xld, obtain D={ D1, D2, D3, D4, D5..., Dg, calculate the difference of element latter and previous item in D
DΔ={ D2-D1, D3-D2, D4-D3, D5-D4..., Dg-Dg-1};Finally find out DΔFrom the 11st element into a last element most
The corresponding line number p of big element, is updated to pth row for head zone lower edge;
Construction safety cap rectangle frame process is as follows: finding out X respectivelyldIn from xlpTo xlgThe smallest element xmlAnd XrdIn from xrpIt arrives
xrgGreatest member xmr, with x=xmlAs accurate head zone left edge, with x=xmrAs right hand edge, in head zone
Edge and lower edge obtain accurate head zone as accurate head top edge and lower edge;Then, to accurate head zone into
One step cutting is specific as follows to obtain safety cap region: on safety cap region, left and right edge it is consistent with accurate head zone, with
Accurate head zone top edge and lower edge 1/3 close to top edge position as safety cap region lower edge;
Step 8, safety cap rectangle frame is extracted in the image of human body cromogram corresponding position, obtains a safety cap region color figure,
It is that foundation one by one judges safety cap region color figure with color;
Its deterministic process is as follows: safety cap region color figure being switched to HSV image, since general factory safety cap has red, indigo plant
The corresponding tone H of these four colors, color, white and yellow are satisfied so taking red, blue, white and yellow as judgment criteria
It is as follows with degree S, the value range of lightness V:
It is red: H:0-10 and 156-180;S:43-255;V:46-255
Blue: H:100-124;S:43-255;V:46-255
White: H:0-180;S:0-30;V:221-255
Yellow: H:26-34;S:43-255;V:46-255
Classify to the color of each pixel, process is as follows: if the H of pixel to be sorted, S, V value is all satisfied certain color
Value range then judges that the color of the pixel belongs to the color;After classification, the picture of red blue, white and yellow is calculated
Element accounts for the ratio of whole pixels, obtains red ratio Tr, Blue Scale Tb, white proportion TwWith yellow ratio Ty;Threshold value T is set1,
If Tr、Tb、Tw、TyIt is middle that there are a values more than T1, then determine safe wearing cap;If Tr、Tb、Tw、TyRespectively less than T1, then
Judge non-safe wearing cap.
5. a kind of Intellectualized monitoring emerging system based on computer vision analysis as described in claim 1, which is characterized in that
The step of whether Car license recognition judgment module vehicle enters unauthorized area in the video structure analyzing system are as follows:
Present frame is extracted, gray processing is carried out to present frame using mean value method and gray scale stretching is handled, obtains present frame gray figure;
Edge extracting is carried out to present frame gray figure using Canny operator, obtains edge graph;Four expansions, four times are carried out to edge graph
Corrosion and median filtering are to eliminate noise;The straight line in edge graph is determined using Hough transform, it is parallel up and down according to license plate, left
Right parallel and license plate length and width ratio characteristic determines license plate position, constructs Car license recognition region;License plate is known
The image in other region carries out vertical direction projection, is split according to the characteristics of peak cluster, obtains each character picture;Use BP
Neural network identifies each character picture, obtains each character, to identify license plate number;After identifying license plate number,
It searches database and determines the corresponding authorized region of the license plate, be compared with preceding camera region is worked as, if camera institute
It is not belonging to license plate authorized region in region, then is warned.
