CN106934819A - A kind of method of moving object segmentation precision in raising image - Google Patents
A kind of method of moving object segmentation precision in raising image Download PDFInfo
- Publication number
- CN106934819A CN106934819A CN201710140027.4A CN201710140027A CN106934819A CN 106934819 A CN106934819 A CN 106934819A CN 201710140027 A CN201710140027 A CN 201710140027A CN 106934819 A CN106934819 A CN 106934819A
- Authority
- CN
- China
- Prior art keywords
- pixel
- image
- moving object
- treatment
- frame
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention belongs to technical field of image processing, a kind of method for improving moving object segmentation precision in image is especially provided, each two field picture to gathering carries out pixel level picture steadiness and judges, if stabilization, image procossing is then carried out using Vibe algorithms, otherwise, image procossing is carried out using frame differential method, morphology filling treatment finally is carried out to the image after the image after frame differential method is processed or Vibe algorithm process；Improve the defect of Vibe algorithms presence using frame differential method and morphology filling, the real-time detection precision of moving object in image can be improved.
Description
Technical field
The invention belongs to technical field of image processing, more particularly to a kind of side for improving moving object segmentation precision in image
Method.
Background technology
Currently, intelligent monitoring technology no matter company and enterprise security monitoring, or society traffic administration, also or
It is that more and more important effect is all played in the life ＆ amusement of people.
Intelligent monitoring technology all possesses important Research Significance in terms of theoretical and practical application.For example, from security standpoint
Say, intelligent monitoring video can carry out effective analysis to video content, extract the potential danger target in video.Such as silver
Possible robber in row, or monitor other moving targets in place.Security control personnel, Ke Yiyou are given these warning reports
Effect improves the efficiency of video monitoring, while the correlative charges of security control can also be saved.From the point of view of traffic administration angle, intelligence
Video technique can carry out intelligent analysis, the chocking-up degree of identification monitoring road, or surveillance and tracking row violating the regulations to monitoring road conditions
Vehicle is sailed, these can help the traffic administration personnel more effectively to carry out traffic administration.In terms of the amusement and recreation of people, can
To recognize the limb action of people by intelligent video, interacted according to the action and games system that identify, by camera
The body kinematics of identification people is shot, the manipulation to playing is completed.
Moving object detection in intelligent monitoring, refer to detected in sequence image region of variation and by moving target from
Extracted in background image, the moving object detection in intelligent monitoring is movement image analysis, intelligent monitoring, visual man-machine friendship
Key technology in mutually.The movable information in image can be obtained by the detection of moving target, moving target in image is extracted
And target is positioned, so as to simplify the difficulty of follow-up motion tracking, identification, analysis.
Conventional motion object detection method mainly includes three kinds：Background subtraction, frame differential method, optical flow method.In recent years
Come, Vibe algorithms are gradually paid attention in background subtraction.
Vibe algorithms are a kind of foreground detection of Pixel-level (or background modeling) methods, and its basic thought includes：For each
Individual pixel stores a sample set, and what the sample set was included is the picture of the past pixel value of this pixel and its surrounding neighbours point
Element value；When a new pixel is run into, just the collection point in the pixel value and sample set of new pixel is contrasted, sentenced
Whether disconnected this new pixel is background dot.The algorithm is the process to the classification of background dot.Vibe algorithms have computing to imitate
The advantages of rate is high, good, committed memory of realizing effect is few and sample decay is optimal.But Vibe algorithms are present by illumination, the moon in environment
Ghost (being moving target by former frame moving object institutes overlay area error detection), missing inspection (are caused by the mutation of the factors such as shadow
Refer to that some moving objects are not detected in IMAQ) the problems such as.Also, Vibe algorithms are in detection moving object
When, the detection of moving object can be made cavity occur and (to be referred to that object is in itself motion in detection, but shown after detecting
Some parts of moving object are but static).As in Fig. 3, whole people is motion originally, so whole people should be white
Color, but the head of people has part black as cavitation during detection.
