CN110414342A - A kind of movement human detection recognition method based on video image processing technology - Google Patents
A kind of movement human detection recognition method based on video image processing technology Download PDFInfo
- Publication number
- CN110414342A CN110414342A CN201910548271.3A CN201910548271A CN110414342A CN 110414342 A CN110414342 A CN 110414342A CN 201910548271 A CN201910548271 A CN 201910548271A CN 110414342 A CN110414342 A CN 110414342A
- Authority
- CN
- China
- Prior art keywords
- video image
- safety cap
- video
- file
- processing technology
- 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
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- 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
- G06V10/30—Noise filtering
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The movement human detection recognition method based on video image processing technology that the present invention relates to a kind of, which comprises the steps of: operation field monitoring system obtains real time video image information, by coding compression storage into file；From monitoring system video recording, the video file newly recorded, the input as operation field staff's identifying system are searched；The video image decoding in video file for newly being recorded monitoring system using decoding tool is the data of yuv format, is stored under specified catalogue；Motion detection is responsible for for being detected and analyzed the moving region in video image；Identify whether moving region is normal staff；For the video image of non-working person, it is encoded to jpeg format, and passes through network transmission to data center's storage system；Data center's storage system is responsible for storing video image file into system, while the storage of the relevant information of image into database；The information such as the description information of database purchase image/video and user's login.
Description
Technical field
The present invention relates to a kind of movement human detection recognition methods more particularly to a kind of based on video image processing technology
Movement human detection recognition method.
Background technique
Electric network operation scene type multiplicity, different work scene rush to repair, test, in maintenance process by different professionals
Be responsible for, for safety without minor matter, safety is greater than day, production safety risk management and control become in power grid routine work most important work it
One.However in routine work, working region range is big, and operation is many and diverse, and personnel are numerous, and management personnel is limited, part work
It is not high to make personnel's consciousness, awareness of safety is inadequate, off one's guard to operation field security risk, causes habituation is violating the regulations to prohibit repeatly not
Only, therefore probably it inadvertently causes the accident, threatens human life's safety, threaten safe operation of electric network.How
Improving becomes problem of the pendulum in face of administrative staff for the control of operation field personnel.
Currently, players based on video image detect identification, in computer vision using very extensive, including
In nurse, safety detection etc. at home.By video image identification technology, the identification degree of monitoring can be enhanced, mitigate monitor
The workload of member.The fast development of video image analysis technology in recent years, gradually incorporates daily life in a variety of manners
In the middle, wherein judging that the analysis of special group personnel is increasingly taken seriously by personnel characteristics.
This programme knows moving target therein by the video image of analysis electric power video monitoring system acquisition
Not, and personnel are worked normally to electric operating scene to analyze and determine, the purpose identified to staff is completed, to assist protecting
Hinder production safety.
Summary of the invention
The movement human detection based on video image processing technology that technical problem to be solved by the invention is to provide a kind of
Recognition methods.
To solve the above problems, the technical solution used in the present invention is:
A kind of movement human detection recognition method based on video image processing technology comprising following steps:
The real time video image information that operation field monitoring system is obtained, by coding compression storage into file；
From monitoring system video recording, the video file newly recorded, the input as operation field staff's identifying system are searched；
The video image decoding in video file for newly being recorded monitoring system using decoding tool, is stored in specified catalogue
Under；
The moving region in video image is detected and analyzed by motion state detection；
Identify whether moving region is normal staff；
For the video image of non-working person, and pass through network transmission to data center's storage system；
Data center's storage system stores video image file into system, while the relevant information of image storage to data
In library.
Further, from monitoring system video recording, the video file newly recorded is searched, is known as operation field staff
The input of other system is realized especially by such as under type:
(1) timed software at regular intervals, such as 5 minutes, checks the file in the catalogue where the video recording of monitoring system；
(2) the generation time for judging file, when the file generated time being greater than the time that last time checks, then it is assumed that newly record
Video file.
Further, the moving region in video image is detected and analyzed by motion state detection, it is specific logical
Under type such as is crossed to realize:
(1) gray level image of Y data is extracted；
(2) processing of noise is removed to gray level image；
(3) moving region in video image is detected using motion detection algorithm；
(4) moving region is identified.
Further, whether identification moving region is normal staff, is realized especially by such as under type:
(1) by human body proportion, confirm the position of human body head；
(2) characteristics of human body of staff is extracted；
(3) characteristics of human body of the staff of extraction is compared in feature templates library, and if a certain template matches；
(4) if match hit, normal work personnel are identified as.