6. a kind of Intellectualized monitoring emerging system based on computer vision analysis as described in claim 1, which is characterized in that
In the video structure analyzing system work clothes identification module judge employee work clothes whether with the consistent step in its working region
Suddenly are as follows:
The collected video information of detection zone camera is read in real time, extracts present frame and former frames, to present frame and former
Frame carries out gray processing and handles to obtain the grayscale image of present frame and former frames;Using gauss hybrid models to the grayscale images of former frames into
Row modeling training updates, and obtains final mask;Judged using each pixel of the model to present frame gray figure, is obtained
Moving object figure;It constructs one or more moving object rectangle frames and extracts all independent parts mutually in moving object figure,
Assuming that there are k independent parts mutually in figure, then k moving object rectangle frame can be obtained;According to human body Aspect Ratio, object is confirmed
Human body rectangle frame in body rectangle frame, it is assumed that judgement has l human body rectangle frame, extracts each human body rectangle frame in present frame coloured silk
The image of chromatic graph piece corresponding position obtains l human body cromograms;For each Zhang Renti rectangle frame, determined according to human body proportion
Upper body position estimates region using upper body position as work clothes, extracts work clothes estimation region in corresponding human body cromogram pair
The image for answering part obtains l work clothes cromograms;Each work clothes cromogram is judged according to color, specifically such as
Under: work clothes cromogram is switched into HSV image, is classified to each pixel, if the H of pixel, S, V value is all satisfied certain
The value range of color is classified as the color, and the pixel for calculating each color accounts for the ratio of whole pixels, and threshold value T is arranged
=0.6, if the ratio of certain color is more than threshold value T, which is determined as the color, database is inquired, is worked as
The corresponding working region work clothes color in the position of preceding camera, compares with the color of judgement, if inconsistent, is warned
Show;If the ratio of all colours is no more than threshold value T, it is determined as that employee does not dress work clothes, is warned.
7. a kind of Intellectualized monitoring emerging system based on computer vision analysis as described in claim 1, which is characterized in that
Take object identification module judges the step of whether employee carries package in the video structure analyzing system are as follows:
The collected video information of detection zone camera is read in real time, extracts present frame and former frames, gray processing is carried out to it
Processing, obtains present frame and former frame grayscale images;It carries out modeling trained update using grayscale image of the vibe algorithm to former frames, obtain
To final mask;Judged using each pixel of the model to present frame gray figure, obtain moving region, constructs human body
Bounding rectangles include all moving regions;Object, which is constructed, according to human body proportion estimates rectangle;With object in object estimation rectangle
Symmetry is that standard judges whether to take object;Specific step is as follows:
Step 1, the collected video information of camera is read, former frames and present frame are extracted;Each frame is carried out at gray processing
Reason, carry out gray processing to image using weighted mean method: each pixel is weighted, to R, tri- component distribution of G, B
Different weights, in f (x, y)=0.3R (x, y)+0.59G (x, y)+0.11B (x, y) formula, x, y refer to pixel coordinate;
Step 2, it carries out modeling trained update using grayscale image of the vibe algorithm to former frames, obtains final mask;
Step 3, judged using each pixel of the final mask to present frame gray figure, obtain moving region, construct people
Body bounding rectangles include all moving regions;
Step 4, object is formed according to human body proportion and estimates rectangle, it is specific as follows: with the top edge of human body bounding rectangles and below
