CN108537823A - Moving target detecting method based on mixed Gauss model - Google Patents
Moving target detecting method based on mixed Gauss model Download PDFInfo
- Publication number
- CN108537823A CN108537823A CN201710122544.9A CN201710122544A CN108537823A CN 108537823 A CN108537823 A CN 108537823A CN 201710122544 A CN201710122544 A CN 201710122544A CN 108537823 A CN108537823 A CN 108537823A
- Authority
- CN
- China
- Prior art keywords
- model
- weight
- matching
- pixel value
- image
- 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.)
- Withdrawn
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of moving target detecting methods based on mixed Gauss model, include the following steps：S1：Piecemeal processing is carried out with W × W block sizes to image, a mixed Gauss model is established to each piece；S2：Edge blurry processing is carried out to image；S3：New pixel value is compared with current K Gauss model, finds the distributed model of matching new pixel value；S4：If the matched model of institute meets context request, which belongs to background, otherwise belongs to foreground；S5：Each model weights are updated, each Model Weight is normalized；S6：Non- Matching Model mean μ and standard deviation sigma are constant, and Matching Model parameter is updated；S7：If not having any model matching in step S3, the model of weight minimum is replaced；S8：Each model is arranged according to ω/σ descendings, and the model that weight is big, standard deviation is small comes front；S9：B model indicates background proportion as background, parameter T before choosing.
Description
Technical field
Present invention relates particularly to a kind of moving target detecting methods based on mixed Gauss model.
Background technology
Moving object detection is a key technology of field of intelligent video surveillance, and research pair is provided for follow-up behavioural analysis
As, accuracy of detection directly influences system performance, common moving object detection algorithm have background subtractions, neighbor frame difference method,
Optical flow method, ViBe（Visual Background Extractor）, mixed Gaussian background modeling method etc..Background subtraction is opposite
Simply, realization is easier, and arithmetic speed is fast, but it does not have context update mechanism, causes Acquiring motion area inaccurate；It is adjacent
Frame difference method realizes that easy and calculation amount is small using adjacent two field pictures difference to extract moving region, but the method can introduce smear
Region；Optical flow method carries out motion detection by calculating the optical flow field of image, and time complexity is higher, is not suitable for real-time and wants
Ask high application；ViBe is a kind of background modeling algorithm of Pixel-level, and calculation amount is small, and time complexity is low, and anti-noise ability is good, energy
It is enough preferably to adapt to the mutation of background, but it is readily incorporated smear region；Mixed Gaussian background modeling method（Gaussian Mixture
Model, GMM）It is a kind of background representation method based on pixels statistics information, there is stronger robustness to noise, but it is counted
Calculation amount is big, and real-time is poor.
Invention content
The technical problem to be solved in the present invention is to provide a kind of moving target detecting methods based on mixed Gauss model.
Moving target detecting method based on mixed Gauss model, includes the following steps：
S1：Piecemeal processing is carried out with W × W block sizes to image, a mixed Gauss model, i.e. K Gauss are established to each piece
Model；
S2：Edge blurry processing is carried out to image；
S3：New pixel valueIt is compared as the following formula with current K Gauss model, finds the distributed model of matching new pixel value,
I.e. with the mean bias of the model within 2.5 σ；
；
S4：If the matched model of institute meets context request, which belongs to background, otherwise belongs to foreground；
S5：With probabilityEach model weights are updated as follows, wherein α is learning rate,
For matched model, otherwise, then each Model Weight is normalized；
；
S6：Non- Matching Model mean μ and standard deviation sigma are constant, and Matching Model parameter is as follows with probabilityIt is updated；
；
；
；
S7：If there is no any model matching in step S3, with probabilityBy the mould of weight minimum
Type is replaced, i.e., the model mean value is arranged to current pixel value, standard deviation and weight are arranged to initial value；
S8：Each model is arranged according to ω/σ descendings, and the model that weight is big, standard deviation is small comes front；
S9：For B model as background, B meets following formula before choosing, and parameter T indicates background proportion；
。
Further, the method for building up of mixed Gauss model is as follows：
Assuming that t moment, pixelValue collection is combined into, wherein
I is video flowing, if characterizing each pixel of image with K Gaussian Profile, the probability of each pixel is：
；
Wherein, K is Gaussian Profile number,For i-th of Gaussian Profile t moment weight, meetAnd it weighs
Weight summation is 1,For t moment pixel value,、It is distributed as i-th of Gaussian Profile t moment mean value and variance,For the probability density of i-th of Gaussian Profile, it is defined as follows：
,。
Further, edge blurry processing is realized using mean filter, specific as follows：
Mean filter process is given by：
,
Wherein, F isOriginal image, U is（M, n are odd number）Filter template, G be output image.
The beneficial effects of the invention are as follows：
Present invention proposition is improved conventional hybrid Gauss model with probability updating strategy using piecemeal treatment technology, improves
Algorithm is taking the spatial domain letter for having obtained very big improvement in terms of the two with memory space, while being effectively utilized between image pixel
Breath, show preferable adaptive capacity to environment, can more completely, accurately extract moving target, improve simultaneously
Time ＆ Space Complexity.