The content of the invention
To solve the deficiency of tradition Vibe algorithms, the present invention proposes a kind of method for improving motion detection accuracy in image.
A kind of method for improving moving object segmentation precision in image of the present invention, each two field picture to gathering carries out pixel
The picture steadiness judgement of point level, if stabilization, then image procossing is carried out using Vibe algorithms, otherwise, using frame differential method
Image procossing is carried out, form finally is carried out to the image after the image after frame differential method is processed or Vibe algorithm process
Learn filling treatment.
Preferably, pixel level picture steadiness judge to include the pixel value of each pixel in present frame with it is preceding
The pixel value of one frame same position pixel is compared, if the number of the pixel of change is less than pixel recognition threshold,
It is judged as stable state, is otherwise judged as unstable state.
Preferably, the frame differential method includes that difference is carried out with the two continuous frames image in image sequence obtains gray scale difference
Component, then the binaryzation grey scale difference image zooming-out movable information, the image district for obtaining is split by interframe Changing Area Detection
Background area and moving object region are separated, and then extracts the moving target to be detected.
Preferably, image filtering treatment is carried out to the image after frame differential method treatment, it is specific to use neighborhood method, i.e., one by one
Pixel in image after scanning frame differential method treatment in neighborhood, by the pixel value of each pixel of pixel neighborhood of a point from
It is small to try to achieve the median of the pixel value of each pixel to being ranked up greatly or from big to small, the median is assigned to frame-to-frame differences
Pixel corresponding with current point in image after point-score treatment.
Preferably, the neighborhood is that, with current pixel point as the center of circle, 3 length of pixel are the border circular areas of radius.
Preferably, image filtering treatment is carried out to the image after frame differential method treatment, it is specific using two-dimentional sleiding form
Method, pixel will be ranked up in two dimension pattern plate W, the 2-D data sequence of generation monotone increasing or decline according to the size of pixel value
Median is arranged and calculates, two dimension median filter exports g (x, y)=med { f (x-k, y-l), (k, l ∈ W) }, wherein, f (x, y), g
(x, y) is respectively image slices vegetarian refreshments after original image pixels are selected and processed, and W is two dimension pattern plate, and x, y represent the coordinate of pixel
Value, k, l are expressed as the step value in two-dimensional reticle, and med { } is expressed as arranging the pixel value of the pixel in two-dimensional reticle
Median after sequence.
Preferably, the morphology filling treatment includes opening operation, i.e., first corrode the treatment for expanding afterwards；
The corrosion is included with each pixel in a structural element scan image, with each in structural element
The pixel that pixel is covered with it does AND-operation, if being all 1, the pixel is 1, is otherwise 0；
The expansion is included with each pixel in a structural element scan image, with each in structural element
The pixel that pixel is covered with it does AND-operation, if being all 0, the pixel is 0, is otherwise 1.
Preferably, the morphology filling treatment includes closed operation, i.e., first expand the treatment of post-etching；
The corrosion is included with each pixel in a structural element scan image, with each in structural element
The pixel that pixel is covered with it does AND-operation, if being all 1, the pixel is 1, is otherwise 0；
The expansion is included with each pixel in a structural element scan image, with each in structural element
The pixel that pixel is covered with it does AND-operation, if being all 0, the pixel is 0, is otherwise 1.
Preferably, whether to terminating collection and judging, if do not gather terminating, collection, and the number that will be gathered are proceeded
According to being saved in buffering area, whole process is repeated；If collection terminates, the data in acquisition buffer area are saved in internal memory,
And discharge buffering area internal memory, end operation.
Compared with prior art, the present invention improves lacking for Vibe algorithms presence using frame differential method and morphology filling
Fall into, high degree improves the accuracy of detection of moving object in image.
Brief description of the drawings
Fig. 1 is the method first preferred embodiment flow chart that the present invention improves moving object segmentation precision in image；
Fig. 2 is method the second preferred embodiment flow chart that the present invention improves moving object segmentation precision in image；
Fig. 3 is the image comparison figure after original image and frame differential method treatment；
Fig. 4 is the image comparison figure after original image and morphology filling treatment.