Further, the human body proportion includes, and the height on head is 0.1-0.2 times of whole body height, and preferably 0.137
Times, the oxter width on both sides is 0.2-0.3 times of whole body height, and preferably 0.224 times, be whole body by the height of foot to waist
0.4-0.5 times of height, preferably 0.435 times.
Further, the characteristics of human body that personnel are worked normally using safety cap as operation field, passes through operation field video
In human body head safety cap, safety cap color characteristic information is extracted, and according to these color characteristics of safety cap, by main body
Then feature and template are carried out matching comparison into specified color template classification by color classification.
Further, whether identification moving region is normal staff, is realized especially by such as under type:
Step 1: on the basis of motion detection, according to human body proportion, carrying out head position judgement；
Step 2: the ratio on the head according to shared by safety cap obtains the position of safety cap；
Step 3: matching being compared by histogram mode, and storage template, carries out the identification of safety cap；
Step 4: if match hit, being identified as normal work personnel.
Further, on the basis of motion detection, according to human body proportion, head position judgement is carried out, especially by such as
Under type is realized:
(1) in detecting the gray level image after the binaryzation of moving region, the height of moving object is calculated, in vertical direction
On, first point of moving region is denoted as initial position, the last one point of moving region is denoted as end position, the difference between the two conduct
The height of moving region；
(2) by the ratio of human body shared by head, the position on head is calculated.
Further, the ratio on the head according to shared by safety cap obtains the position of safety cap, especially by such as under type reality
It is existing:
(1) in the head position being had determined in step 1, the ratio on the head according to shared by safety cap, positioning security cap
Region；
(2) upright position of safety cap is identified in gray level image；
(3) in the region of upright position, the width of moving object is defined as the width of safety cap；
(4) location of pixels for the point that gray value is 255 in record security cap region.
Further, it is compared by histogram mode, and storage template, carries out the identification of safety cap, especially by
As under type is realized:
(1) image of YUV format corresponding with motion detection is converted into RGB picture format；
(2) on RGB image, each color value of RGB of corresponding safety cap region all pixels point is taken out；
(3) histogram of tri- kinds of color values of RGB of safety cap region all pixels point is calculated；
(4) maximum value in three kinds of color value histograms is taken out；
(5) it is compared using three maximum values and the safety cap color template deposited；
(6) it in a template if there is match hit, is denoted as and works normally personnel safety cap；Otherwise non-normal working personnel are denoted as.
Further, safety cap accounts for about 1/2 or so of entire head.
Further, the database is used to store the description information and user login information of image/video.
Further, the video image decoding in video file newly recorded monitoring system using decoding tool, decoding
Data format afterwards is the data of yuv format.
Further, for the video image of non-working person, coded format is jpeg format.
The beneficial effects of adopting the technical scheme are that
Method provided by the present invention, can be by the video image of analysis electric power video monitoring system acquisition, to movement therein
Target is identified, and is worked normally personnel to electric operating scene and analyzed and determined, the mesh identified to staff is completed
, to assist ensureing production safety.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, can also be obtained according to these attached drawings other attached drawings.
Fig. 1 is the basic procedure schematic diagram that moving object detection is carried out in video image.
Fig. 2 is the schematic diagram of expansive working in moving target detecting method.
Fig. 3 is the operation result schematic diagram of expansive working.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, invention is carried out combined with specific embodiments below
Clear, complete description, it is to be understood that term " center ", " vertical ", " transverse direction ", "upper", "lower", "front", "rear",
The orientation or positional relationship of the instructions such as "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outside" is based on attached drawing institute
The orientation or positional relationship shown, is merely for convenience of description of the present invention and simplification of the description, rather than the dress of indication or suggestion meaning
It sets or element must have a particular orientation, be constructed and operated in a specific orientation, therefore should not be understood as to limit of the invention
System.
A kind of movement human detection recognition method based on video image processing technology comprising following steps:
Step 1: operation field monitoring system obtains real time video image information, by coding compression storage into file.
Step 2: from monitoring system video recording, searching the video file newly recorded, identified as operation field staff
The input of system.This step is decomposed as follows:
(1) timed software at regular intervals, such as 5 minutes, checks the file in the catalogue where the video recording of monitoring system；
(2) the generation time of file is judged, if the file generated time is greater than the time of last time inspection, then it is assumed that be new video recording text
Part；
(3) new file will enter step 3 and handle.
