CN108508425A - Foreground target detection method based on neighborhood characteristics under a kind of radar near-earth ambient noise - Google Patents
Foreground target detection method based on neighborhood characteristics under a kind of radar near-earth ambient noise Download PDFInfo
- Publication number
- CN108508425A CN108508425A CN201810256423.8A CN201810256423A CN108508425A CN 108508425 A CN108508425 A CN 108508425A CN 201810256423 A CN201810256423 A CN 201810256423A CN 108508425 A CN108508425 A CN 108508425A
- Authority
- CN
- China
- Prior art keywords
- radar
- image
- background model
- pixels
- detection method
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
Foreground target detection method based on neighborhood characteristics under a kind of radar near-earth ambient noise, including：Continuous N frames radar image is obtained, and the pixel addition of all frames of N frame radar images is taken by the average estimation for being used as background by formula, to establish out the background model of fixed object reflectogram；Establish the sample set of each pixel under background model；The single sector image of current radar is obtained according to radar scanning line sequence；The image of current single sector is subjected to template matches with background model as template according to radar scanning line sequence and continues below step if successful match to find corresponding identical region；According to matching result, the distance of the pixel mean value and each sample value in corresponding sample set of the N*M block of pixels centered on the current pixel value of successful match sector is calculated, when distance is less than pre-determined distance threshold value R, then approximation sample point number increases, otherwise judge the block of pixels for background, it is on the contrary then be foreground.Detection method accuracy rate is high, and efficient.
Description
Technical field
The present invention relates to radar equipment fields, and in particular under a kind of radar near-earth ambient noise based on neighborhood characteristics before
Scape object detection method.
Background technology
Moving foreground object detection is always one of vision monitoring area research emphasis, and the purpose is to will from sequence image
Region of variation is extracted from background image, and effective detection of sport foreground object is for image tracing, target classification, behavior
The post-processings such as understanding are most important, distinguish a very crucial problem of foreground object and are to determine a suitable background, mesh
Preceding several common methods mainly have：Frame difference method, optical flow method etc..Frame difference method is difficult to extract the complete area of object, Zhi Nengti
Take out boundary；And object then can't detect when object is almost overlapped in front and back two frame for the object of microinching
Body；Optical flow method calculates complicated, processing in real time difficult to realize.
In Radar Technology field, radar image feature has quite big difference, radar to pass through transmitting with optical imagery feature
Electromagnetic wave and receive echo target is detected, radar image is then that receiver receives scatter echo and is formed by image；And
Camera is then to be then converted into electric signal by collecting the light of object reflection, therefore be imaged and be generally visible images.Therefore
Radar image includes only object reflected signal strength information and is not illuminated by the light variation influence, even if fixed in radar,
Form of the same object reflection signal in different scanning period in imaging can also have a greater change, and size and pulse are wide
Degree and beam angle are related, currently, in radar foreground target detection method, it is usually that existing foreground moving-target detection algorithm is straight
It scoops out and is used for radar image, cause the accuracy for the sport foreground extracted and efficiency not high.
Invention content
The present invention is provided and is based under a kind of radar near-earth ambient noise to solve the above problem of the existing technology
The foreground target detection method of neighborhood characteristics directly applies to radar image to solve existing foreground moving-target detection algorithm,
Lead to the accuracy for the sport foreground extracted and the problem that efficiency is not high.
To achieve the above object, the present invention provides the foreground mesh based on neighborhood characteristics under a kind of radar near-earth ambient noise
Detection method is marked, is included the following steps：
S1, continuous N frames radar image is obtained, and passes through formulaBy all of N frame radar images
The pixel addition of frame takes the average estimation as background, to establish out the background model of fixed object reflectogram, wherein BM is
Background model, N are to be currently located frame and not all frame；
S2, the sample set for establishing each pixel under background model；
S3, the single sector image that current radar is obtained according to radar scanning line sequence；
S4, template matches are carried out using the image of current single sector as template and background model according to radar scanning line sequence
Continue below step if successful match to find corresponding identical region；It abandons, and handles next if matching is unsuccessful
Sector data；
S5, according to matching result, the pixel for calculating the N*M block of pixels centered on the current pixel value of successful match sector is equal
The distance of each sample value in value and corresponding sample set, when distance is less than pre-determined distance threshold value R, then approximate sample point number increases
Add, and when approximate sample point number is more than predetermined threshold value #, then judges the block of pixels for background, it is on the contrary then be foreground.