Specific embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, below in conjunction with accompanying drawing to of the invention real
Apply example further description.
As shown in figure 1, the present invention carries out IMAQ first, pixel level figure is then carried out to each two field picture for gathering
Picture judgement of stability, if stabilization, then image procossing is carried out using Vibe algorithms, otherwise, figure is carried out using frame differential method
As treatment, morphology filling finally is carried out to the image after the image after frame differential method is processed or Vibe algorithm process
Treatment, so as to eliminate the cavitation produced in motion detection process, more than by way of improve detection moving object
Precision.
For IMAQ, can be saved in the image of collection in the internal memory of computer by camera by the present invention.Realizing
When, the application programming interface (Application of the webcam driver that can be provided by using windows systems
Programming Interface, abbreviation API) camera is directly opened, carry out IMAQ.For example, opening camera device simultaneously
Camera collection image parameter is set；It is mainly every here, it is necessary to give camera arrange parameter after program is when opening camera
The pixel value of frame.Run in windows systems due to program, and the camera device for being used is that system carries driving
, eliminate the time overhead for needing oneself to write webcam driver code, it is only necessary to windows systems are used in program code
The api interface of the webcam driver for providing of uniting.
Judge for pixel level picture steadiness, including each pixel in present frame pixel value and former frame phase
It is compared with the pixel value of position pixel, if the number of the pixel of change is less than pixel recognition threshold, is judged as
Stable state, is otherwise judged as unstable state
When treatment multiple image is repeated, when moving object is detected for unstable state, frame differential method is switched to,
Then when detecting moving object and having stablized, then reinitialize background sample collection and be switched to Vibe algorithms.
Because Vibe algorithms are initialized from the first two field picture, when just having started initialization, whole collection can be given tacit consent to
Image be i.e. whole image of foreground image all be motion.Accordingly, it would be desirable to several two field pictures carry out the extensive of scene to compare below
It is multiple, it is judged the foreground image and background image of whole scene, and isolate foreground image and background image.When there is motion
After object enters into acquisition zone, pixel level picture steadiness is carried out to image first and is judged.If unstable rule uses frame-to-frame differences
Point-score, if stabilization, is switched to Vibe algorithms, if being judged as stabilization, shows that the object of now whole motion has been in stabilization
State, will eliminate when at this moment being detected to moving image using Vibe algorithms and drag shadow.
For frame differential method, the region changed in adjacent two field pictures is detected.The method is to use image sequence
In two continuous frames image carry out difference obtain gray scale difference component, then the binaryzation grey scale difference image come extract motion letter
Breath, the image obtained by the segmentation of interframe Changing Area Detection distinguishes background area and moving object region, and then extraction will be examined
The moving target of survey.
Frame differential method of the present invention is by front and rear two field pictures corresponding pixel points gray value, two frames in movement images sequence
Corresponding pixel points gray value subtracts each other, if difference very little, it is believed that this without motion object passes through, otherwise grey scale change is very
Greatly, then it is assumed that there is object to pass through.
For example, kth frame and k+1 two field pictures fk(x, y), fk+1Change between (x, y) with a two-value difference image D (x,
Y) represent, such as formula (1):
T is difference image binary-state threshold, because background model need not be updated, when moving object is unstable
Frame differential method can be preferably considered as.
Above-described embodiment adds frame differential method in traditional Vibe algorithms, and lacking for shadow is dragged so as to improve Vibe algorithms and exist
Fall into.
Preferably as optional mode, the present invention carries out image filtering treatment to the image after frame differential method treatment, such as
Shown in Fig. 2, can be carried out using various ways, including but not limited to neighborhood method, two-dimentional sleiding form method etc..
The neighborhood method, including procedure below：
First, the first address of source images and the wide and height of image are obtained.