Step 3: the video image decoding in video file for newly being recorded monitoring system using decoding tool is YUV lattice
The data of formula are stored under specified catalogue.
Step 4: motion detection is responsible for for being detected and analyzed the moving region in video image.Motion detection it is detailed
Thin step is decomposed as follows:
(1) gray level image of Y data is extracted；
(2) processing of noise is removed to gray level image；
(3) moving region in video image is detected using motion detection algorithm；
(4) moving region is identified.
Step 5: whether identification moving region is normal staff.The step of staff identifies is decomposed as follows:
(1) by human body proportion, confirm the position of human body head；
(2) characteristics of human body of staff is extracted；
(3) characteristics of human body of extraction is compared in feature templates library, and if a certain template matches；
(4) if match hit, normal work personnel are identified as.
Step 6: for the video image of non-working person, being encoded to JPEG format, and pass through network transmission to data
Central store system.
Step 7: data center's storage system is responsible for storing video image file into system, while the phase of image
Information storage is closed into database.
Step 8: the information such as the description information of database purchase image/video and user's login.
In Video Image processing, the method encoded to color of image is referred to as " color space " or " color
Domain ", there are many kinds of the formats for describing video image color space, and common color space format includes RGB, YUV etc..Reality
In life, the color space of any color all may be defined to a fixed number or variable.
YUV is a kind of colour coding method used by eurovision system, is that simulation color television system uses earliest
Color model.Wherein " Y " indicates brightness (Luminance or Luma), that is, grayscale value, and what " U " was indicated is then
Coloration (Chrominance or Chroma), " V " indicate the saturation degree (Saturation) of color.
Each data value of YUV can be established by RGB input signal, and wherein Y data are by the spy of RGB signal
Partial stack is determined to together, U reflects the difference between RGB input signal RED sector and RGB luminance signals value, and
What V reflected is the difference between RGB input signal blue portion and RGB luminance signals value.
It can mutually be converted between YUV and RGB, conversion formula are as follows:
Y=0.257R+0.504G+ 0.098B+16 (2-1)
U=- 0.148R- 0.291G+ 0.439B+128 (2-2)
V=0.439R -0.368G -0.071B+128 (2-3)
YUV color model is mainly used for optimizing the transmission of colour-video signal, makes the old-fashioned black-and-white television of its back compatible.Pass through
Largely it was verified that human eye is significantly larger than the susceptibility of color difference to luminance sensitivity, the benefit of YUV picture format is used in this way
It is that color difference channel can be stored with specific luminance channel using lower sample rate.Therefore in terms of transmission with RGB color model phase
Than lesser space can be used in YUV, and RGB then requires three independent vision signal simultaneous transmissions.
The sample mode of YUV color model usually has following several:
The channel 4:4:4 UV is identical as the channel Y in level sampling and Vertical Sampling.
4:2:2 indicates that UV level sampling is the half of Y level sampling, and Vertical Sampling is constant, i.e., every 4 Y samples
2 UV samples of this correspondence.
4:2:0 indicates UV level sampling and Vertical Sampling is the half of Y, i.e., every 4 Y samples corresponding 1
A UV sample.
Video signal is often subject to the noise jamming of imaging device or external environment in digitlization and transmission process
Deng influence, referred to as noisy image or noise image, the process for reducing these noises in video image is known as image denoising.
According to the relationship of noise and video signal, noise can be divided into three kinds of forms:
(1) this noise like of additive noise is unrelated with inputted video image signal, and noisy image may be expressed as:
F (x, y)=g (x, y)+n (x, y) (2-4)
In formula: f (x, y) indicates noisy image signal；
G (x, y) indicates original image signal；
N (x, y) indicates noise signal.
Additive noise is to be added relationship with picture signal, and regardless of original image signal whether there is, noise is all always existed.
The noise generated when interchannel noise and the camera-scanning image of vidicon just belongs to additive noise.
(2) this noise like of multiplicative noise is related with picture signal, and noisy image may be expressed as:
F (x, y)=g (x, y)+n (x, y) g (x, y) (2-5)
In formula: f (x, y) is represented to the picture signal of Noise；
G (x, y) indicates original image signal；
N (x, y) indicates noise signal.