The 1/5 of edge close to top edge position as object estimation rectangle top edge, with human body bounding rectangles top edge and lower edge
Between 2/5 close to the position of lower edge as object estimation rectangle lower edge, with the left edge and right hand edge of human body bounding rectangles
Between be respectively object estimation rectangle left edge and right hand edge, formed object estimate rectangle;
Step 5, judge whether to take object as standard using the symmetry in object estimation region, process is as follows: left with object estimation rectangle
Inferior horn is that origin establishes coordinate system, and object estimates rectangle in first quartile, takes the object parts of object estimation rectangle top edge
Midpoint m obtains its abscissa xm, with x=xmFor symmetry axis, the object area S1 of symmetry axis left area is calculated, calculates symmetry axis
The object area S2 on the right, calculates symmetrical degreeThreshold value th and tl are set, if tl < β < th, is judged as and does not carry
Otherwise object is judged as carrying object.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910134732.2A CN110008831A (en) | 2019-02-23 | 2019-02-23 | A kind of Intellectualized monitoring emerging system based on computer vision analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910134732.2A CN110008831A (en) | 2019-02-23 | 2019-02-23 | A kind of Intellectualized monitoring emerging system based on computer vision analysis |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110008831A true CN110008831A (en) | 2019-07-12 |
Family
ID=67165928
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910134732.2A Pending CN110008831A (en) | 2019-02-23 | 2019-02-23 | A kind of Intellectualized monitoring emerging system based on computer vision analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110008831A (en) |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110428587A (en) * | 2019-07-19 | 2019-11-08 | 国网安徽省电力有限公司建设分公司 | A kind of engineering site early warning interlock method and system |
CN110517461A (en) * | 2019-08-30 | 2019-11-29 | 成都智元汇信息技术股份有限公司 | A method of it prevents people from carrying package and escapes safety check |
CN110533811A (en) * | 2019-08-28 | 2019-12-03 | 深圳市万睿智能科技有限公司 | The method and device and system and storage medium of safety cap inspection are realized based on SSD |
CN110930624A (en) * | 2019-12-06 | 2020-03-27 | 深圳北斗国芯科技有限公司 | Safety in production monitored control system based on big dipper |
CN111083441A (en) * | 2019-12-18 | 2020-04-28 | 广州穗能通能源科技有限责任公司 | Construction site monitoring method and device, computer equipment and storage medium |
CN111898514A (en) * | 2020-07-24 | 2020-11-06 | 燕山大学 | Multi-target visual supervision method based on target detection and action recognition |
CN112071006A (en) * | 2020-09-11 | 2020-12-11 | 湖北德强电子科技有限公司 | High-efficiency low-resolution image area intrusion recognition algorithm and device |
CN112466086A (en) * | 2020-10-26 | 2021-03-09 | 福州微猪信息科技有限公司 | Visual identification early warning device and method for farm work clothes |
CN112528821A (en) * | 2020-12-06 | 2021-03-19 | 杭州晶一智能科技有限公司 | Pedestrian crosswalk pedestrian detection method based on motion detection |
CN112532927A (en) * | 2020-11-17 | 2021-03-19 | 南方电网海南数字电网研究院有限公司 | Intelligent safety management and control system for construction site |
CN112669505A (en) * | 2019-12-16 | 2021-04-16 | 丰疆智能科技股份有限公司 | Integrated management system for shower and entrance guard of farm and management method thereof |
CN112883969A (en) * | 2021-03-01 | 2021-06-01 | 河海大学 | Rainfall intensity detection method based on convolutional neural network |
CN112949367A (en) * | 2020-07-07 | 2021-06-11 | 南方电网数字电网研究院有限公司 | Method and device for detecting color of work clothes based on video stream data |
CN113379144A (en) * | 2021-06-24 | 2021-09-10 | 深圳开思信息技术有限公司 | Store purchase order generation method and system for online automobile distribution purchase platform |
CN113408501A (en) * | 2021-08-19 | 2021-09-17 | 北京宝隆泓瑞科技有限公司 | Oil field park detection method and system based on computer vision |
CN113977603A (en) * | 2021-10-29 | 2022-01-28 | 连云港福润食品有限公司 | Monitoring robot based on target detection, identification and tracking for worker production specification |
US11346938B2 (en) | 2019-03-15 | 2022-05-31 | Msa Technology, Llc | Safety device for providing output to an individual associated with a hazardous environment |
