CN110348327A - Realize the method and device that Articles detecting is left in monitoring scene - Google Patents
Realize the method and device that Articles detecting is left in monitoring scene Download PDFInfo
- Publication number
- CN110348327A CN110348327A CN201910547197.3A CN201910547197A CN110348327A CN 110348327 A CN110348327 A CN 110348327A CN 201910547197 A CN201910547197 A CN 201910547197A CN 110348327 A CN110348327 A CN 110348327A
- Authority
- CN
- China
- Prior art keywords
- pixel
- picture
- single frames
- monitoring
- article
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30242—Counting objects in image
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The embodiment provides a kind of realize, and the method and device of Articles detecting is left in monitoring scene.This method comprises: putting the background model for obtaining co-located pixels point in previous frame picture pixel-by-pixel for the monitoring single frames picture in monitor video;According to the background model of acquisition, the sport foreground pixel in the monitoring single frames picture is calculated, includes the sport foreground pixel and static background pixel in the monitoring single frames picture;It is positioned in the monitoring single frames picture by the sport foreground pixel and leaves article candidate region;Article candidate region is left according to the organism historical track filtering in the monitor video, the region that article is stopped is left in acquisition.The technical solution of the embodiment of the present invention, which can be realized, accurately detects the article of leaving in monitoring scene.
Description
Technical field
The present invention relates to technical field of video monitoring, in particular to residue product examine in a kind of realization monitoring scene
The method and device of survey.
Background technique
Passageway for fire apparatus refers to that fire fighter implements the channel of rescue and trapped person's evacuation, such as stairs port and passageway,
Occur to play the role of when various dangerous situations immeasurable, it is therefore desirable to guarantee that the passageway for fire apparatus moment keeps unimpeded.
Passageway for fire apparatus is typically provided with monitoring camera, and deploying has the Security Personnel of patent to monitor monitor video in real time,
In order to find and clear up in time stacked in passageway for fire apparatus leave article.
And in the scene that some passageways for fire apparatus monitor automatically, usually using artificial nerve network model to passageway for fire apparatus pair
The monitor video answered is monitored, and automatic alarm when leaving article is stacked in monitoring passageway for fire apparatus, and notify relevant people
Article is left in member's cleaning, there is no need to deploy monitoring personnel, saves security cost.
But in training artificial nerve network model, need to obtain in a large amount of monitor videos and calibration monitor video
What is occurred leaves article as training data, causes training process very many and diverse.And very in view of actual monitoring scene
Ambient lighting variation under complexity, such as different passageways for fire apparatus acutely stacks and leaves many kinds of of article, it is difficult to train
It is suitble to the artificial nerve network model of various complex scenes, causes through artificial nerve network model to residue in passageway for fire apparatus
The detection of product is easy to appear situations such as missing inspection and erroneous detection.
It would therefore be highly desirable to which solve can not be to the skill leaving article and accurately being detected occurred in monitor video in existing realization
Art problem.
Summary of the invention
In order to solve the above-mentioned technical problem, the embodiment provides residue product examines in a kind of realization monitoring scene
The method and device of survey, electronic equipment, computer readable storage medium.
Wherein, the technical scheme adopted by the invention is as follows:
It is a kind of to realize the method that Articles detecting is left in monitoring scene, comprising: for the monitoring single frames figure in monitor video
Piece puts the background model for obtaining co-located pixels point in previous frame picture pixel-by-pixel, and the background model on airspace for simulating institute
State the background pixel variation of co-located pixels point in previous frame picture;According to the background model of acquisition, it is single to calculate the monitoring
Sport foreground pixel in frame picture, the monitoring single frames picture include the sport foreground pixel and static background pixel;By
The sport foreground pixel positions in the monitoring single frames picture and leaves article candidate region;According in the monitor video
Article candidate region is left described in the filtering of organism historical track, the region that article is stopped is left in acquisition.
It is a kind of to realize the device that Articles detecting is left in monitoring scene, comprising: background model obtains module, for monitoring
Monitoring single frames picture in video is put pixel-by-pixel obtains background model of the co-located pixels point on airspace in previous frame picture, described
Background model is used to simulate the background pixel variation of co-located pixels point in the previous frame picture;Sport foreground computing module is used
The sport foreground pixel in the monitoring single frames picture is calculated in the relatively described background model, the monitoring single frames picture includes institute
State sport foreground pixel and static background pixel;Candidate region locating module is used for by the sport foreground pixel in the prison
Article candidate region is left in positioning in control single frames picture;Candidate region filtering module, for according to the life in the monitor video
Article candidate region is left described in the filtering of object historical track, the region that article is stopped is left in acquisition.
A kind of electronic equipment, including processor and memory are stored with computer-readable instruction on the memory, described
The method realized leave Articles detecting in monitoring scene as described above is realized when computer-readable instruction is executed by the processor.
A kind of computer readable storage medium, is stored thereon with computer-readable instruction, when the computer-readable instruction
When being executed by the processor of computer, computer is made to execute the method realized leave Articles detecting in monitoring scene as described above.
In the above-mentioned technical solutions, it is contemplated that leaving article is to carry via the mankind or other organisms to monitoring scene
In, it leaves article and shows as being the image-region moved in monitor video, therefore by the monitoring list in monitor video
Frame picture puts the background model for obtaining co-located pixels point in previous frame picture pixel-by-pixel, and it is single then to calculate monitoring with respect to background model
Sport foreground pixel in frame picture can obtain the pixel that may indicate to leave article in monitoring single frames picture, thus fixed
Position obtains leaving article candidate region.
Being still based on and leaving article is to carry via organism to monitoring scene, and the area peripheral edge for leaving article appearance is bound to
There is organism historical track, therefore was carried out according to the organism historical track in monitor video to article candidate region is left
Filter, obtained article candidate region of leaving is accurately to leave article dwell regions.
Compared with prior art, above-mentioned technical proposal Articles detecting and leaves the type of article to leaving in monitor video
It is unrelated, be filtered according to the organism historical track in monitor video to article candidate region is left, can also filter out by
Change in ambient lighting and article candidate region is left by erroneous detection, to realize the accurate detection to article is left.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
It can the limitation present invention.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows and meets implementation of the invention
Example, and be used to explain the principle of the present invention together with specification.
Fig. 1 is the schematic diagram of implementation environment according to the present invention;
Fig. 2 is a kind of block diagram of computer equipment shown according to an exemplary embodiment;
Fig. 3 is shown according to an exemplary embodiment a kind of to realize the stream that the method for Articles detecting is left in monitoring scene
Cheng Tu;
Fig. 4 is that show according to another exemplary embodiment a kind of realizes the method that Articles detecting is left in monitoring scene
Flow chart;
Fig. 5 is to step 210 in Fig. 3 corresponding embodiment in the flow chart of one embodiment;
Fig. 6 is to step 230 in Fig. 3 corresponding embodiment in the flow chart of one embodiment;
Fig. 7 is to step 230 in Fig. 3 corresponding embodiment in the flow chart of another embodiment;
Fig. 8 is to step 230 in Fig. 3 corresponding embodiment in the flow chart of another embodiment;
Fig. 9 is that show according to another exemplary embodiment a kind of realizes the method that Articles detecting is left in monitoring scene
Flow chart;
Figure 10 is that show according to another exemplary embodiment a kind of realizes the method that Articles detecting is left in monitoring scene
Flow chart;
Figure 11 is that show according to another exemplary embodiment a kind of realizes the method that Articles detecting is left in monitoring scene
Flow chart;
Figure 12 is a kind of specific implementation signal for realizing the method that Articles detecting is left in monitoring scene in an application scenarios
Figure;
Figure 13 is shown according to an exemplary embodiment a kind of to realize the device that Articles detecting is left in monitoring scene
Block diagram.