Multiplicative noise is the relationship of being multiplied with picture signal, and in the presence of picture signal, multiplicative noise exists；Work as picture signal
In the absence of, multiplicative noise is just not present.The noise generated when flying-spot scanner scan image, the grain noise in film all belong to
In multiplicative noise
(3) this noise like of quantizing noise is unrelated with received image signal, is that there are quantization errors for quantizing process, then reflect reception
It holds and generates.
Digital filter is that noise signal is removed from video signal using most methods.Common number filter
Wave device has following several:
(1) mean filter uses field mean value method, the basic principle is that, a template is given for pixel to be processed,
The template includes pixel value neighbouring around it, and the mean value of the ensemble of pixel values in template is substituted to pixel to be processed
Value.Mean filter is suitable for removing the grain noise the image generated due to scanning；Its shortcomings that is image to be made to become
It is fuzzy, the reason is that it treats all pixels point in template using same, while being shared noise, by picture
The boundary of element is also shared.In order to improve the effect of mean filter, filter can be constructed using average weighted mode
Wave device.
(2) median filter is a kind of common Nonlinear Smoothing Filter, the basic principle is that, digital video image
Or in Serial No. certain point value, be replaced with the intermediate value of each point value in a field of the point.Its major function is, right
In pixel to be processed, if it is bigger with the difference of surrounding pixel gray value, the value close with the pixel value of surrounding is taken,
So as to eliminate isolated noise spot, so median filtering is highly effective for the salt-pepper noise for filtering out image.Median filtering
Device can accomplish the edge for not only having removed noise but also capable of utmostly having protected image, to obtain relatively satisfied recovery effect
Fruit, but the more image of the image more to some details, especially point, line, pinnacle details should not use the side of median filtering
Method.
(3) morphology noise filter be by morphological image unlatching and closed procedure combine and filter out making an uproar
Sound.First to there is noise image to carry out opening operation, the process range of opening operation is bigger than the size of noise, to make to open
Result can carry out closed procedure, then making an uproar image by the noise remove in background, followed by the image that obtains after unlatching
Sound removes.According to the characteristics of this method, it can be seen that the applicable image type of Morphologic filters is the object size in image
It is all bigger, and not tiny details, it is relatively preferable to the effect of such scene image partition.
(4) Wavelet Denoising Method filter is a kind of denoising filter of the foundation on wavelet transform base.Its basic principle
It is to have the characteristics that varying strength distribution according to signal and the wavelet decomposition of noise on different frequency bands, it will be corresponding on each frequency band
Wavelet coefficient be removed, retain the wavelet decomposition system of original signal, small echo weight then carried out to treated coefficient again
Structure, to obtain pure signal.It compares with several filters in front, Wavelet Denoising Method filter is to low signal-to-noise ratio
It is relatively more preferable to denoise effect, meanwhile, the denoising effect of time varying signal and jump signal is especially apparent.
Expanding (dilation) is one of two basic operations of morphological image, another is etching operation.Expansion
Its definition is: structure B being carried out convolution operation on structure A, if during moving structure B, deposited with structure X
In overlapping region, then the position is recorded, there are the position of intersection, referred to as structure X exists by all moving structure B and structure X
Expansion results under structure B effect.As shown in Figure 2.
In Fig. 2, structure X is processed object, and B is structural element, is not difficult to know, any one is being tied
Point in structure X, after B hits the point, then X is exactly dash area in figure by the B result expanded.Dash area includes
All ranges of X, just as X expands a circle, this is why the reason of crying expansion.
Actual dilation operation, as shown in Figure 3.The left side is processed video image X, and centre is structural element B.It is swollen
Swollen method is collide by the point on the central point and X of B and around X, if there is a point to fall on B
In the range of X, then the point is just black；The right is the result after expansion.As can be seen that expansion results include all models of X
It encloses.
In video image processing technology, the method for common human testing has:
(1) based on the detection method of skin color image higher for quality, the skin color of human body is one obvious
Characteristics of human body, can be used to carry out the detection of human body.Under normal conditions, face can it is exposed outside, and color relatively come
Say fixation, therefore, the modeling that the colour of skin is carried out by HSV spatial model is a kind of very effective detection mode.In addition, skin
Coloration and saturation degree be illuminated by the light the variation with weather influence it is smaller, therefore skin can with dimensional Gaussian model HSV sky
Between indicate, in order to reduce background color interference, the acquisition of the dimensional Gaussian model mean value, covariance can pass through foreground extraction
Afterwards, then by hand determine that skin color position classifies to region, calculated in assorting process between pixel and mean value away from
From if the distance is less than some preset thresholding, which may be considered skin, be otherwise judged as non-skin.