CN115925243A (en) * | 2022-12-24 | 2023-04-07 | 山西百澳智能玻璃股份有限公司 | Method and system for regulating and controlling heating temperature of glass |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102262729A (en) * | 2011-08-03 | 2011-11-30 | 山东志华信息科技股份有限公司 | Fused face recognition method based on integrated learning |
CN103235938A (en) * | 2013-05-03 | 2013-08-07 | 北京国铁华晨通信信息技术有限公司 | Method and system for detecting and identifying license plate |
CN107632565A (en) * | 2017-08-28 | 2018-01-26 | 上海欧忆能源科技有限公司 | Power construction building site intellectualized management system and method |
CN108833831A (en) * | 2018-06-15 | 2018-11-16 | 陈在新 | A kind of power construction intelligent safety monitor system |
CN109034535A (en) * | 2018-06-21 | 2018-12-18 | 中国化学工程第六建设有限公司 | Construction site wisdom monitoring method, system and computer readable storage medium |
CN109117827A (en) * | 2018-09-05 | 2019-01-01 | 武汉市蓝领英才科技有限公司 | Work clothes work hat wearing state automatic identifying method and alarm system based on video |
CN109215155A (en) * | 2018-09-29 | 2019-01-15 | 东莞方凡智能科技有限公司 | A kind of building site management system based on technology of Internet of things |
CN109218673A (en) * | 2018-09-20 | 2019-01-15 | 国网江苏省电力公司泰州供电公司 | The system and method for power distribution network construction safety coordinated management control is realized based on artificial intelligence |
CN109240311A (en) * | 2018-11-19 | 2019-01-18 | 国网四川省电力公司电力科学研究院 | Outdoor power field construction operation measure of supervision based on intelligent robot |
-
2019
- 2019-02-23 CN CN201910134732.2A patent/CN110008831A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102262729A (en) * | 2011-08-03 | 2011-11-30 | 山东志华信息科技股份有限公司 | Fused face recognition method based on integrated learning |
CN103235938A (en) * | 2013-05-03 | 2013-08-07 | 北京国铁华晨通信信息技术有限公司 | Method and system for detecting and identifying license plate |
CN107632565A (en) * | 2017-08-28 | 2018-01-26 | 上海欧忆能源科技有限公司 | Power construction building site intellectualized management system and method |
CN108833831A (en) * | 2018-06-15 | 2018-11-16 | 陈在新 | A kind of power construction intelligent safety monitor system |
CN109034535A (en) * | 2018-06-21 | 2018-12-18 | 中国化学工程第六建设有限公司 | Construction site wisdom monitoring method, system and computer readable storage medium |
CN109117827A (en) * | 2018-09-05 | 2019-01-01 | 武汉市蓝领英才科技有限公司 | Work clothes work hat wearing state automatic identifying method and alarm system based on video |
CN109218673A (en) * | 2018-09-20 | 2019-01-15 | 国网江苏省电力公司泰州供电公司 | The system and method for power distribution network construction safety coordinated management control is realized based on artificial intelligence |
CN109215155A (en) * | 2018-09-29 | 2019-01-15 | 东莞方凡智能科技有限公司 | A kind of building site management system based on technology of Internet of things |
CN109240311A (en) * | 2018-11-19 | 2019-01-18 | 国网四川省电力公司电力科学研究院 | Outdoor power field construction operation measure of supervision based on intelligent robot |
Non-Patent Citations (2)
Title |
---|
富吉勇: ""基于全方位视觉的遗留物及其放置者检测的研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
陈曦: ""面向大型工地的视觉监管***关键技术研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11346938B2 (en) | 2019-03-15 | 2022-05-31 | Msa Technology, Llc | Safety device for providing output to an individual associated with a hazardous environment |
CN110428587A (en) * | 2019-07-19 | 2019-11-08 | 国网安徽省电力有限公司建设分公司 | A kind of engineering site early warning interlock method and system |
CN110533811A (en) * | 2019-08-28 | 2019-12-03 | 深圳市万睿智能科技有限公司 | The method and device and system and storage medium of safety cap inspection are realized based on SSD |
CN110517461A (en) * | 2019-08-30 | 2019-11-29 | 成都智元汇信息技术股份有限公司 | A method of it prevents people from carrying package and escapes safety check |
CN110930624A (en) * | 2019-12-06 | 2020-03-27 | 深圳北斗国芯科技有限公司 | Safety in production monitored control system based on big dipper |
CN112669505A (en) * | 2019-12-16 | 2021-04-16 | 丰疆智能科技股份有限公司 | Integrated management system for shower and entrance guard of farm and management method thereof |
CN111083441A (en) * | 2019-12-18 | 2020-04-28 | 广州穗能通能源科技有限责任公司 | Construction site monitoring method and device, computer equipment and storage medium |
CN112949367A (en) * | 2020-07-07 | 2021-06-11 | 南方电网数字电网研究院有限公司 | Method and device for detecting color of work clothes based on video stream data |