Specific embodiment
Here will the description is performed on the exemplary embodiment in detail, the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistented with the present invention.On the contrary, they be only with it is such as appended
The example of device and method being described in detail in claims, some aspects of the invention are consistent.
Fig. 1 is a kind of schematic diagram for realizing implementation environment involved in the method for leaving Articles detecting in monitoring scene.It should
Implementation environment includes computer equipment 100 and monitoring camera 200.
Wherein, monitoring camera 200 is installed on passageway for fire apparatus or any other ground for carrying out leaving article monitoring
Point, to be monitored in real time and obtain monitor video.Under normal conditions, monitoring place is equipped with multiple 200 (figures of monitoring camera
4 are shown in 1), to carry out monitoring in all directions.
Monitoring camera 200 is established between computer equipment 100 in advance communication connection, with the monitor video that will acquire
It is transmitted to the detection for carrying out leaving article in computer equipment 100.
In one embodiment, computer equipment 100 detects when occurring leaving article in monitor video, can be to residue
The region respective markers that product occur, and the article region of leaving of label is shown, it is accurately positioned and leaves convenient for related personnel
Article is located at the position in monitoring place, to cleared up in time article is left.
In a further embodiment, computer equipment 100 detects in monitor video is also reported when there is leaving article
It is alert, to notify related personnel clears up in time to leave article.
Fig. 2 is a kind of block diagram of computer equipment shown according to an exemplary embodiment.
It should be noted that the computer equipment, which is one, adapts to example of the invention, must not believe that there is provided
To any restrictions of use scope of the invention.The computer equipment can not be construed to the figure that needs to rely on or must have
One or more component in illustrative computer equipment 100 shown in 2.
As shown in Fig. 2, computer equipment 100 includes processing component 101, memory 102, power supply module 103, multimedia group
Part 104, audio component 105, sensor module 107 and communication component 108.Wherein, said modules and be not all it is necessary, calculate
Machine equipment 100 can increase other assemblies according to itself functional requirement or reduce certain components, and this embodiment is not limited.
Processing component 101 usually control computer equipment 100 integrated operation, such as with display, data communication and day
Associated operation of will data processing etc..Processing component 101 may include one or more processors 109 to execute instruction, with
Complete all or part of the steps of aforesaid operations.In addition, processing component 101 may include one or more modules, convenient for processing
Interaction between component 101 and other assemblies.For example, processing component 101 may include multi-media module, to facilitate multimedia group
Interaction between part 104 and processing component 101.
Memory 102 is configured as storing various types of data to support the operation in computer equipment 100, these are counted
According to example include any application or method for being operated in computer equipment 100 instruction.It is deposited in memory 102
One or more modules are contained, which is configured to be executed by the one or more processors 109, to complete
The all or part of step in the method for Articles detecting is left in realization monitoring scene described in following embodiments.
Power supply module 103 provides electric power for the various assemblies of computer equipment 100.Power supply module 103 may include power supply
Management system, one or more power supplys and other with for computer equipment 100 generate, manage, and distribute associated group of electric power
Part.
Multimedia component 104 includes the screen of one output interface of offer between the computer equipment 100 and user
Curtain.In some embodiments, screen may include TP (Touch Panel, touch panel) and LCD (Liquid Crystal
Display, liquid crystal display).If screen includes touch panel, screen may be implemented as touch screen, be used by oneself with receiving
The input signal at family.Touch panel includes one or more touch sensors to sense the hand on touch, slide, and touch panel
Gesture.The touch sensor can not only sense the boundary of a touch or slide action, but also detect and the touch or sliding
Operate relevant duration and pressure.
Audio component 105 is configured as output and/or input audio signal.For example, audio component 105 includes a Mike
Wind, when computer equipment 100 is in operation mode, when such as call mode, recording mode, and voice recognition mode, microphone is matched
It is set to reception external audio signal.The received audio signal can be further stored in memory 102 or via communication set
Part 108 is sent.In some embodiments, audio component 105 further includes a loudspeaker, is used for output audio signal.
Sensor module 107 includes one or more sensors, for providing the shape of various aspects for computer equipment 10
State assessment.For example, sensor module 107 can detecte computer equipment 100 open/close state, meter can also be detected
Calculate the temperature change of machine equipment 100.
Communication component 108 is configured to facilitate the logical of wired or wireless way between computer equipment 100 and other equipment
Letter.Computer equipment 100 can access the wireless network based on communication standard, such as WiFi (WIreless-Fidelity, nothing
Gauze network).In one exemplary embodiment, communication component 108 receives via broadcast channel and comes from external broadcasting management system
Broadcast singal or broadcast related information.
It is appreciated that structure shown in Fig. 2 is only to illustrate, computer equipment 100 this may include more than shown in Fig. 2
Or less component, or with the component different from shown in Fig. 2.Each component shown in Fig. 2 can use hardware, software
Or a combination thereof realize.
Referring to Fig. 3, in an exemplary embodiment, a method of it realizes in monitoring scene and leaves Articles detecting
The structure of computer equipment suitable for implementation environment shown in Fig. 1, the computer equipment can be as shown in Figure 2.
This kind realizes that the method that Articles detecting is left in monitoring scene can be executed by computer equipment, may include following
Step:
Step 210, it is put pixel-by-pixel for the monitoring single frames picture in monitor video and obtains co-located pixels in previous frame picture
The background model of point.
As previously mentioned, for passageway for fire apparatus or some other do not allow to stack and leave the place of article, it usually needs into
Row real time monitoring leads to occur endangering the serious problems such as human security to avoid due to leaving article violation stacking.
In existing realization, trained artificial nerve network model is generallyd use to the residue occurred in monitor video
Product are detected automatically, to save human cost, but due to actual monitored scene is sufficiently complex, it is various to leave type of goods, with
And ambient lighting changes the factors such as frequent, it is difficult to training obtains the artificial nerve network model for being adapted to different monitoring scene, from
And lead to not accurately be detected to leaving article.
The present embodiment is based on leaving article being to bring monitoring this property of place by the mankind or other organisms, can obtain
Article is left out and shows as being the image-region moved in monitor video, therefore, by carrying out sport foreground to monitor video
Detection, what the sport foreground that can be will test was retrieved as occurring in monitor video leaves article, without training of human
Artificial neural networks model, so that more convenient to the detection for leaving article.
In the present embodiment, article (i.e. sport foreground) detection is left to what monitor video carried out, is for monitor video
In monitoring single frames picture carry out, if detect in monitoring single frames picture and leave the corresponding picture region of article, then it represents that
Occur leaving article in monitor video.