It after classification, may determine whether with the presence of skin area, if with the presence of skin area in moving object, object can be sentenced
Break as human body, is otherwise judged as non-human.The judgment method has the details of the details of image, especially skin area higher
It is required that therefore, this method carries out human testing for the fuzzy image of remote and image detail, and effect can be poor, or even sentences
Dislocation misses.
(2) the detection method human body based on motion feature can show certain posture with when running in stabilized walking, and
And this posture can periodically occur, and therefore, can use this periodically variable feature and carry out human testing.For example,
State of the object at the t moment is x (t), if there is a constant p, so that:
X (t+p)=x (t)+T (t) (2-5)
In formula: x (t+p) is feature of the object at the t moment；
X (t) is state of the object at the t moment；
T (t) is translational movement.
Then this smallest time interval may be considered the period of motion of the object.During stable motion, human body
Posture can be using the axis of people as reference axis, the symmetry of human body is in cyclically-varying, is constructed in the time domain using this characteristic
Time similar matrix, by the motion feature of the cycle movement of human body variation reflection human body, and using this similar matrix into
The relevant analysis of row is to come out human testing.
3) detection method based on body configuration's feature is compared with other objects, and body configuration has very special ground
Side, because body configuration includes the important information of human testing.It has been deposited according to the algorithm that body configuration carries out human testing
In many research, but the difficulty for analyzing body configuration is, the non-rigid that human body is showed, to portraying human body band
Very big uncertainty is carried out.If this uncertainty is relatively easy, some mathematical models, such as single Gauss can be used
Model or other such as mixed Gauss model are modeled, and parameter prediction and calculating are then carried out.If using non-ginseng
The detection method of numberization needs a large amount of sample then to cover all uncertainties of human body, wherein applying wider distribution
Model is typical imparametrization method, it describes the form of human body, this side using a point orderly, labeled
Method has biggish flexibility, but because brought a large amount of calculating is marked for point set, in addition it is based on complete
The method of office, for blocking situations such as, are difficult to be effectively treated, so not being widely used in actual application.It is based on
The detection method of body local shape, comparatively, data volume to be treated are smaller, while aiding in others and judging hand
Section, obtains relatively broad application during human testing.
In working normally personal identification algorithm, the present embodiment calculates human head location, root by human body proportion method
According to the position on head where safety cap, the position of safety cap is calculated, and works normally personnel using safety cap as operation field
Feature extracts, and tri- kinds of color histograms of RGB of safety cap position are calculated finally by histogram mode, take histogram
Maximum value and safety cap template carry out matching comparison, effectively identify and work normally personnel and improper work in operation field video
Make personnel.
For the position orientation problem of safety cap, first by human body proportion come the position of positioning head, by difference
Human Height proportion grading is statisticallyd analyze by a large amount of human sample, and the height on head is about 0.137 times of whole body height
Meet actual experimental data.
It is analyzed by actual data, calculates ratio of the safety cap in the position on head and show that safety cap accounts for about entire head
1/2 or so of portion.
In moving region, compared according to head, positions human head location.In example image resolution ratio be 352 ×
264, using the upper left corner of image as coordinate origin, calculate the vertical end position on head:
End=(height-begin) × 0.137+begin(4-3)
In formula: begin is head initial position；
Height is the height of whole image.
Head initial position is at the top of safety cap, and according to analysis before, safety cap bottom position is that head is high
The 1/2 of degree, the vertical end position calculation formula of safety cap are as follows:
Mid=(height-begin) × 0.137/2+begin (4-4)
In formula: mid is the vertical end position of safety cap.
In the present embodiment, 264 height, begin value are calculated 91, mid for 64, end and are calculated
It is 77.From moving region, the parts of images under the image and mid line on height line is removed, obtains safety cap place
Region.
In network system, safety cap is divided into stringent wearing grade, and according to the rules, such as white safety cap represents neck
It leads, blue safety cap represents administrative staff, and yellow safety cap represents construction personnel, and red safety cap represents nonnative personnel, orange
Safety cap represents supply or collator.