As present invention further optimization technical solution, gone back after continuous N frames radar image is obtained in the step S1
Including：
The N frames radar image of acquisition is pre-processed respectively, the preprocess method includes low-pass filtering treatment and oneself
Adapt to threshold binarization treatment；
Pretreated image and reference picture are subjected to image registration, the reference picture is first frame image.
As present invention further optimization technical solution, the background of fixed object reflectogram is established out in the step S1
Model specifically includes：
Take different parameter combinations, establish multiple background models tool, wherein parameter include radar operation mode, pulsewidth,
Beam angle, gain, transmission power, season, meteorology.
As present invention further optimization technical solution, the step S2 establishes each pixel under background model
Sample set specifically includes：
If any pixel point is x under background model, 20 pixels are randomly selected from 24 neighborhoods of x, then calculate separately with this
Sample set of the pixel mean value of N*M block of pixels centered on 20 pixels as x：NG (x)=V1, V2,
V3..........V20 }, NG (x) is the sample set of x, and v1, v2.....v20 are 20 sample value V (x), wherein 24 neighborhoods are
Other 24 pixels of the block of pixels of 5*5 centered on x in addition to x, picture of the point in the block of pixels of the N*M of x centered on V (x)
Plain mean value, the mean value of the block of pixels are that all pixels point of block of pixels is added again divided by block of pixels area.
As present invention further optimization technical solution, obtained currently according to radar scanning line sequence in the step S3
Further include after the single sector image of radar：
The single sector image of acquisition is pre-processed, the preprocess method includes low-pass filtering treatment and adaptive thresholding
It is worth binary conversion treatment.
As present invention further optimization technical solution, during the template matches in the step S4, single sector
When image carries out template matches as template and background model, template need to be carried out in the background model that relevant parameter combines
Match.
As present invention further optimization technical solution, the foreground target detection method further includes：Using adaptive
The method of iteration updates background model.
As present invention further optimization technical solution, the background model update is counted using triangle recursive function
It calculates, formula is：
BM=(N-1)/N*BM+1/N*current
Wherein, BM is background model, and current is currently available input picture, and N is bigger, and renewal rate is slower, no
Then, renewal rate gets over block.
The foreground target detection method based on neighborhood characteristics can reach as follows under the radar near-earth ambient noise of the present invention
Advantageous effect：
1) compared with traditional ViBe algorithms are using the method for single frames modeling, the present invention uses mean value background modeling, the method
The advantages of be spatial and temporal distributions information that background image contains pixel, the echo-signal form factors of instability are taken into account
And the regions Ghost will not be introduced；
2) the imaging size of object is related to radar parameter in parameter matched pixel block comparison radar image, so in foreground
Used when detection block of pixels comparison replace single pixel comparison method, can fast filtering fall noise spot improve algorithm effect
Rate；
3) the traditional ViBe of parameter model is nonparametric model, but due to the pulsewidth and beam shape of radar image and radar
Correlation, and can be with the adjustment and use of radar parameter when weather conditions and generate larger variation, therefore in modified hydrothermal process
In we these change conditions are all set as adjustable parameter, so that algorithm is had preferably adaptive, parameter includes radar
Operating mode, pulsewidth, beam angle, gain, transmission power, season, meteorology (fine, cloudy, light rain, heavy rain, mist, sand and dust) etc.；
4) the traditional Video Detection Algorithm of sector models is handled in face of a frame image, with the proviso that passing through camera
The image of acquisition is disposably to obtain, and radar image is to scan to obtain according to location order, and the image formation period is longer, is
Identification delay is reduced, we are compared using single sector with background model rather than full figure comparison when carrying out foreground detection.
Description of the drawings
The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
Fig. 1 is the reality that the foreground target detection method based on neighborhood characteristics provides under radar near-earth ambient noise of the present invention
The method flow diagram of example；
The object of the invention is realized, the embodiments will be further described with reference to the accompanying drawings for functional characteristics and advantage.