Specific implementation mode
The present invention is further elaborated for following specific examples, but not as a limitation of the invention.
Moving target detecting method based on mixed Gauss model, includes the following steps：
S1：Piecemeal processing is carried out with W × W block sizes to image, a mixed Gauss model, i.e. K Gauss are established to each piece
Model；
S2：Edge blurry processing is carried out to image；
S3：New pixel valueIt is compared as the following formula with current K Gauss model, finds the distributed model of matching new pixel value,
I.e. with the mean bias of the model within 2.5 σ；
；
S4：If the matched model of institute meets context request, which belongs to background, otherwise belongs to foreground；
S5：With probabilityEach model weights are updated as follows, wherein α is learning rate,
For matched model, otherwise, then each Model Weight is normalized；
；
S6：Non- Matching Model mean μ and standard deviation sigma are constant, and Matching Model parameter is as follows with probabilityIt is updated；
；
；
；
S7：If there is no any model matching in step S3, with probabilityBy the mould of weight minimum
Type is replaced, i.e., the model mean value is arranged to current pixel value, standard deviation and weight are arranged to initial value；
S8：Each model is arranged according to ω/σ descendings, and the model that weight is big, standard deviation is small comes front；
S9：For B model as background, B meets following formula before choosing, and parameter T indicates background proportion；
。
The method for building up of mixed Gauss model is as follows：
Assuming that t moment, pixelValue collection is combined into, wherein
I is video flowing, if characterizing each pixel of image with K Gaussian Profile, the probability of each pixel is：
；
Wherein, K is Gaussian Profile number,For i-th of Gaussian Profile t moment weight, meetAnd it weighs
Weight summation is 1,For t moment pixel value,、It is distributed as i-th of Gaussian Profile t moment mean value and variance,For the probability density of i-th of Gaussian Profile, it is defined as follows：
,。
Edge blurry processing is realized using mean filter, specific as follows：
Mean filter process is given by：
,
Wherein, F isOriginal image, U is（M, n are odd number）Filter template, G be output image.
Claims (3)
1. the moving target detecting method based on mixed Gauss model, which is characterized in that include the following steps：
S1：Piecemeal processing is carried out with W × W block sizes to image, a mixed Gauss model, i.e. K Gauss are established to each piece
Model；
S2：Edge blurry processing is carried out to image；
S3：New pixel valueIt is compared as the following formula with current K Gauss model, finds the distributed model of matching new pixel value,
I.e. with the mean bias of the model within 2.5 σ；
；
S4：If the matched model of institute meets context request, which belongs to background, otherwise belongs to foreground；
S5：With probabilityEach model weights are updated as follows, wherein α is learning rate,
For matched model, otherwise, then each Model Weight is normalized；
；
S6：Non- Matching Model mean μ and standard deviation sigma are constant, and Matching Model parameter is as follows with probabilityIt is updated；
；
；
；
S7：If there is no any model matching in step S3, with probabilityBy the mould of weight minimum
Type is replaced, i.e., the model mean value is arranged to current pixel value, standard deviation and weight are arranged to initial value；
S8：Each model is arranged according to ω/σ descendings, and the model that weight is big, standard deviation is small comes front；
S9：For B model as background, B meets following formula before choosing, and parameter T indicates background proportion；
。
2. moving target detecting method according to claim 1, which is characterized in that the method for building up of mixed Gauss model is such as
Under：
Assuming that t moment, pixelValue collection is combined into, wherein
I is video flowing, if characterizing each pixel of image with K Gaussian Profile, the probability of each pixel is：
；
Wherein, K is Gaussian Profile number,For i-th of Gaussian Profile t moment weight, meetAnd weight
Summation is 1,For t moment pixel value,、It is distributed as i-th of Gaussian Profile t moment mean value and variance,For the probability density of i-th of Gaussian Profile, it is defined as follows：
,。
3. moving target detecting method according to claim 1, which is characterized in that edge blurry processing uses mean filter
It realizes, it is specific as follows：
Mean filter process is given by：
,
Wherein, F isOriginal image, U is（M, n are odd number）Filter template, G be output image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710122544.9A CN108537823A (en) | 2017-03-03 | 2017-03-03 | Moving target detecting method based on mixed Gauss model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710122544.9A CN108537823A (en) | 2017-03-03 | 2017-03-03 | Moving target detecting method based on mixed Gauss model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108537823A true CN108537823A (en) | 2018-09-14 |
Family
ID=63489080
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710122544.9A Withdrawn CN108537823A (en) | 2017-03-03 | 2017-03-03 | Moving target detecting method based on mixed Gauss model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108537823A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109543608A (en) * | 2018-11-22 | 2019-03-29 | 中国科学院西安光学精密机械研究所 | Hyperspectral marine small target real-time detection method based on Gaussian mixture model |
| CN109919964A (en) * | 2019-03-01 | 2019-06-21 | 南阳理工学院 | The method that Gaussian Background modeling technique based on mathematical morphology carries out image procossing |
| CN113240611A (en) * | 2021-05-28 | 2021-08-10 | 中建材信息技术股份有限公司 | Foreign matter detection method based on picture sequence |
| CN114693504A (en) * | 2022-04-11 | 2022-07-01 | 武汉大学 | Image processing method of mixed Gaussian model based on FPGA |
| CN114693504B (en) * | 2022-04-11 | 2024-07-26 | 武汉大学 | Image processing method of Gaussian mixture model based on FPGA |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101908141A (en) * | 2010-08-04 | 2010-12-08 | 丁天 | Video smoke detection method based on mixed Gaussian model and morphological characteristics |
| CN104867144A (en) * | 2015-05-15 | 2015-08-26 | 广东工业大学 | IC element solder joint defect detection method based on Gaussian mixture model |
| CN105976612A (en) * | 2016-04-27 | 2016-09-28 | 东南大学 | Urban traffic scene vehicle detection method based on robust mixed Gaussian model |
-
2017
- 2017-03-03 CN CN201710122544.9A patent/CN108537823A/en not_active Withdrawn
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101908141A (en) * | 2010-08-04 | 2010-12-08 | 丁天 | Video smoke detection method based on mixed Gaussian model and morphological characteristics |
| CN104867144A (en) * | 2015-05-15 | 2015-08-26 | 广东工业大学 | IC element solder joint defect detection method based on Gaussian mixture model |
| CN105976612A (en) * | 2016-04-27 | 2016-09-28 | 东南大学 | Urban traffic scene vehicle detection method based on robust mixed Gaussian model |
Non-Patent Citations (1)
| Title |
|---|
| 王慎波等: "基于块模型的混合高斯模型运动目标检测方法", 《信息技术》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109543608A (en) * | 2018-11-22 | 2019-03-29 | 中国科学院西安光学精密机械研究所 | Hyperspectral marine small target real-time detection method based on Gaussian mixture model |
| CN109543608B (en) * | 2018-11-22 | 2022-12-09 | 中国科学院西安光学精密机械研究所 | Hyperspectral marine small target real-time detection method based on Gaussian mixture model |
| CN109919964A (en) * | 2019-03-01 | 2019-06-21 | 南阳理工学院 | The method that Gaussian Background modeling technique based on mathematical morphology carries out image procossing |
| CN113240611A (en) * | 2021-05-28 | 2021-08-10 | 中建材信息技术股份有限公司 | Foreign matter detection method based on picture sequence |
| CN113240611B (en) * | 2021-05-28 | 2024-05-07 | 中建材信息技术股份有限公司 | Foreign matter detection method based on picture sequence |
| CN114693504A (en) * | 2022-04-11 | 2022-07-01 | 武汉大学 | Image processing method of mixed Gaussian model based on FPGA |
| CN114693504B (en) * | 2022-04-11 | 2024-07-26 | 武汉大学 | Image processing method of Gaussian mixture model based on FPGA |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Huang | An advanced motion detection algorithm with video quality analysis for video surveillance systems | |
| CN105740945B (en) | A kind of people counting method based on video analysis | |
| CN103325112B (en) | Moving target method for quick in dynamic scene | |
| Yi et al. | Moving object detection based on running average background and temporal difference | |
| CN106846359A (en) | Moving target method for quick based on video sequence | |
| CN108009493B (en) | Human face anti-cheating recognition method based on motion enhancement | |
| CN107301376B (en) | Pedestrian detection method based on deep learning multi-layer stimulation | |
| TWI441096B (en) | Motion detection method for comples scenes | |
| CN102663405B (en) | Prominence and Gaussian mixture model-based method for extracting foreground of surveillance video | |
| CN113762009B (en) | Crowd counting method based on multi-scale feature fusion and double-attention mechanism | |
| CN103208105B (en) | A kind of infrared image details strengthens and noise Adaptive Suppression method | |
| CN107067416A (en) | A kind of detection method of moving target | |
| Huang et al. | Motion detection with pyramid structure of background model for intelligent surveillance systems | |
| WO2014201971A1 (en) | Object detection method and device for online training | |
| CN110222572A (en) | Tracking, device, electronic equipment and storage medium | |
| CN108537823A (en) | Moving target detecting method based on mixed Gauss model | |
| CN108549901A (en) | A kind of iteratively faster object detection method based on deep learning | |
| CN105913002A (en) | On-line adaptive abnormal event detection method under video scene | |
| CN103886324B (en) | Scale adaptive target tracking method based on log likelihood image | |
| CN109558790B (en) | Pedestrian target detection method, device and system | |
| CN104036526A (en) | Gray target tracking method based on self-adaptive window | |
| CN107871315B (en) | Video image motion detection method and device | |
| CN105426928B (en) | A kind of pedestrian detection method based on Haar feature and EOH feature | |
| CN106203360A (en) | Intensive scene crowd based on multistage filtering model hives off detection algorithm | |
| CN103646397B (en) | Real-time synthetic aperture perspective imaging method based on multisource data fusion |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20180914 |