The Articles detecting of leaving executed to monitoring single frames picture is that point carries out pixel-by-pixel, and depends in previous frame picture
Background model of the co-located pixels point on airspace.It should be appreciated that the airspace refers to the corresponding spatial domain of pixel namely pixel
Domain can execute the processing such as image superposition of Pixel-level on airspace.The co-located pixels point refers in different monitoring single frames picture
The identical pixel in position, for example, it is located at the pixel co-located pixels point each other of line n m column in different monitoring single frames picture,
And co-located pixels point corresponds to same airspace.
It is then by the first frame to monitor video for background model of the co-located pixels point on airspace in previous frame picture
Co-located pixels neighborhood of a point pixel is sampled in picture, obtains background model of the co-located pixels point on airspace in first frame picture
Afterwards, it is left in Articles detecting what is executed to subsequent frame picture, if detecting, co-located pixels point is the corresponding pixel of static background
Point, then according to the co-located pixels point to the background mould of co-located pixels point and neighbor pixel on airspace in previous frame picture
Type is updated obtained.
In other words, in addition to background model of the co-located pixels point on airspace in first frame picture, co-located pixels point is corresponding
It in the background model of other frame pictures is updated based on the corresponding background model of previous frame picture, because
This should be containing in historical frames picture for the background model of co-located pixels point in previous frame picture acquired in the present embodiment
The neighbor pixel of some co-located pixels points and co-located pixels point, and be made of in previous frame picture these pixels with position picture
The background pixel point of vegetarian refreshments.
As a result, relative to background model of the co-located pixels point on airspace in first frame picture, upper one acquired in the present embodiment
The background model of co-located pixels point in frame picture can simulate the background pixel of co-located pixels point in previous frame picture on airspace
Variation.
Step 230, according to the background model of acquisition, the sport foreground pixel in monitoring single frames picture is calculated.
Wherein, the sport foreground pixel monitored in single frames picture refers to, monitors and shows as being sport foreground in single frames picture
Pixel, that is, show as being the pixel for leaving article.
Monitoring single frames picture also accordingly includes static background pixel, which refers in monitoring single frames picture
Show as be pixel, such as metope, the floor of passageway for fire apparatus of static background etc. shown as in monitor video it is static or
Mobile slow image-region.
As previously mentioned, history should be contained to the background model of co-located pixels point in previous frame picture acquired in step 210
The pixel value of some co-located pixels points in frame picture and the pixel value of neighbor pixel are therefore, right as background pixel point
Pixel in the monitoring single frames picture for currently carrying out leaving Articles detecting, if itself and co-located pixels point in previous frame picture
The included background pixel point of background model distribution character it is identical, indicate co-located pixels point in the pixel and historical frames picture
Represented picture material is same or similar, therefore, it is determined that the pixel is static background pixel.
If monitoring pixel and the distribution character of the included background pixel point of corresponding background model in single frames picture not
Together, then it represents that the pixel changes with picture material represented by co-located pixels point in historical frames picture, which is
It shows as being to leave article.
As a result, with respect to background model of the co-located pixels point on airspace in previous frame image, prison is calculated by putting pixel-by-pixel
The similarity in single frames picture between pixel and the distribution character of the included background pixel point of corresponding background model is controlled, it will be similar
The pixel that degree reaches certain threshold value is judged to monitoring the sport foreground pixel in single frames picture, and similarity degree is less than certain
The pixel of threshold value is judged to monitoring the static background pixel in single frames picture.
Step 250, by sport foreground pixel, article candidate region is left in positioning in monitoring single frames picture.
As previously mentioned, sport foreground pixel is to show as being the pixel for leaving article in monitoring single frames picture, monitoring is single
It then accordingly shows as being to leave the corresponding picture region of article by the pixel region of several sport foreground pixel connection in frame picture.
As a result, according to the link relation between sport foreground pixel, can position to obtain leaving in monitoring single frames picture
Article candidate region.
In one exemplary embodiment, it is contemplated that influence leaving article and should having centainly for the smoothness of passageway for fire apparatus
Size will by judging whether the area for monitoring the sport foreground pixel being connected in single frames picture reaches setting area threshold
The connection zone location for reaching area threshold is to leave article candidate region, thus promotes the positioning standard for leaving article candidate region
True property.
It, can also be by judgement monitoring single frames picture since the size of each pixel in monitoring single frames picture is all the same
Whether the sport foreground pixel quantity being connected is greater than given threshold, and positioning sport foreground pixel quantity is greater than the connection of given threshold
Logical region is to leave article candidate region.
It can be concluded that compared with prior art, before the present embodiment by putting calculating movement to monitoring single frames picture pixel-by-pixel
Scene element, can leave article candidate region according in gained sport foreground pixel locating and monitoring single frames picture, not need
Complicated artificial nerve network model training is carried out, the numerous influences of type of goods will not be left.
Step 270, article candidate region is left according to the organism historical track filtering in monitor video, obtains residue
The region that product are stopped.
Wherein, the organism historical track in monitor video refers to, has organism to occur in monitor video and is carried out continuously
Mobile motion profile.
Leaving the sport foreground pixel of article candidate region in view of composition is obtained via aforementioned movement foreground detection
, these sport foreground pixels may also show as be the organism moved in monitor video or other movements object, also
May be since static background pixel is mistaken for being sport foreground pixel by the influence of ambient lighting, according to step 250 institute
Position obtain a possibility that leaving article candidate region there are still erroneous detections.
Being still based on and leaving article is that the region week that article appearance is left to this property of monitoring scene is carried via organism
Side should have organism historical track, thus establish organism and leave the strong association of article, according to the life in monitor video
The article candidate region of leaving that object historical track positions step 250 is filtered, can largely will be above-mentioned
Erroneous detection leaves the removal of article candidate region, and article stacking area is as accurately left in finally obtained article candidate region of leaving
Domain.
As a result, compared with prior art, the present embodiment institute's providing method leaves Articles detecting and something lost in monitor video
It stays the type of article unrelated, and can also be filtered out according to the organism historical track in monitor video since ambient lighting becomes
Change and article candidate region is left by erroneous detection, can eliminate and leave type of goods and ambient lighting variation to leaving Articles detecting
The accurate detection to article is left is realized in the influence of accuracy.
Referring to Fig. 4, in an exemplary embodiment, before step 210, realizing in monitoring scene and leaving article
The method of detection is further comprising the steps of:
Step 310, ambient lighting influence is eliminated by putting pixel-by-pixel on monitoring single frames picture, rgb color is empty where obtaining
Between on pixel value update.
Wherein, since different ambient lightings will lead to monitoring camera acquired image color and true colors exist
Certain deviation is easy to cause to the erroneous detection of sport foreground pixel and static background pixel in monitoring single frames picture, in certain journey
Also it will affect the accuracy for leaving Articles detecting on degree.
In order to further eliminate influence of the ambient lighting to Articles detecting is left in monitoring single frames picture, residue is being carried out
Before product examine is surveyed, each monitoring single frames picture in monitor video can be put in advance pixel-by-pixel and eliminate ambient lighting influence, with
The original scene of reduction monitoring single frames picture.Based on the original scene of monitoring single frames picture, it is able to ascend to monitoring single frames picture
In each pixel identification accuracy.