Secondly, one piece of core buffer is opened up, is configured to temporarily store result images, and be initialized as 0.Then, scanning figure one by one
Pixel as in, by being ranked up from small to large for its neighborhood each element, the median that will be tried to achieve is assigned in target image
Pixel corresponding with current point.The neighborhood is that, with current pixel point as the center of circle, 3 length of pixel are the circle of radius
Region.
Then previous step is circulated, the whole pixels until having processed source images.
Finally, in the data buffer zone for result being copied into source images from core buffer.
The neighborhood method that the present embodiment is used is the medium filtering means based on spatial domain.Median filtering method is a kind of non-linear
Smoothing technique, the pixel value of each pixel is set to all pixels point pixel value in the range of the point adjacent domain for it
Intermediate value.Medium filtering is based on a kind of theoretical nonlinear signal processing technology that can effectively suppress noise of sequencing statistical, intermediate value
The general principle of filtering is each pixel in a neighborhood the pixel value of any in digital picture or Serial No. with the point
The Mesophyticum of pixel value replace, the actual value for making the pixel value of surrounding pixel point close, so as to eliminate isolated noise spot.
According to the size of pixel value be ranked up pixel in template by the two-dimentional sleiding form method, generates monotone increasing
The 2-D data sequence of (or decline).Two dimension median filter exports g (x, y)=med { f (x-k, y-l), (k, l ∈ W) }, wherein,
F (x, y), g (x, y) are respectively original image pixels and select and image slices vegetarian refreshments after processing, and W is two dimension pattern plate, and usually 3 × 3,5
× 5 regions, or different shapes, such as wire, circular, cross, annular etc.；X, y represent the coordinate value of pixel,
K, l are expressed as the step value in two-dimensional reticle, after med { } is expressed as being ranked up the pixel value of the pixel in two-dimensional reticle
Median.
For by the Vibe algorithms after improvement moving target can be detected, but the characteristics of due to Vibe algorithms itself
And and reason the characteristics of frame differential method, the target detected when detecting the object of motion using Vibe algorithms is often
There is cavitation, although the phenomenon in cavity is not it is obvious that but being implicitly present in.As Fig. 3 is original image (left side) and is used
Image (right side) after frame differential method treatment, from entirely seeing that profile is apparent from, treatment effect preferably, but still has cavitation
Presence, particularly head cavitation is more serious.It is then necessary to the cavitation to existing is further processed.
The present invention is using the method for morphology filling to the image after frame differential method is processed or Vibe algorithms
Image after treatment or the image after processing after filtering are filled, and to eliminate cavitation, further improve detection fortune
The precision of animal body.
Morphology filling using one be referred to as structural element " probe " collect image information, when " probe " in the picture
When constantly mobile, relation that just can be between image under consideration various pieces, so as to understand the architectural feature of image.Its core is thought
Want to include opening operation and closed operation.
The computation model of opening operation is：
Dst1=open (src, element)=dilate (erode (src, element)) (2)
Open () represents opening operation wherein in formula (2), and src represents artwork, the figure after element representatives treatment,
Dilate () represents etching operation, and erode () represents expansive working
Opening operation is first to corrode the process for expanding afterwards.For eliminating wisp, the separating objects, smooth larger at very thin point
Substantially do not change its area while the border of object.Opening operation is typically needing to remove little particle noise, and disconnection
Used when being connected between object, have the advantages that substantially keep the original size of target constant.
The computation model of closed operation is：
Dst2=close (src, element)=dilate (erode (src, element)) (3)
Close () represents closed operation wherein in formula (3), and src represents artwork, the figure after element representatives treatment,
Dilate () represents etching operation, and erode () represents expansive working
Closed operation is the process for first expanding post-etching, can be with cavity tiny in filler body, and smooth object border.
It is preferred that, can also first carry out carrying out closed operation again after opening operation, per two field picture by opening operation simultaneously
And after closed operation filling, the cavitation of moving object will be filled in whole picture.