Specific implementation mode
Below in conjunction with attached drawing and specific implementation mode, the present invention is described further.Drawn in preferred embodiment
Such as "upper", "lower", "left", "right", " centre " and " one " term, only being illustrated for ease of narration, rather than to limit
The enforceable range of the present invention, relativeness are altered or modified, in the case where changing technology contents without essence, when being also considered as this hair
Bright enforceable scope.
As shown in Figure 1, the foreground target detection method based on neighborhood characteristics includes following step under radar near-earth ambient noise
Suddenly：
Step S1, continuous N frames radar image is obtained, and passes through formulaBy N frame radar images
The pixel addition of all frames takes the average estimation as background, to establish out the background model of fixed object reflectogram, wherein
BM is background model, and N is to be currently located frame and not all frame；
Radar image is in transmission process, and image data uses network transmission in a sequential manner after being packaged, wherein radar one
Image obtained by a antenna scan period, i.e. radar image.
In specific implementation, in the step S1, further include after obtaining continuous N frames radar image：
The N frames radar image of acquisition is pre-processed respectively, the preprocess method includes low-pass filtering treatment and oneself
Adapt to threshold binarization treatment, wherein the sliding window of low-pass filter is dimensioned to the 1/4 of object minimum dimension, i.e., orientation and
Distance is respectively set as 1/2, and the threshold value of adaptive threshold binaryzation is chosen using the energy of radar image and radar gain etc. as ginseng
Number carries out adaptive；
Pretreated image and reference picture are subjected to image registration, the reference picture is first frame image, using figure
The image slight shift caused by radar shake etc. will definitely be solved the problems, such as matching, while improving system stability.
In specific implementation, in the step S1, the background model for establishing out fixed object reflectogram specifically includes：
Different parameter combinations is taken, establishes multiple background models tool, when use, each parameter should be corresponding to it, wherein ginseng
Number includes radar operation mode, pulsewidth, beam angle, gain, transmission power, season, meteorology, for example, defining radar background mould
Type is Clear Days Summer template under so-and-so working condition dust and sand weather of lower winter template or so-and-so working condition etc., due to radar image
It is related to the pulsewidth of radar and beam shape, and can be with the adjustment and use of radar parameter when weather conditions and generate larger
Variation, therefore these change conditions are all set as adjustable parameter by we in modified hydrothermal process, make algorithm can preferably certainly
It adapts to.
Step S2, the sample set of each pixel under background model is established；
In specific implementation, the sample set that the step S2 establishes each pixel under background model specifically includes：
If any pixel point is x under background model, 20 pixels are randomly selected from 24 neighborhoods of x, then calculate separately with this
The pixel mean value conduct of N*M (wherein, N, M choose related to pulse width and beam angle) block of pixels centered on 20 pixels
The sample set of x：NG (x)={ V1, V2, V3..........V20 }, NG (x) are the sample set of x, and v1, v2.....v20 are 20
Sample value V (x), wherein 24 neighborhoods are other 24 pixels of the block of pixels of the 5*5 centered on x in addition to x, during V (x) is
For heart point in the pixel mean value of the block of pixels of the N*M of x, the mean value of the block of pixels is that all pixels point of block of pixels is added again divided by picture
Plain block area.
Step S3, the single sector image of current radar is obtained according to radar scanning line sequence；
Radar image is to scan to obtain according to location order, and the image formation period is longer, in order to reduce identification delay,
We are compared using single sector and background model when carrying out foreground detection rather than full figure compares, and are obtained from network transmission port
Radar data by scan line sequence sort, buffering single sector data enter detection process.
In specific implementation, in the step S3, after the single sector image that current radar is obtained according to radar scanning line sequence
Further include：
The single sector image of acquisition is pre-processed, the preprocess method includes low-pass filtering treatment and adaptive thresholding
It is worth binary conversion treatment, single sector image is radar present image, and the full figure that radar image is radar, single sector image and thunder
Similar up to the radar image characteristic properties in initialization, the two is same radar after same position, different antennae scan period
Obtained image, therefore identical pre-treatment step can be used to carry out pretreatment denoising to it.