The process eliminating ambient lighting and influencing is put pixel-by-pixel on monitoring single frames picture, substantially in monitoring single frames picture
The process that pixel value of each pixel on rgb color space is updated, the pixel value updated are that corresponding pixel points are gone
Except the original pixel value of environment illumination effect.It should be appreciated that the Image Acquisition that monitoring camera is carried out is all in existing realization
It is to be realized based on rgb color space, therefore the monitoring single frames picture in monitor video should also be as corresponding to rgb color sky
Between.
In one exemplary embodiment, it can realize to put monitoring single frames picture pixel-by-pixel using gray world algorithm and disappear
Except environment illumination effect.Gray world algorithm assume nature scenery for the average reflection of ambient lighting mean value on the whole
It is definite value, which is approximately " grey ", then can be from monitoring single frames figure by the hypothesis mandatory use in monitoring single frames picture
The influence of environment light, the original scene of reduction monitoring single frames picture are eliminated in piece.
It is as follows that the process that elimination ambient lighting influences is put to monitoring single frames picture using gray world algorithm pixel-by-pixel:
The average reflection mean value of ambient lighting is determined firstIllustratively, can calculate monitoring single frames picture R,
G, the average channel value in tri- channels B, and take the average value of these three average channel values for the average reflection mean value of ambient lighting,
HaveThe half of maximum gradation value that monitoring single frames picture, which can also be directly acquired, to be shown is
The average reflection mean value of ambient lighting, this place is not limited.
Then calculating each pixel in monitoring single frames picture, respectively in the gain coefficient in tri- channels R, G, B, these increase
Beneficial coefficient is the ratio between the average reflection mean value of ambient lighting and each channel value of current pixel point.Wherein, gain coefficient
Correspondence is expressed as
The product of the gain coefficient of each channel value and corresponding channel of each pixel in monitoring single frames picture is finally calculated,
To obtain removing each Src Chan value that ambient lighting influences to each pixel.To each picture in monitoring single frames picture
The original of removal ambient lighting influence can be obtained by the way that each Src Chan value to be updated current channel value in vegetarian refreshments
Pixel value.
In a further exemplary embodiment, the speed that ambient lighting influences is eliminated on monitoring single frames picture for quickening, and
Computing resource is saved, before putting elimination ambient lighting pixel-by-pixel on monitoring single frames picture and influencing, also monitoring single frames picture is held
The down-sampled processing of row, reduces the resolution ratio of monitoring single frames picture.Illustratively, the resolution ratio for monitoring single frames picture can be reduced
To 400 × 320.
Step 330, according to gained pixel value is updated, monitoring single frames picture is converted by rgb color space to HSV color
Space.
Wherein, HSV color space is by representation method of the point in inverted cone in rgb color space, includes form and aspect
(Hue), three channels of saturation degree (Saturation) and lightness (Value) are more closely similar to the mankind's for the expression of color
Color perception, compared to rgb color space, HSV color space is more intuitive.
As a result, after eliminating ambient lighting on monitoring single frames picture and influencing, single frames picture will be also monitored by rgb color space
It converts to HSV color space, and the monitoring single frames picture being converted to is used to execute the detection for leaving article, to simulate the mankind
Actual visual detected to article is left, further improve to monitoring single frames picture in leave the accurate of Articles detecting
Degree.
In one exemplary embodiment, monitoring single frames picture is converted by rgb color space to the mistake of HSV color space
Journey is as follows:
To each pixel in monitoring single frames picture, the color component and monitoring single frames figure of rgb color space are first calculated
The ratio between maximum gradation value that piece can be shown, that is, calculate separately R'=R/255, G'=G/255, B'=B/255, and
Obtain maximum ratio Cmax=max (R', G', B'), minimum ratio Cmin=min (R', G', B') and maximum ratio and minimum
Difference DELTA=C between ratiomax-Cmin。
When difference DELTA is zero, form and aspect component of the pixel in HSV color space is 0 °;Work as CmaxWhen=R', pixel
Form and aspect component beWork as CmaxWhen=G', the form and aspect component of pixel isWhen
CmaxWhen=B', the form and aspect component of pixel is
Work as CmaxWhen=0, saturation degree component of the pixel in HSV color space is zero;Work as CmaxWhen ≠ 0, pixel
Saturation degree component is
Lightness component of the pixel in HSV color space is Cmax。
As a result, according to obtained pixel in HSV color space form and aspect component, saturation degree component and lightness component, it is right
R, G of rgb color space, channel B value are replaced where pixel, and the pixel can be realized by rgb color space to HSV
The conversion of color space.
Referring to Fig. 5, in an exemplary embodiment, leaving the monitoring single frames picture of Articles detecting in current execution
When for first frame picture in monitor video, before step 210, the above method is further comprising the steps of:
Step 211, to each pixel in monitoring single frames picture, stochastical sampling is carried out by the neighborhood territory pixel to pixel,
Obtain background pixel point set of the pixel on airspace;
Step 213, it obtains background pixel point set of the pixel on airspace and forms background model.
Wherein, it monitors pixel neighborhood of a point in single frames picture to refer to, the certain area around current pixel point, example
Such as, to a certain pixel in monitoring single frames picture, neighborhood can be the surrounding of the pixel, be also possible to the pixel
Upper area, this place are not limited.
The neighborhood territory pixel of pixel then refers in monitoring single frames picture, the picture in contiguous range where current pixel point
Vegetarian refreshments.
Assuming that the pixel value of each pixel and its neighborhood territory pixel has similar distribution on airspace in monitoring single frames picture,
On the basis of this assumption, each pixel can be expressed as corresponding background model according to the pixel in its neighborhood.But it needs
It is noted that in order to guarantee that background model coincidence statistics rule, the range for monitoring pixel neighborhood of a point in single frames picture are answered
When sufficiently large, the neighborhood territory pixel of certain scale is obtained to guarantee sampling, namely sampling obtains the background pixel of certain scale
Point.
Stochastical sampling several times is carried out by the neighborhood territory pixel to each pixel in first frame picture, accordingly to obtain each pixel
Background pixel point set of the point on airspace, and back of each pixel on airspace is formed by sampling gained background pixel point set
Scape model.
In an exemplary embodiment, pass through the neighborhood territory pixel stochastical sampling 8 to the pixel P (x) in first frame picture
It is secondary, background pixel point P (1) to P (8) successively is obtained, then pixel P (x) is constituted in airspace by background pixel point P (1) to P (8)
On background model.
The second frame picture can be carried out after background model of each pixel on airspace in obtaining first frame picture as a result,
Point executes the identification for leaving article pixel-by-pixel.Illustratively, to a certain pixel in the second frame picture, relatively in head frame picture
Background model on co-located pixels idea airspace executes the calculating of sport foreground pixel, if obtaining the pixel is static background
Pixel is then carried out the static background pixel as background model of the new background pixel point to co-located pixels point in first frame picture
It updates, background model corresponding to the second frame picture can be obtained.It so repeats, the monitoring single frames of Articles detecting is left to execution
Picture all leaves Articles detecting according to the corresponding background model execution of co-located pixels point in previous frame picture.