The effect of the corrosion is to eliminate object boundary point, makes shrinking of object, can eliminate the noise less than structural element
Point.The concrete operations of corrosion are：With each pixel in a structural element (usually 3 × 3 sizes) scan image, use
The pixel that each pixel in structural element is covered with it does AND-operation, if being all 1, the pixel is 1, is otherwise 0.
The effect of the expansion is that all background dots contacted with object are merged into object, increases target, can be added
Mend the cavity in target.The concrete operations of expansion are：With in a structural element (usually 3 × 3 sizes) scan image
Each pixel, the pixel covered with it with each pixel in structural element does AND-operation, if being all 0, the picture
Element is 0, is otherwise 1.
Picture can be filled well by the cavitation of whole image after morphology filling treatment so that whole figure
Piece is by after present invention treatment, moving object can be good at displaying.If Fig. 4 is original image (left side) and by morphology
Image (right side) after filling treatment, the profile of whole moving object is perfectly clear, and, cavitation is reduced relative to before processing
A lot, improvement effect is obvious.
Finally, whether to gathering end and judging, if collection does not terminate, collection, and the number that will be gathered are proceeded
According to being saved in buffering area, operation above is repeated；If collection terminates, pass hull closure, and by the data in acquisition buffer area
It is saved in internal memory, and discharges buffering area internal memory, end operation.
The present invention improves the defect of Vibe algorithms presence using frame differential method and morphology filling, can improve image
The real-time detection precision of middle moving object.
The object, technical solutions and advantages of the present invention have been carried out further detailed description, institute by embodiment provided above
It should be understood that embodiment provided above is only the preferred embodiment of the present invention, be not intended to limit the invention, it is all
Any modification, equivalent substitution and improvements made for the present invention etc., should be included in the present invention within the spirit and principles in the present invention
Protection domain within.
Claims (9)
1. it is a kind of improve image in moving object segmentation precision method, it is characterised in that：Each two field picture to gathering is carried out
Pixel level picture steadiness judgement, if stabilization, then image procossing is carried out using Vibe algorithms, otherwise, using frame-to-frame differences
Point-score carries out image procossing, and finally the image after the image after frame differential method is processed or Vibe algorithm process is carried out
Morphology filling is processed.
2. the method for improving moving object segmentation precision in image according to claim 1, it is characterised in that：The pixel
Level picture steadiness judges to include the pixel of the pixel value with former frame same position pixel of each pixel in present frame
Value is compared, if the number of the pixel of change is less than pixel recognition threshold, is judged as stable state, is otherwise judged as
Unstable state.
3. the method for improving moving object segmentation precision in image according to claim 1, it is characterised in that：The frame-to-frame differences
Point-score includes that difference is carried out with the two continuous frames image in image sequence obtains gray scale difference component, then the binaryzation grey scale difference
Image zooming-out movable information, the image obtained by the segmentation of interframe Changing Area Detection distinguishes background area and moving object area
Domain, and then extract the moving target to be detected.
4. the method for improving moving object segmentation precision in image according to claim 1, it is characterised in that：To inter-frame difference
Image after method treatment carries out image filtering treatment, specific to use neighborhood method, i.e., figure one by one after the treatment of scanning frame differential method
Pixel as in neighborhood, the pixel value of each pixel of pixel neighborhood of a point is arranged from small to large or from big to small
Sequence, tries to achieve the median of the pixel value of each pixel, and the median is assigned in the image after frame differential method treatment and is worked as
The corresponding pixel of preceding point.
5. the method for improving moving object segmentation precision in image according to claim 4, it is characterised in that：The neighborhood is
With current pixel point as the center of circle, 3 length of pixel are the border circular areas of radius.