Safety cap judges principle, judges that safety cap whether there is using histogram (histogram) algorithm.Histogram is again
Claim quality distribution diagram, is a kind of statistical report figure.It, which is provided in a frame video image or one group of video image, possesses given numerical value
Pixel quantity.The histogram of gray level image has 256 entries (or being container), and No. 0 container provides the picture that value is 0
Plain number, No. 1 container provide the number of pixels that value is 1, and so on.So histogram can to a certain extent effectively
Ground describes the content of image, therefore is frequently used for content-based retrieval.
In the present embodiment, matching comparison is carried out for using orange safety cap as safety cap color template.In RGB face
Color defines in table, orange standard value be RGB(255,90,0), it is contemplated that the influence of the environmental factors such as illumination, in conjunction with orange
Color defines table, orange value range extension are as follows: RGB(255,190,160) --- RGB(180,60,0).
The histogram of safety cap region all pixels point is counted, the maximum corresponding color value of value is as peace in histogram
The color value in full cap region.
Judgement to tri- color object of RGB, needs the stereogram using RGB.
Calculate in safety cap region RGB maximum color value in stereogram: RGB (230,130,69) and orange
Value range RGB (255,190,160) --- RGB (180,60,0) is compared:
180 < 230 < 255,60 < 130 < 190,0 < 69 < 160
It is known that RGB (230,130,69) in orange range, i.e. match hit.It is verified in color diagram again: safety
Histogram maximum color value in cap region: the corresponding actual color of RGB (230,130,69) is orange.
The motion detection algorithm is first to carry out background using background differential technique in conjunction with background differential technique and frame difference method
It updates, obtains the complete image of moving object, interference information therein is eliminated in conjunction with frame difference method, to more be added
The algorithm of whole moving object contours.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify to technical solution documented by previous embodiment or equivalent replacement of some of the technical features；And
These are modified or replaceed, the spirit and model of technical solution of the embodiment of the present invention that it does not separate the essence of the corresponding technical solution
It encloses.
Claims (10)
1. a kind of movement human detection recognition method based on video image processing technology characterized by comprising
The real time video image information that operation field monitoring system is obtained, by coding compression storage into file；
From monitoring system video recording, the video file newly recorded, the input as operation field staff's identifying system are searched；
The video image decoding in video file for newly being recorded monitoring system using decoding tool, is stored in specified catalogue
Under；
The moving region in video image is detected and analyzed by motion state detection；
Identify whether moving region is normal staff；
For the video image of non-working person, and pass through network transmission to data center's storage system；
Data center's storage system stores video image file into system, while the relevant information of image storage to data
In library.
2. a kind of movement human detection recognition method based on video image processing technology according to claim 1, special
Sign is, from monitoring system video recording, the video file newly recorded is searched, as the defeated of operation field staff's identifying system
Enter, realized especially by such as under type:
(1) timed software at regular intervals, checks the file in the catalogue where the video recording of monitoring system；
(2) the generation time for judging file, when the file generated time being greater than the time that last time checks, then it is assumed that newly record
Video file.
3. a kind of movement human detection recognition method based on video image processing technology according to claim 1, special
Sign is, is detected and analyzed the moving region in video image by motion state detection, especially by such as under type
It realizes:
(1) gray level image of Y data is extracted；
(2) processing of noise is removed to gray level image；
(3) moving region in video image is detected using motion detection algorithm；
(4) moving region is identified.
4. a kind of movement human detection recognition method based on video image processing technology according to claim 1, special
Sign is whether identification moving region is normal staff, realizes especially by such as under type:
(1) by human body proportion, confirm the position of human body head；
(2) characteristics of human body of staff is extracted；
(3) characteristics of human body of the staff of extraction is compared in feature templates library, and if a certain template matches；
(4) if match hit, normal work personnel are identified as.
5. a kind of movement human detection recognition method based on video image processing technology according to claim 4, special
Sign is that the human body proportion includes that the height on head is 0.1-0.2 times of whole body height, and the oxter width on both sides is whole body
0.2-0.3 times of height, by 0.4-0.5 times that the height of foot to waist is whole body height.
6. detecting identification side according to a kind of any movement human based on video image processing technology of claim 4 or 5
Method, which is characterized in that the characteristics of human body that personnel are worked normally using safety cap as operation field, by operation field video
The safety cap of human body head extracts safety cap color characteristic information, and according to these color characteristics of safety cap, by body color
It is categorized into specified color template classification, feature and template is then subjected to matching comparison.
7. a kind of movement human detection recognition method based on video image processing technology according to claim 1, special
Sign is whether identification moving region is normal staff, realizes especially by such as under type:
Step 1: on the basis of motion detection, according to human body proportion, carrying out head position judgement；
Step 2: the ratio on the head according to shared by safety cap obtains the position of safety cap；
Step 3: matching being compared by histogram mode, and storage template, carries out the identification of safety cap；
Step 4: if match hit, being identified as normal work personnel.
8. a kind of movement human detection recognition method based on video image processing technology according to claim 7, special
Sign is, on the basis of motion detection, according to human body proportion, carries out head position judgement, realizes especially by such as under type:
(1) in detecting the gray level image after the binaryzation of moving region, the height of moving object is calculated, in vertical direction
On, first point of moving region is denoted as initial position, the last one point of moving region is denoted as end position, the difference between the two conduct
The height of moving region；
(2) by the ratio of human body shared by head, the position on head is calculated.
9. a kind of movement human detection recognition method based on video image processing technology according to claim 7, special
Sign is that the ratio on the head according to shared by safety cap obtains the position of safety cap, realizes especially by such as under type:
(1) in the head position being had determined in step 1, the ratio on the head according to shared by safety cap, positioning security cap
Region；
(2) upright position of safety cap is identified in gray level image；
(3) in the region of upright position, the width of moving object is defined as the width of safety cap；
(4) location of pixels for the point that gray value is 255 in record security cap region.
10. a kind of movement human detection recognition method based on video image processing technology according to claim 7, special
Sign is, is compared by histogram mode, and storage template, carries out the identification of safety cap, especially by such as under type reality
It is existing:
(1) image of YUV format corresponding with motion detection is converted into RGB picture format；
(2) on RGB image, each color value of RGB of corresponding safety cap region all pixels point is taken out；
(3) histogram of tri- kinds of color values of RGB of safety cap region all pixels point is calculated；
(4) maximum value in three kinds of color value histograms is taken out；
(5) it is compared using three maximum values and the safety cap color template deposited；
(6) it in a template if there is match hit, is denoted as and works normally personnel safety cap；Otherwise non-normal working personnel are denoted as.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910548271.3A CN110414342A (en) | 2019-06-24 | 2019-06-24 | A kind of movement human detection recognition method based on video image processing technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910548271.3A CN110414342A (en) | 2019-06-24 | 2019-06-24 | A kind of movement human detection recognition method based on video image processing technology |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110414342A true CN110414342A (en) | 2019-11-05 |
Family
ID=68359602
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910548271.3A Pending CN110414342A (en) | 2019-06-24 | 2019-06-24 | A kind of movement human detection recognition method based on video image processing technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110414342A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111508038A (en) * | 2020-04-17 | 2020-08-07 | 北京百度网讯科技有限公司 | Image processing method, image processing device, electronic equipment and computer readable storage medium |
| CN112488031A (en) * | 2020-12-11 | 2021-03-12 | 华能华家岭风力发电有限公司 | Safety helmet detection method based on color segmentation |
| CN113688301A (en) * | 2021-08-24 | 2021-11-23 | 杨成兴 | Online accurate teaching tutoring method and system based on artificial intelligence and big data |
| CN116189089A (en) * | 2023-02-14 | 2023-05-30 | 深圳市巨龙创视科技有限公司 | Intelligent video monitoring method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105469069A (en) * | 2015-12-09 | 2016-04-06 | 上海工业自动化仪表研究院 | Safety helmet video detection method for production line data acquisition terminal |
| CN106503716A (en) * | 2016-09-13 | 2017-03-15 | 中国电力科学研究院 | A kind of safety cap recognition methods that is extracted based on color and contour feature and system |
| CN107545224A (en) * | 2016-06-29 | 2018-01-05 | 珠海优特电力科技股份有限公司 | The method and device of transformer station personnel Activity recognition |