It therefore deduces that, the present embodiment need to only execute the first frame picture in monitor video at the initialization of background model
Reason executes the random sampling procedure of neighborhood territory pixel, without taking a significant amount of time as each pixel in each monitoring single frames picture
Point generates corresponding background model, can be realized the real-time for leave to monitor video Articles detecting.
Referring to Fig. 6, in an exemplary embodiment, step 230 may comprise steps of:
Step 231, correspond to monitoring single frames picture where color space, by calculate monitoring single frames picture in pixel with
The distance of each background pixel point in background model obtains the background pixel that pixel value is similar with the pixel in background model
Point.
Wherein, for the background pixel point in background model, pixel value is close with the pixel in monitoring single frames picture
Seemingly refer to, background pixel point is similar with image content represented by the pixel in monitoring single frames picture, and the two pixel value is also answered
When similar, distance of the two on color space should also be as being less than certain threshold value.
Therefore, in background model is calculated each background pixel point respectively with monitoring single frames picture in pixel it
Between distance after, obtaining distance to be less than the background pixel point of set distance threshold value is similar background pixel in background model
Point.
In one exemplary embodiment, in HSV color space, since form and aspect channel is insensitive to ambient lighting, lead to
Cross calculate background pixel point with monitor in single frames picture pixel on form and aspect channel at a distance from obtain similitude, can into one
Step eliminates the influence of ambient lighting, monitors the accuracy that Articles detecting is left in single frames picture with further promoted.
And in a further embodiment, the pixel in background pixel point and monitoring single frames picture can also be calculated in HSV
Space length in color space, this place are limited not to this.
Step 233, when similar background pixel point quantity is less than amount threshold, the picture in monitoring single frames picture is obtained
Vegetarian refreshments is sport foreground pixel.
As previously mentioned, if current pixel point and the included background pixel point of corresponding background model in monitoring single frames picture
Distribution character it is different, then it represents that current pixel point becomes with picture material represented by co-located pixels point in historical frames picture
Change, which shows as being to leave article.
When background pixel point quantity similar with corresponding pixel points in monitoring single frames picture in background model is less than quantity
When threshold value, i.e., the corresponding pixel points in the distribution character of expression the included background pixel point of background model and monitoring single frames picture are not
Identical, thus obtaining the pixel in monitoring single frames picture is sport foreground pixel.
The present embodiment by calculate monitoring single frames picture in pixel in corresponding background model each background pixel point away from
From accurately to obtain the pass in monitoring single frames picture in pixel and background model between the distribution character of each background pixel point
System, thus, it is possible to obtain the sport foreground pixel in monitoring single frames picture to accurate detection.
Referring to Fig. 7, in an exemplary embodiment, step 230 is further comprising the steps of:
Step 232, reach amount threshold in similar background pixel point quantity, or detect in monitoring single frames picture
Pixel when belonging to special area, obtaining the pixel is static background pixel.
Wherein, when background pixel point quantity similar in background model reaches amount threshold, background mould is indicated
The distribution character of the included background pixel point of type is identical as the corresponding pixel points in monitoring single frames picture, thus obtains monitoring single frames
Pixel in picture is static background pixel.
Also, when the pixel in monitoring single frames picture is via being detected as belonging to highlight area, shadow region etc. due to ring
When border illumination effect leads to the special area occurred, also obtaining these pixels is static background pixel.It should be noted that right
The detection that the pixel in single frames picture carries out bloom or shadow region is monitored, is differentiated according to bloom distinguished number, shade
What the respective algorithms such as algorithm were realized, herein without detailed description.
Step 234, random to the background pixel point in background model using static background pixel as new background pixel point
It updates, obtains the background model of update.
Wherein, detection obtain monitoring single frames picture in static background pixel after, need using the static background pixel as
New background pixel point is updated in background model.
Randomly updating for background pixel point is referred to, is in background model by the pixel value random replacement of static background pixel
A background pixel point.
By being updated to background model corresponding to static background pixel, background model is enabled to contain historical frames figure
Co-located pixels point in piece so that background model can accurate simulation background pixel variation, thus be conducive to carry out subsequent frame picture
In leave the identification of article.
Step 236, according to setting probability, background model corresponding to the neighborhood territory pixel to static background pixel is carried out at random more
Newly.
Wherein, setting probability refers to that it has the probability of 1/rate more when a pixel is detected as static background
Background model corresponding to new neighborhood territory pixel, wherein rate indicates that the time sampling factor, general value are 16.
Background model corresponding to neighborhood territory pixel to static background pixel randomly update referring to, according to setting probability,
First an adjacent pixel is randomly selected from the neighborhood of static background pixel, then will the static background pixel pixel
Value randomly selects in background model corresponding to neighbor pixel background pixel point and is replaced as new background pixel point
It changes.
In the present embodiment, not only the corresponding background model of static background pixel is randomly updated, also to its neighborhood
The corresponding background model of pixel randomly updates, and takes full advantage of the spatial characteristic of pixel value so that background model gradually to
Neighborhood territory pixel diffusion is conducive to the identification for carrying out ghost region.
Referring to Fig. 8, in an exemplary embodiment, after step 233 can with the following steps are included:
Step 410, whether ghost region is belonged to according to sport foreground pixel, identifies the erroneous judgement of sport foreground pixel, and
The sport foreground pixel of amendment erroneous judgement is static background pixel.
Wherein, in the sport foreground pixel that step 233 detection obtains, it is understood that there may be belong to ghost caused by ambient lighting influence
Shadow zone domain, and being by erroneous detection is sport foreground pixel, this also will affect the accuracy for leaving Articles detecting.Thus, it is desirable into one
Step eliminates the ghost region in monitoring single frames picture.
It is considered as the sport foreground pixel of sport foreground pixel for co-located pixels point continuous multiple frames, according to place HSV
The information and setting channel threshold value of color space, can be made whether to belong to the differentiation in ghost region.
As previously described, because form and aspect channel is insensitive to ambient lighting, ghost can be preset on form and aspect channel
Channel value distributed area, if the form and aspect channel value of sport foreground pixel is located in the form and aspect channel value distributed area of setting,
It indicates that the sport foreground pixel belongs to ghost region, is then modified to static background pixel.
Step 430, the update to correct the corresponding background model of obtained static background pixel execution.
It is randomly updated as previously mentioned, corresponding background model should be executed to each static background pixel, specifically
This place of renewal process repeats no more.
The present embodiment can further eliminate ghost region caused by ambient lighting as a result, eliminate environment most possibly
Influence of the illumination to Articles detecting is left further improves pair so that the present embodiment has robustness to ambient lighting variation
Leave the accuracy of Articles detecting.
Referring to Fig. 9, in an exemplary embodiment, realizing that the method for leaving Articles detecting in monitoring scene is also wrapped
Include following steps:
Step 510, by carrying out living body detection to the monitoring single frames picture in monitor video, monitoring single frames picture is obtained
In correspond to organism picture region.
As previously described, because leaving article must carry via organism to monitoring scene, article appearance is left
Area peripheral edge certainly exists organism historical track, is positioned according to the organism historical track in monitor video to step 250
Article candidate region of leaving be filtered, can accurately left article dwell regions.
As a result, when starting to leave the detection of article to the monitoring single frames picture execution in monitor video, need to monitoring
Single frames picture is synchronous to carry out living body detection and maintenance, with during leaving Articles detecting to monitoring single frames picture execution,
Based on the organism historical track of institute's synchronous maintenance, the article candidate region of leaving in monitoring single frames picture is filtered.