6. the method for improving moving object segmentation precision in image according to claim 1, it is characterised in that：To inter-frame difference
Image after method treatment carries out image filtering treatment, specific using two-dimentional sleiding form method, will in two dimension pattern plate W pixel according to
The size of pixel value is ranked up, and the 2-D data sequence of generation monotone increasing or decline simultaneously calculates median, two-dimentional intermediate value filter
Ripple exports g (x, y)=med { f (x-k, y-l), (k, l ∈ W) }, wherein, f (x, y), g (x, y) are respectively original image pixels point
With image slices vegetarian refreshments after treatment, W is two dimension pattern plate, and x, y represent the coordinate value of pixel, and k, l are expressed as the step in two-dimensional reticle
Long value, med { } is expressed as the median after being ranked up to the pixel value of the pixel in two-dimensional reticle.
7. the method for improving moving object segmentation precision in image according to claim 1, it is characterised in that：The morphology
Filling treatment includes opening operation, i.e., first corrode the treatment for expanding afterwards；
The corrosion is included with each pixel in a structural element scan image, with each pixel in structural element
The pixel covered with it does AND-operation, if being all 1, the pixel is 1, is otherwise 0；
The expansion is included with each pixel in a structural element scan image, with each pixel in structural element
The pixel covered with it does AND-operation, if being all 0, the pixel is 0, is otherwise 1.
8. the method for improving moving object segmentation precision in image according to claim 1, it is characterised in that：The morphology
Filling treatment includes closed operation, i.e., first expand the treatment of post-etching；
The corrosion is included with each pixel in a structural element scan image, with each pixel in structural element
The pixel covered with it does AND-operation, if being all 1, the pixel is 1, is otherwise 0；
The expansion is included with each pixel in a structural element scan image, with each pixel in structural element
The pixel covered with it does AND-operation, if being all 0, the pixel is 0, is otherwise 1.
9. the method for improving moving object segmentation precision in image according to claim 1, it is characterised in that：To whether terminating
Collection is judged, if do not gather terminating, proceeds collection, and the data of collection are saved in buffering area, repeats whole
Process；If collection terminates, the data in acquisition buffer area are saved in internal memory, and discharge buffering area internal memory, end operation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710140027.4A CN106934819A (en) | 2017-03-10 | 2017-03-10 | A kind of method of moving object segmentation precision in raising image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710140027.4A CN106934819A (en) | 2017-03-10 | 2017-03-10 | A kind of method of moving object segmentation precision in raising image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106934819A true CN106934819A (en) | 2017-07-07 |
Family
ID=59433563
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710140027.4A Pending CN106934819A (en) | 2017-03-10 | 2017-03-10 | A kind of method of moving object segmentation precision in raising image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106934819A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107249122A (en) * | 2017-07-26 | 2017-10-13 | 杨建伟 | A kind of moving target monitoring system for intelligent monitoring |
| CN108280841A (en) * | 2018-01-16 | 2018-07-13 | 北京联合大学 | A kind of foreground extracting method based on neighborhood territory pixel intensity correction |
| CN108303745A (en) * | 2018-03-19 | 2018-07-20 | 贵州电网有限责任公司 | A kind of inversion method of the buried cable detection based on electromagnetic wave saturating ground technology |
| CN109934870A (en) * | 2019-01-30 | 2019-06-25 | 西安天伟电子系统工程有限公司 | Object detection method, device, equipment, computer equipment and storage medium |
| CN110599523A (en) * | 2019-09-10 | 2019-12-20 | 江南大学 | ViBe ghost suppression method fused with interframe difference method |
| CN113570578A (en) * | 2021-07-29 | 2021-10-29 | 歌尔光学科技有限公司 | Lens ghost phenomenon detection method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101621615A (en) * | 2009-07-24 | 2010-01-06 | 南京邮电大学 | Self-adaptive background modeling and moving target detecting method |