CN111898514A (en) * | 2020-07-24 | 2020-11-06 | 燕山大学 | Multi-target visual supervision method based on target detection and action recognition |
CN111898514B (en) * | 2020-07-24 | 2022-10-18 | 燕山大学 | Multi-target visual supervision method based on target detection and action recognition |
CN112071006A (en) * | 2020-09-11 | 2020-12-11 | 湖北德强电子科技有限公司 | High-efficiency low-resolution image area intrusion recognition algorithm and device |
CN112466086A (en) * | 2020-10-26 | 2021-03-09 | 福州微猪信息科技有限公司 | Visual identification early warning device and method for farm work clothes |
CN112532927A (en) * | 2020-11-17 | 2021-03-19 | 南方电网海南数字电网研究院有限公司 | Intelligent safety management and control system for construction site |
CN112528821A (en) * | 2020-12-06 | 2021-03-19 | 杭州晶一智能科技有限公司 | Pedestrian crosswalk pedestrian detection method based on motion detection |
CN112883969A (en) * | 2021-03-01 | 2021-06-01 | 河海大学 | Rainfall intensity detection method based on convolutional neural network |
CN112883969B (en) * | 2021-03-01 | 2022-08-26 | 河海大学 | Rainfall intensity detection method based on convolutional neural network |
CN113379144A (en) * | 2021-06-24 | 2021-09-10 | 深圳开思信息技术有限公司 | Store purchase order generation method and system for online automobile distribution purchase platform |
CN113408501A (en) * | 2021-08-19 | 2021-09-17 | 北京宝隆泓瑞科技有限公司 | Oil field park detection method and system based on computer vision |
CN113977603A (en) * | 2021-10-29 | 2022-01-28 | 连云港福润食品有限公司 | Monitoring robot based on target detection, identification and tracking for worker production specification |
CN115925243A (en) * | 2022-12-24 | 2023-04-07 | 山西百澳智能玻璃股份有限公司 | Method and system for regulating and controlling heating temperature of glass |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110008831A (en) | A kind of Intellectualized monitoring emerging system based on computer vision analysis | |
CN108596277B (en) | Vehicle identity recognition method and device and storage medium | |
CN109271554B (en) | Intelligent video identification system and application thereof | |
CN102521565B (en) | Garment identification method and system for low-resolution video | |
CN101715111B (en) | Method for automatically searching abandoned object in video monitoring | |
CN106778684A (en) | deep neural network training method and face identification method | |
CN110569772A (en) | Method for detecting state of personnel in swimming pool | |
CN109190475A (en) | A kind of recognition of face network and pedestrian identify network cooperating training method again | |
CN110378179A (en) | Subway based on infrared thermal imaging is stolen a ride behavioral value method and system | |
CN110619277A (en) | Multi-community intelligent deployment and control method and system | |
CN112115761B (en) | Countermeasure sample generation method for detecting vulnerability of visual perception system of automatic driving automobile | |
CN104143077B (en) | Pedestrian target search method and system based on image | |
JP4975801B2 (en) | Monitoring method and monitoring apparatus using hierarchical appearance model | |
CN113989858B (en) | Work clothes identification method and system | |
CN103996045A (en) | Multi-feature fused smoke identification method based on videos | |
CN112966736B (en) | Vehicle re-identification method based on multi-view matching and local feature fusion | |
CN105005773A (en) | Pedestrian detection method with integration of time domain information and spatial domain information | |
CN114612823A (en) | Personnel behavior monitoring method for laboratory safety management | |
CN111507320A (en) | Detection method, device, equipment and storage medium for kitchen violation behaviors | |
CN112434545A (en) | Intelligent place management method and system | |
CN110222735A (en) | A kind of article based on neural network and background modeling is stolen to leave recognition methods | |
CN115861940A (en) | Working scene behavior evaluation method and system based on human body tracking and recognition technology | |
CN109101925A (en) | Biopsy method | |
CN109284759A (en) | One kind being based on the magic square color identification method of support vector machines (svm) | |
CN109672847A (en) | Intelligent safety defense monitoring system based on image recognition technology |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190712 |
|
WD01 | Invention patent application deemed withdrawn after publication |