Illustratively, living body detection is carried out to monitoring single frames picture, and by training in advance and for executing biology
What the artificial nerve network model that physical examination is surveyed was realized, such as yolov3 model, fastercnn model etc., it can accordingly obtain
Monitor the picture region for corresponding to organism in single frames picture.
Step 530, the picture region being consecutively detected is safeguarded, obtains the organism history rail in monitor video
Mark.
Wherein, the picture region being consecutively detected refers to, to biology performed by continuous several monitoring single frames pictures
During physical examination is surveyed, each frame picture detects the corresponding picture region of organism, therefore these picture regions are safeguarded, i.e.,
Organism historical track can be obtained.
In an exemplary embodiment, when detecting monitoring single frames picture for the first time, there are the corresponding picture regions of organism
When domain, start to carry out the accumulative of picture region, and the picture region by continuously detecting is accumulated by organism historical track.
When the organism being not detected in monitoring single frames picture, then start to be counted, when counting reaches certain threshold value
When, organism is not detected in expression within a certain period of time, safeguarded organism historical track is emptied at this time, to detect again
When to organism, the accumulative of picture region corresponding to organism is re-executed.
As a result, in the filtering for leaving article candidate region performed by step 270, work as to carrying out leaving Articles detecting
Previous frame picture can obtain the corresponding picture region of organism in present frame picture according to the organism historical track safeguarded,
To be included in, the picture region is interior or the leave article candidate region adjacent with the picture region is retrieved as leaving article
The region stopped can obtain and accurately leave article dwell regions.
And leave article candidate region for unfiltered in monitoring single frames picture, that is, it leaves article candidate region and is not contained in
It is in the corresponding picture region of organism and non-conterminous, then it represents that it is the corresponding figure of non-legacy article that these, which leave article candidate region,
Panel region, therefore these are left into the total movement foreground pixel that article candidate region is included and is modified to static background pixel,
And these modified static background pixels are executed with the update of corresponding background model.
Referring to Fig. 10, in an exemplary embodiment, realizing in monitoring scene and leaving the method for Articles detecting also
The following steps are included:
Step 610, it is left in Articles detecting what is executed to monitoring single frames picture, if detecting monitoring single frames figure for the first time
Containing the region that article is stopped is left in piece, start to be counted;
Step 630, alarm is executed when the numerical value of counting is more than alarm threshold value.
Wherein, it for the application scenarios left the region that article is stopped and execute automatic alarm are detected, is detected for the first time
When into monitoring single frames picture containing region that article is stopped is left, start to be counted, when the numerical value of counting is more than alarm
When threshold value, expression leaves article and has stayed for some time in monitoring scene, therefore executes automatic alarm.
Compared to the mode for being immediately performed alarm after article occurs in monitoring scene is left, the present embodiment can be avoided something lost
Stay article in monitoring scene short stay and cause unnecessary alarm.
Figure 11 is please referred to, in another exemplary embodiment, it is contemplated that it is of short duration in monitoring scene to leave article
The case where stop can cause unnecessary alarm, before step 630, this method is further comprising the steps of:
Step 710, when in detecting monitoring single frames picture without containing leaving region that article is stopped, stop counting;
Step 730, it if counting the time stopped reaching setting frame number, is reset to counting.
Wherein, detect that expression has left article without containing when leaving the region that article is stopped in monitoring single frames picture
Through removing monitoring scene, is counted by stopping, numerical value is avoided to continue growing and trigger false alarm.
When the time that counting stops reaching setting frame number, expression leaves article and determines removal monitoring scene, no longer executes
Automatic alarm, therefore current counting is reset, leave what article was stopped to detect to contain in monitoring single frames picture again
When region, start from scratch and count again, to meet actual monitoring scene.
And in a further embodiment, it is contemplated that under actual monitored scene, long-time of the organism in monitoring scene is funny
It stays and will cause false alarm, and leave situations such as article can be taken away in monitoring scene after short stay by organism, if lost
There are the corresponding picture regions of organism for the area peripheral edge for staying article to be stopped, even if the numerical value of counter is more than alarm threshold value,
Alarm will not be executed.
Figure 12 is a kind of specific implementation signal for realizing the method that Articles detecting is left in monitoring scene in an application scenarios
Figure.The application scenarios are applied particularly to leave Articles detecting in passageway for fire apparatus.
As shown in figure 12, the present frame picture that Articles detecting is left for executing in monitor video, on the one hand passes through execution
Resolution adjustment, gray world algorithm are eliminated illumination effect and are converted by rgb color space to HSV color space, to institute
Obtained present frame picture puts the judgement for executing static background pixel or sport foreground pixel pixel-by-pixel, and is more than threshold by area
The sport foreground pixel connection region of value is retrieved as leaving article candidate region.
On the other hand, it needs to carry out living body detection to present frame picture, and contains organism in detecting picture
When corresponding picture region, organism history rail is carried out based on picture region corresponding to the organism detected in history picture
The maintenance of mark;If be not detected in picture containing the corresponding picture region of organism, counting is executed by counter, is being counted
Number empties safeguarded organism historical track when being more than threshold value.
Article candidate region is left by acquiring in present frame picture in terms of two above, and the life safeguarded
After object historical track, it is filtered, is lost to article candidate region is left according to the organism historical track safeguarded
The region for staying article to stop, and counted containing the picture for leaving article dwell regions by another counter to detecting
Number executes automatic alarm when counting and being more than threshold value.
When occurring leaving article in detecting passageway for fire apparatus as a result, automatic alarm can be triggered, with notify related personnel and
When clear up to leaving article, so that the unimpeded of passageway for fire apparatus be effectively ensured, due to fire-fighting when avoiding occurring various dangerous situations
Channel blockage causes major accident.
Following is apparatus of the present invention embodiment, can be used for executing residue in realization monitoring scene according to the present invention
The method that product examine is surveyed.For undisclosed details in apparatus of the present invention embodiment, realization monitoring according to the present invention is please referred to
The embodiment of the method for Articles detecting is left in scene.
Figure 13 is please referred to, it is in an exemplary embodiment, a kind of to realize the dress that Articles detecting is left in monitoring scene
It sets and obtains module 810, sport foreground computing module 830, candidate region locating module 850 and candidate region mistake including background model
Filter module 870.
Background model obtains module 810 and is used to put acquisition previous frame pixel-by-pixel for the monitoring single frames picture in monitor video
The background model of co-located pixels point in picture, the background model on airspace for simulating co-located pixels point in previous frame picture
Background pixel variation.
Sport foreground computing module 830 is used for opposite background model and calculates the sport foreground pixel monitored in single frames picture,
The monitoring single frames picture includes sport foreground pixel and static background pixel.
Candidate region locating module 850, which is used to be positioned in monitoring single frames picture by sport foreground pixel, leaves article candidate
Region.
Candidate region filtering module 870 is used to leave article candidate according to the organism historical track filtering in monitor video
The region that article is stopped is left in region, acquisition.
In an exemplary embodiment, the device of Articles detecting is left in above-mentioned realization monitoring scene further includes environment
Illumination cancellation module and color space conversion module.