| CN103824070A (en) * | 2014-03-24 | 2014-05-28 | 重庆邮电大学 | Rapid pedestrian detection method based on computer vision |
| CN104331905A (en) * | 2014-10-31 | 2015-02-04 | 浙江大学 | Surveillance video abstraction extraction method based on moving object detection |
-
2017
- 2017-03-10 CN CN201710140027.4A patent/CN106934819A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101621615A (en) * | 2009-07-24 | 2010-01-06 | 南京邮电大学 | Self-adaptive background modeling and moving target detecting method |
| CN103824070A (en) * | 2014-03-24 | 2014-05-28 | 重庆邮电大学 | Rapid pedestrian detection method based on computer vision |
| CN104331905A (en) * | 2014-10-31 | 2015-02-04 | 浙江大学 | Surveillance video abstraction extraction method based on moving object detection |
Non-Patent Citations (1)
| Title |
|---|
| 谢红等: "一种基于改进三帧差分和ViBe算法的运动目标检测算法", 《应用科技》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107249122A (en) * | 2017-07-26 | 2017-10-13 | 杨建伟 | A kind of moving target monitoring system for intelligent monitoring |
| CN108280841A (en) * | 2018-01-16 | 2018-07-13 | 北京联合大学 | A kind of foreground extracting method based on neighborhood territory pixel intensity correction |
| CN108280841B (en) * | 2018-01-16 | 2022-03-29 | 北京联合大学 | Foreground extraction method based on neighborhood pixel intensity correction |
| CN108303745A (en) * | 2018-03-19 | 2018-07-20 | 贵州电网有限责任公司 | A kind of inversion method of the buried cable detection based on electromagnetic wave saturating ground technology |
| CN109934870A (en) * | 2019-01-30 | 2019-06-25 | 西安天伟电子系统工程有限公司 | Object detection method, device, equipment, computer equipment and storage medium |
| CN110599523A (en) * | 2019-09-10 | 2019-12-20 | 江南大学 | ViBe ghost suppression method fused with interframe difference method |
| CN113570578A (en) * | 2021-07-29 | 2021-10-29 | 歌尔光学科技有限公司 | Lens ghost phenomenon detection method and device |
| CN113570578B (en) * | 2021-07-29 | 2024-05-17 | 歌尔光学科技有限公司 | Method and device for detecting lens ghost phenomenon |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106934819A (en) | A kind of method of moving object segmentation precision in raising image | |
| CN104392468B (en) | Based on the moving target detecting method for improving visual background extraction | |
| Mikic et al. | Moving shadow and object detection in traffic scenes | |
| JP4668921B2 (en) | Object detection in images | |
| CN110874592B (en) | Forest fire smoke image detection method based on total bounded variation | |
| US8280106B2 (en) | Shadow and highlight detection system and method of the same in surveillance camera and recording medium thereof | |
| WO2022027931A1 (en) | Video image-based foreground detection method for vehicle in motion | |
| CN110517288A (en) | Real-time target detecting and tracking method based on panorama multichannel 4k video image | |
| CN106846359A (en) | Moving target method for quick based on video sequence | |
| CN108022249B (en) | Automatic extraction method for target region of interest of remote sensing video satellite moving vehicle | |
| CN103903278A (en) | Moving target detection and tracking system | |
| CN103093198B (en) | A kind of crowd density monitoring method and device | |
| CN104063885A (en) | Improved movement target detecting and tracking method | |
| CN104952256A (en) | Video information based method for detecting vehicles at intersection | |
| CN110309765B (en) | High-efficiency detection method for video moving target | |
| KR20170015299A (en) | Method and apparatus for object tracking and segmentation via background tracking | |
| CN108198208A (en) | A kind of mobile detection method based on target following | |
| CN111539980A (en) | Multi-target tracking method based on visible light | |
| CN111967394A (en) | Forest fire smoke root node detection method based on dynamic and static grid fusion strategy | |
| CN110751635A (en) | Oral cavity detection method based on interframe difference and HSV color space | |
| CN116229359A (en) | Smoke identification method based on improved classical optical flow method model | |
| CN109583414B (en) | Indoor road occupation detection method, device, medium and processor based on video detection | |
| CN112801021B (en) | Method and system for detecting lane line based on multi-level semantic information | |
| Balisavira et al. | Real-time object detection by road plane segmentation technique for ADAS | |
| CN113657264A (en) | Forest fire smoke root node detection method based on fusion of dark channel and KNN algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170707 |