Step S4, template is carried out using the image of current single sector as template and background model according to radar scanning line sequence
Matching continues below step to find corresponding identical region if successful match；It abandons, and handles if matching is unsuccessful
Next sector data；Inevitably there are the factors such as shake in step S4 execution steps, therefore adjacent mainly due to because of radar
Two scan period obtained images might have offset；
In specific implementation, during the template matches in the step S4, the image of single sector is as template and background mould
When type carries out template matches, template matches, i.e. single sector image need to be carried out in the background model that relevant parameter combines, need to use
Background model under the parameter combination that is corresponding to it is matched.
Step S5, according to matching result, the picture of the N*M block of pixels centered on the current pixel value of successful match sector is calculated
The distance of each sample value in plain mean value and corresponding sample set, when distance is less than pre-determined distance threshold value R, then approximate sample point number
Increase, and when approximate sample point number is more than predetermined threshold value #, then judges the block of pixels for background, it is on the contrary then be foreground.
The predetermined threshold value # can specifically be set by designer, such as it can first pass through many experiments in advance and be met
The concrete numerical value of condition does not limit its design parameter herein.
In specific implementation, the foreground target detection method further includes：Using the method for adaptive iteration to background model
Update.
It is further preferred that the background model update is calculated using triangle recursive function, formula is：
BM=(N-1)/N*BM+1/N*current
Wherein, BM is background model, and current is currently available input picture, and N is bigger, and renewal rate is slower, no
Then, renewal rate gets over block.This algorithm uses non-conservative more new strategy, the target occurred in the short time will not be classified as carrying on the back
Scape, and for such as newly-built building of new fixed object, other large-scale fixed objects, also can after update after a period of time
It is included into background model.
Foreground detection compares current input image and the established template of model initialization, to extract foreground fortune
Moving-target, core of the invention thinking with traditional ViBe algorithms substantially compared with, the difference is that, initialization establish background model
In, a sample set is stored for each block of pixels, sampled value is exactly that the block of pixels corresponds to area in background model in sample set
The pixel mean value of the pixel mean value in domain and its corresponding block of pixels of neighbours' point, then by each new value (thunder currently inputted
Up to image pixel value) it is compared with sample set to determine whether belonging to background dot so that the accuracy of the detection of sport foreground
Height, excellent in efficiency.
Although specific embodiments of the present invention have been described above, those skilled in the art should be appreciated that this
It is merely illustrative of, various changes or modifications can be made to present embodiment, without departing from the principle and substance of the present invention,
Protection scope of the present invention is only limited by the claims that follow.
Claims (8)
1. the foreground target detection method based on neighborhood characteristics under a kind of radar near-earth ambient noise, which is characterized in that including with
Lower step：
S1, continuous N frames radar image is obtained, and passes through formulaBy all frames of N frame radar images
Pixel addition takes the average estimation as background, to establish out the background model of fixed object reflectogram, wherein BM is background
Model, N are to be currently located frame and not all frame；
S2, the sample set for establishing each pixel under background model；
S3, the single sector image that current radar is obtained according to radar scanning line sequence；
S4, template matches are carried out to seek using the image of current single sector as template and background model according to radar scanning line sequence
Corresponding identical region is looked for continue below step if successful match；It is abandoned if matching is unsuccessful, and handles next sector
Data；
S5, according to matching result, calculate the N*M block of pixels centered on the current pixel value of successful match sector pixel mean value and
The distance of each sample value in corresponding sample set, when distance is less than pre-determined distance threshold value R, then approximate sample point number increases, and
When approximate sample point number is more than predetermined threshold value #, then the block of pixels is judged for background, it is on the contrary then be foreground.
2. the foreground target detection method based on neighborhood characteristics under radar near-earth ambient noise according to claim 1,
It is characterized in that, further includes after continuous N frames radar image is obtained in the step S1：
The N frames radar image of acquisition is pre-processed respectively, the preprocess method includes low-pass filtering treatment and adaptive
Threshold binarization treatment；
Pretreated image and reference picture are subjected to image registration, the reference picture is first frame image.