Ambient lighting cancellation module is used to eliminate ambient lighting influence by putting monitoring single frames picture pixel-by-pixel, obtains institute
Pixel value on rgb color space updates.
Color space conversion module is used to turn monitoring single frames picture by rgb color space according to gained pixel value is updated
Shift to HSV color space.
In an exemplary embodiment, in the case where monitoring single frames picture is the first frame picture in monitor video,
The device that Articles detecting is left in above-mentioned realization monitoring scene further includes that neighborhood territory pixel sampling module and background pixel obtain module.
Neighborhood territory pixel sampling module be used for monitoring single frames picture in each pixel, by the neighborhood territory pixel to pixel into
Capable stochastical sampling several times obtains background pixel point set of the pixel on airspace.
Background pixel obtains module and is used to obtain background pixel point set formation background model of the pixel on airspace.
In an exemplary embodiment, sport foreground computing module 830 includes space length computing unit and movement
Foreground pixel acquiring unit.
Space length computing unit is used to correspond to color space where monitoring single frames picture, monitors single frames figure by calculating
Pixel obtains in background model in pixel value and monitoring single frames picture at a distance from background pixel point each in background model in piece
The similar background pixel point of pixel.
Sport foreground pixel acquisition unit is used for when similar background pixel point quantity is less than amount threshold, and obtaining should
Pixel is sport foreground pixel.
In an exemplary embodiment, sport foreground computing module 830 further include static background pixel acquisition unit,
Background model updating unit and neighborhood territory pixel updating unit.
Static background pixel acquisition unit is used to reach amount threshold or detection in similar background pixel point quantity
When belonging to special area to pixel, obtaining the pixel is static background pixel, which includes highlight area and yin
Shadow zone domain.
Background model updating unit is used for using static background pixel as new background pixel point to the back in background model
Scene vegetarian refreshments randomly updates, and obtains the background model of update.
Neighborhood territory pixel updating unit is used for according to setting probability, background mould corresponding to the neighborhood territory pixel to static background pixel
Type is randomly updated.
In an exemplary embodiment, sport foreground computing module 830 further includes ghost erroneous judgement recognition unit and repairs
Positive updating unit.
Ghost erroneous judgement recognition unit is used to whether belong to ghost region recognition sport foreground picture according to sport foreground pixel
The erroneous judgement of element, and the sport foreground pixel for correcting erroneous judgement is static background pixel.
The static background pixel that amendment updating unit is used to obtain for amendment executes the update of corresponding background model.
In an exemplary embodiment, the device of Articles detecting is left in above-mentioned realization monitoring scene further includes biology
Body detection module and historical track maintenance module.
Living body detection module is used to be supervised by carrying out living body detection to the monitoring single frames picture in monitor video
Control the picture region for corresponding to organism in single frames picture.
Historical track maintenance module obtains the life in monitor video for safeguarding to the picture region being consecutively detected
Object historical track.
In an exemplary embodiment, candidate region filtering module 870 includes picture region acquiring unit and target
Acquiring unit.
Picture region acquiring unit is used for the monitoring single frames picture for leave Articles detecting, according to the biology safeguarded
Body historical track obtains the picture region for corresponding to organism.
Target Acquisition unit is used to be included in picture region or adjacent with picture region leaves article candidate regions
Domain is retrieved as leaving the region that article is stopped.
In an exemplary embodiment, the device that Articles detecting is left in above-mentioned realization monitoring scene further includes counting
Module and alarm module.
Counting module is used to leave in Articles detecting execute monitoring single frames picture, if detecting that monitoring is single for the first time
Containing the region that article is stopped is left in frame picture, start to be counted.
Alarm module is used to execute alarm when the numerical value of counting is more than alarm threshold value.
In an exemplary embodiment, which further includes counting to stop control module and counting dump block.
It counts and stops control module for not containing the region leaving article and being stopped in detecting monitoring single frames picture
When, control stops counting.
Dump block is counted to be used to reset when the time for counting stopping reaching setting frame number to counting.
It should be noted that method provided by device provided by above-described embodiment and above-described embodiment belongs to same structure
Think, the concrete mode that wherein modules and unit execute operation is described in detail in embodiment of the method, herein
It repeats no more.
The present invention also provides a kind of electronic equipment, including processor and memory, wherein calculating is stored on memory
Machine readable instruction is realized when the computer-readable instruction is executed by processor and realizes residue product examine in monitoring scene as previously described
The method of survey.
The present invention also provides a kind of computer readable storage mediums, are stored thereon with computer program, the computer journey
The method realized leave Articles detecting in monitoring scene as previously described is realized when sequence is executed by processor.
Above content, only the preferable examples embodiment of the application, the embodiment for being not intended to limit the application, this
Field those of ordinary skill can very easily carry out corresponding flexible or repair according to the central scope and spirit of the application
Change, therefore the protection scope of the application should be subject to protection scope required by claims.
Claims (14)
1. a kind of realize the method for leaving Articles detecting in monitoring scene, which is characterized in that the described method includes:
Put the background model for obtaining co-located pixels point in previous frame picture pixel-by-pixel for the monitoring single frames picture in monitor video,
The background model is used to simulate the background pixel variation of co-located pixels point in the previous frame picture on airspace;
According to the background model of acquisition, the sport foreground pixel in the monitoring single frames picture, the monitoring single frames are calculated
Picture includes the sport foreground pixel and static background pixel;
It is positioned in the monitoring single frames picture by the sport foreground pixel and leaves article candidate region;
Article candidate region is left according to the organism historical track filtering in the monitor video, article institute is left in acquisition
The region of stop.
2. the method according to claim 1, wherein the monitoring single frames picture in monitor video by
Pixel obtains in previous frame picture before the background model of co-located pixels point, the method also includes:
Ambient lighting influence is eliminated by putting pixel-by-pixel on the monitoring single frames picture, the picture where obtaining on rgb color space
Element value updates;
According to pixel value obtained by the update, the monitoring single frames picture is converted by the rgb color space to HSV color sky
Between.
3. the method according to claim 1, wherein the monitoring single frames picture is the head in the monitor video
Frame picture is put pixel-by-pixel in the monitoring single frames picture in monitor video and obtains co-located pixels point in previous frame picture
Before background model, the method also includes:
To each pixel in the monitoring single frames picture, stochastical sampling is carried out by the neighborhood territory pixel to the pixel, is obtained
Background pixel point set of the pixel on airspace;
It obtains background pixel point set of the pixel on airspace and forms the background model.
4. the method according to claim 1, wherein the relatively described background model, it is single to calculate the monitoring
Sport foreground pixel in frame picture, comprising:
Corresponding to color space where the monitoring single frames picture, respectively carried on the back by calculating in the pixel and the background model
The distance of scene vegetarian refreshments obtains the background pixel point that pixel value is similar with the pixel in the background model;
When the similar background pixel point quantity is less than amount threshold, obtaining the pixel is sport foreground pixel.
5. according to the method described in claim 4, it is characterized in that, the method also includes:
Reach the amount threshold in the similar background pixel point quantity or to detect that the pixel belongs to special
When region, obtaining the pixel is static background pixel, and the special area includes highlight area and shadow region;
The background pixel point in the background model is randomly updated using the static background pixel as new background pixel point,
Obtain the background model updated;
According to setting probability, randomly updated described in background model progress corresponding to the neighborhood territory pixel to the static background pixel.
6. according to the method described in claim 5, it is characterized in that, calculating the monitoring in the relatively described background model
After sport foreground pixel in single frames picture, the method also includes:
Ghost region whether is belonged to according to the sport foreground pixel, identifies the erroneous judgement of the sport foreground pixel, and correct
The sport foreground pixel of erroneous judgement is static background pixel;
To correct the update that the obtained static background pixel executes corresponding background model.
7. the method according to claim 1, wherein the method also includes:
By carrying out living body detection to the monitoring single frames picture in the monitor video, it is right in the monitoring single frames picture to obtain
The picture region of organism described in Ying Yu;
The picture region being consecutively detected is safeguarded, the organism historical track in the monitor video is obtained.
8. according to the method described in claim 7, described according to the organism historical track filtering in the monitor video
Article candidate region is left, the region that article is stopped is left in acquisition, comprising:
The monitoring single frames picture that Articles detecting is left described in progress corresponds to according to the organism historical track acquisition safeguarded
The picture region of the organism;
It is included in the picture region or the leave article candidate region adjacent with the picture region is retrieved as leaving
The region that article is stopped.
9. according to the method described in claim 8, it is characterized in that, the method also includes:
Article candidate region is left to unfiltered in the monitoring single frames picture, the total movement foreground pixel for being included is repaired
Just it is static background pixel, and executes the update of corresponding background model to the static background pixel.
10. the method according to claim 1, wherein the method also includes:
It is left in Articles detecting what is executed to the monitoring single frames picture, if detected in the monitoring single frames picture for the first time
Containing the region leaving article and being stopped, start to be counted;
Alarm is executed when the numerical value of the counting is more than alarm threshold value.
11. according to method described in right 10, which is characterized in that at described when the numerical value of the counting is more than alarm threshold value
Before executing alarm, the method also includes:
When leaving the region that article is stopped without containing described in detecting the monitoring single frames picture, stop the counting;
If the time for counting stopping reaches setting frame number, the counting is reset.
12. a kind of realize the device for leaving Articles detecting in monitoring scene, which is characterized in that described device includes:
Background model obtains module, obtains in previous frame picture for putting pixel-by-pixel for the monitoring single frames picture in monitor video
The background model of co-located pixels point, the background model on airspace for simulating co-located pixels point in the previous frame picture
Background pixel variation;
Sport foreground computing module, before the movement in the monitoring single frames picture is calculated for the background model according to acquisition
Scene element, the monitoring single frames picture includes the sport foreground pixel and static background pixel;
Candidate region locating module leaves article time for being positioned in the monitoring single frames picture by the sport foreground pixel
Favored area;
Candidate region filtering module is waited for leaving article according to the organism historical track filtering in the monitor video
The region that article is stopped is left in favored area, acquisition.
13. a kind of electronic equipment characterized by comprising
Memory is stored with computer-readable instruction;
Processor reads the computer-readable instruction of memory storage, is required described in any one of 1-11 with perform claim
Method.
14. a kind of computer readable storage medium, which is characterized in that computer-readable instruction is stored thereon with, when the calculating
When machine readable instruction is executed by the processor of computer, computer perform claim is made to require side described in any one of 1-11
Method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910547197.3A CN110348327A (en) | 2019-06-24 | 2019-06-24 | Realize the method and device that Articles detecting is left in monitoring scene |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910547197.3A CN110348327A (en) | 2019-06-24 | 2019-06-24 | Realize the method and device that Articles detecting is left in monitoring scene |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110348327A true CN110348327A (en) | 2019-10-18 |
Family
ID=68182812
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910547197.3A Pending CN110348327A (en) | 2019-06-24 | 2019-06-24 | Realize the method and device that Articles detecting is left in monitoring scene |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110348327A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111914670A (en) * | 2020-07-08 | 2020-11-10 | 浙江大华技术股份有限公司 | Method, device and system for detecting left-over article and storage medium |
CN112417978A (en) * | 2020-10-26 | 2021-02-26 | 刘英杰 | Vision-based method and device for detecting foreign matters between train and shield door |
CN113378602A (en) * | 2020-03-10 | 2021-09-10 | 顺丰科技有限公司 | Method and device for detecting illegal stacking of articles, server and storage medium |
CN113470013A (en) * | 2021-07-28 | 2021-10-01 | 浙江大华技术股份有限公司 | Method and device for detecting moved article |
-
2019
- 2019-06-24 CN CN201910547197.3A patent/CN110348327A/en active Pending
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113378602A (en) * | 2020-03-10 | 2021-09-10 | 顺丰科技有限公司 | Method and device for detecting illegal stacking of articles, server and storage medium |
CN111914670A (en) * | 2020-07-08 | 2020-11-10 | 浙江大华技术股份有限公司 | Method, device and system for detecting left-over article and storage medium |
CN112417978A (en) * | 2020-10-26 | 2021-02-26 | 刘英杰 | Vision-based method and device for detecting foreign matters between train and shield door |
CN113470013A (en) * | 2021-07-28 | 2021-10-01 | 浙江大华技术股份有限公司 | Method and device for detecting moved article |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110348327A (en) | Realize the method and device that Articles detecting is left in monitoring scene | |
JP6504364B2 (en) | Monitoring device, monitoring system and monitoring method | |
CN110044486B (en) | Method, device and equipment for avoiding repeated alarm of human body inspection and quarantine system | |
US9536154B2 (en) | Monitoring method and camera | |
CN105160297B (en) | Masked man's event automatic detection method based on features of skin colors | |
CN109409238B (en) | Obstacle detection method and device and terminal equipment | |
KR20080054368A (en) | Flame detecting method and device | |
TW201633215A (en) | System and method for protecting eyes | |
JP2007286995A (en) | Attention level measurement device and attention level measurement system | |
CN107948465A (en) | A kind of method and apparatus for detecting camera and being disturbed | |
KR100822476B1 (en) | Remote emergency monitoring system and method | |
CA2845950A1 (en) | System and method of audio information display on video playback timeline | |
WO2023005403A1 (en) | Respiratory rate detection method and apparatus, and storage medium and electronic device | |
CN113591701A (en) | Respiration detection area determination method and device, storage medium and electronic equipment | |
CN108288361A (en) | A kind of passageway for fire apparatus door state detection method | |
CN111611630A (en) | Mobile terminal peeping prevention method and mobile terminal | |
CN104506850A (en) | Method and device for judging video occlusion | |
CN107317992A (en) | video communication control device and control method | |
CN105761275A (en) | Fire-fighting early warning aircraft with binocular visual structure | |
CN107169885A (en) | A kind of customer's article leaves detection method and system | |
CN108460357A (en) | A kind of windowing alarm detection system and method based on image recognition | |
JP2015169618A (en) | visible smoke determination device and visible smoke determination method | |
CN102567951A (en) | Binary image display technology for protecting user privacy in video monitoring | |
CN115761638A (en) | Online real-time intelligent analysis method based on image data and terminal equipment | |
JP5586383B2 (en) | Image monitoring device |
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 |