| CN107679524A (en) * | 2017-10-31 | 2018-02-09 | 天津天地伟业信息***集成有限公司 | A kind of detection method of the safety cap wear condition based on video |
| CN108460358A (en) * | 2018-03-20 | 2018-08-28 | 武汉倍特威视***有限公司 | Safety cap recognition methods based on video stream data |
| CN108921004A (en) * | 2018-04-27 | 2018-11-30 | 淘然视界(杭州)科技有限公司 | Safety cap wears recognition methods, electronic equipment, storage medium and system |
-
2019
- 2019-06-24 CN CN201910548271.3A patent/CN110414342A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105469069A (en) * | 2015-12-09 | 2016-04-06 | 上海工业自动化仪表研究院 | Safety helmet video detection method for production line data acquisition terminal |
| CN107545224A (en) * | 2016-06-29 | 2018-01-05 | 珠海优特电力科技股份有限公司 | The method and device of transformer station personnel Activity recognition |
| CN106503716A (en) * | 2016-09-13 | 2017-03-15 | 中国电力科学研究院 | A kind of safety cap recognition methods that is extracted based on color and contour feature and system |
| CN107679524A (en) * | 2017-10-31 | 2018-02-09 | 天津天地伟业信息***集成有限公司 | A kind of detection method of the safety cap wear condition based on video |
| CN108460358A (en) * | 2018-03-20 | 2018-08-28 | 武汉倍特威视***有限公司 | Safety cap recognition methods based on video stream data |
| CN108921004A (en) * | 2018-04-27 | 2018-11-30 | 淘然视界(杭州)科技有限公司 | Safety cap wears recognition methods, electronic equipment, storage medium and system |
Non-Patent Citations (3)
| Title |
|---|
| JIE LI: "Safety helmet wearing detection based on image processing and machine learning", 《2017 NINTH INTERNATIONAL CONFERENCE ON ADVANCED COMPUTATIONAL INTELLIGENCE (ICACI)》 * |
| 毕林: "基于卷积神经网络的矿工安全帽佩戴识别研究", 《万方数据知识服务平台》 * |
| 郝存明: "基于深度学***台》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111508038A (en) * | 2020-04-17 | 2020-08-07 | 北京百度网讯科技有限公司 | Image processing method, image processing device, electronic equipment and computer readable storage medium |
| CN112488031A (en) * | 2020-12-11 | 2021-03-12 | 华能华家岭风力发电有限公司 | Safety helmet detection method based on color segmentation |
| CN113688301A (en) * | 2021-08-24 | 2021-11-23 | 杨成兴 | Online accurate teaching tutoring method and system based on artificial intelligence and big data |
| CN116189089A (en) * | 2023-02-14 | 2023-05-30 | 深圳市巨龙创视科技有限公司 | Intelligent video monitoring method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110414342A (en) | A kind of movement human detection recognition method based on video image processing technology | |
| CN105160297B (en) | Masked man's event automatic detection method based on features of skin colors | |
| CN106408846A (en) | Image fire hazard detection method based on video monitoring platform | |
| CN111275696B (en) | Medical image processing method, image processing method and device | |
| CN104298996B (en) | A kind of underwater active visual tracking method applied to bionic machine fish | |
| CN107657619B (en) | A kind of low-light (level) Forest fire image dividing method | |
| CN109427055B (en) | Remote sensing image sea surface ship detection method based on visual attention mechanism and information entropy | |
| CN103942539B (en) | A kind of oval accurate high efficiency extraction of head part and masking method for detecting human face | |
| CN105184216A (en) | Cardiac second region palm print digital extraction method | |
| CN105893946A (en) | Front face image detection method | |
| CN101635835A (en) | Intelligent video monitoring method and system thereof | |
| CN101359365A (en) | Iris positioning method based on Maximum between-Cluster Variance and gray scale information | |
| CN109784148A (en) | Biopsy method and device | |
| CN108009479A (en) | Distributed machines learning system and its method | |
| Monwar et al. | Pain recognition using artificial neural network | |
| CN107222660A (en) | A kind of distributed network visual monitor system | |
| CN109259528A (en) | A kind of home furnishings intelligent mirror based on recognition of face and skin quality detection | |
| CN114898261A (en) | Sleep quality assessment method and system based on fusion of video and physiological data | |
| CN115937237A (en) | Local feature extraction method based on edge transform domain | |
| KR100755800B1 (en) | Face detector and detecting method using facial color and adaboost | |
| Yusuf et al. | Human face detection using skin color segmentation and watershed algorithm | |
| CN109359530A (en) | Intelligent video monitoring method and device | |
| CN102542304B (en) | Region segmentation skin-color algorithm for identifying WAP (Wireless Application Protocol) mobile porn image | |
| CN107707975A (en) | Video intelligent clipping method based on monitor supervision platform | |
| CN109948570B (en) | Real-time detection method for unmanned aerial vehicle in dynamic environment |
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 |