3. the foreground target detection method based on neighborhood characteristics under radar near-earth ambient noise according to claim 2,
It is characterized in that, the background model that fixed object reflectogram is established out in the step S1 specifically includes：
Different parameter combinations is taken, establishes multiple background model tools, wherein parameter includes radar operation mode, pulsewidth, wave beam
Width, gain, transmission power, season, meteorology.
4. the foreground target detection method based on neighborhood characteristics under radar near-earth ambient noise according to claim 3,
It is characterized in that, the sample set that the step S2 establishes each pixel under background model specifically includes：
If any pixel point is x under background model, 20 pixels are randomly selected from 24 neighborhoods of x, then calculate separately with this 20
Sample set of the pixel mean value of N*M block of pixels centered on pixel as x：NG (x)={ V1, V2, V3..........V20 },
NG (x) is the sample set of x, and v1, v2.....v20 are 20 sample value V (x), wherein 24 neighborhoods are the 5*5's centered on x
Other 24 pixels of block of pixels in addition to x, centered on V (x) point the block of pixels of the N*M of x pixel mean value, the block of pixels
Mean value is that all pixels point of block of pixels is added again divided by block of pixels area.
5. the foreground target detection method based on neighborhood characteristics under radar near-earth ambient noise according to claim 4,
It is characterized in that, further includes after obtaining the single sector image of current radar in the step S3 according to radar scanning line sequence：
The single sector image of acquisition is pre-processed, the preprocess method includes low-pass filtering treatment and adaptive threshold two
Value is handled.
6. the foreground target detection method based on neighborhood characteristics under radar near-earth ambient noise according to claim 5,
It is characterized in that, during the template matches in the step S4, the image of single sector carries out template as template and background model
When matching, template matches need to be carried out in the background model that relevant parameter combines.
7. the foreground target detection method based on neighborhood characteristics under radar near-earth ambient noise according to claim 6,
It is characterized in that, the foreground target detection method further includes：Background model is updated using the method for adaptive iteration.
8. the foreground target detection method based on neighborhood characteristics under radar near-earth ambient noise according to claim 7,
It is characterized in that, the background model update is calculated using triangle recursive function, and formula is：
BM=(N-1)/N*BM+1/N*current
Wherein, BM is background model, and current is currently available input picture, and N is bigger, and renewal rate is slower, otherwise, more
New rate gets over block.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810256423.8A CN108508425B (en) | 2018-03-26 | 2018-03-26 | Method for detecting foreground target based on neighborhood characteristics under radar near-earth background noise |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810256423.8A CN108508425B (en) | 2018-03-26 | 2018-03-26 | Method for detecting foreground target based on neighborhood characteristics under radar near-earth background noise |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108508425A true CN108508425A (en) | 2018-09-07 |
| CN108508425B CN108508425B (en) | 2020-08-04 |
Family
ID=63378669
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810256423.8A Active CN108508425B (en) | 2018-03-26 | 2018-03-26 | Method for detecting foreground target based on neighborhood characteristics under radar near-earth background noise |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108508425B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113994404A (en) * | 2019-06-21 | 2022-01-28 | 松下电器产业株式会社 | Monitoring device and monitoring method |
| CN115320669A (en) * | 2022-08-31 | 2022-11-11 | 南京慧尔视智能科技有限公司 | Method, device, equipment and medium for detecting railway coming car based on radar map |
| CN115424455A (en) * | 2022-08-31 | 2022-12-02 | 南京慧尔视智能科技有限公司 | Vehicle early warning method and device, electronic equipment and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103745216A (en) * | 2014-01-02 | 2014-04-23 | 中国民航科学技术研究院 | Radar image clutter suppression method based on airspace feature |
| CN104199009A (en) * | 2014-09-18 | 2014-12-10 | 中国民航科学技术研究院 | Radar image clutter suppression method based on time-domain characteristics |
| CN105184814A (en) * | 2015-07-27 | 2015-12-23 | 成都天奥信息科技有限公司 | Moving target detecting and tracking method based on multi-frame radar image |
| CN106447680A (en) * | 2016-11-23 | 2017-02-22 | 湖南华诺星空电子技术有限公司 | Method for radar and vision fused target detecting and tracking in dynamic background environment |
| CN107016691A (en) * | 2017-04-14 | 2017-08-04 | 南京信息工程大学 | Moving target detecting method based on super-pixel feature |
| JP2017219377A (en) * | 2016-06-06 | 2017-12-14 | 三菱電機株式会社 | Monitoring device, monitoring method, and airport monitoring system |
-
2018
- 2018-03-26 CN CN201810256423.8A patent/CN108508425B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103745216A (en) * | 2014-01-02 | 2014-04-23 | 中国民航科学技术研究院 | Radar image clutter suppression method based on airspace feature |
| CN104199009A (en) * | 2014-09-18 | 2014-12-10 | 中国民航科学技术研究院 | Radar image clutter suppression method based on time-domain characteristics |
| CN105184814A (en) * | 2015-07-27 | 2015-12-23 | 成都天奥信息科技有限公司 | Moving target detecting and tracking method based on multi-frame radar image |
| JP2017219377A (en) * | 2016-06-06 | 2017-12-14 | 三菱電機株式会社 | Monitoring device, monitoring method, and airport monitoring system |
| CN106447680A (en) * | 2016-11-23 | 2017-02-22 | 湖南华诺星空电子技术有限公司 | Method for radar and vision fused target detecting and tracking in dynamic background environment |
| CN107016691A (en) * | 2017-04-14 | 2017-08-04 | 南京信息工程大学 | Moving target detecting method based on super-pixel feature |
Non-Patent Citations (1)
| Title |
|---|
| 尹宏鹏 等: "基于视觉的目标检测与跟踪综述", 《自动化学报》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113994404A (en) * | 2019-06-21 | 2022-01-28 | 松下电器产业株式会社 | Monitoring device and monitoring method |
| CN113994404B (en) * | 2019-06-21 | 2023-11-07 | 松下控股株式会社 | Monitoring device and monitoring method |
| CN115320669A (en) * | 2022-08-31 | 2022-11-11 | 南京慧尔视智能科技有限公司 | Method, device, equipment and medium for detecting railway coming car based on radar map |
| CN115424455A (en) * | 2022-08-31 | 2022-12-02 | 南京慧尔视智能科技有限公司 | Vehicle early warning method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108508425B (en) | 2020-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108230264B (en) | Single image defogging method based on ResNet neural network | |
| CN105427626B (en) | A kind of statistical method of traffic flow based on video analysis | |
| Sen-Ching et al. | Robust techniques for background subtraction in urban traffic video | |
| CN102494675B (en) | High-speed visual capturing method of moving target features | |
| CN107204006A (en) | A kind of static target detection method based on double background difference | |
| CN104268519B (en) | Image recognition terminal and its recognition methods based on pattern match | |
| CN102542289A (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
| CN104299229A (en) | Infrared weak and small target detection method based on time-space domain background suppression | |
| CN104616290A (en) | Target detection algorithm in combination of statistical matrix model and adaptive threshold | |
| CN108508425A (en) | Foreground target detection method based on neighborhood characteristics under a kind of radar near-earth ambient noise | |
| CN110569782A (en) | Target detection method based on deep learning | |
| CN107944354B (en) | Vehicle detection method based on deep learning | |
| CN111462027B (en) | Multi-focus image fusion method based on multi-scale gradient and matting | |
| CN105095869A (en) | Pedestrian detection method and apparatus | |
| CN102663362A (en) | Moving target detection method t based on gray features | |
| CN103106796A (en) | Vehicle detection method and device of intelligent traffic surveillance and control system | |
| CN108614998B (en) | Single-pixel infrared target detection method | |
| CN113763427A (en) | Multi-target tracking method based on coarse-fine shielding processing | |
| CN111915558A (en) | Pin state detection method for high-voltage transmission line | |
| CN112070717A (en) | Power transmission line icing thickness detection method based on image processing | |
| CN111027564A (en) | Low-illumination imaging license plate recognition method and device based on deep learning integration | |
| CN111027542A (en) | Target detection method improved based on fast RCNN algorithm | |
| Hu et al. | Big data oriented novel background subtraction algorithm for urban surveillance systems | |
| CN106780390A (en) | Single image to the fog method based on marginal classification Weighted Fusion | |
| CN107871315B (en) | Video image motion detection method and 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |