CN106919889A - The method and apparatus detected to the number of people in video image - Google Patents
The method and apparatus detected to the number of people in video image Download PDFInfo
- Publication number
- CN106919889A CN106919889A CN201510993978.7A CN201510993978A CN106919889A CN 106919889 A CN106919889 A CN 106919889A CN 201510993978 A CN201510993978 A CN 201510993978A CN 106919889 A CN106919889 A CN 106919889A
- Authority
- CN
- China
- Prior art keywords
- people
- detection
- image
- detector
- detectors
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
Landscapes
- Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
According to the invention it is proposed that a kind of method that number of people in video image is detected, including：The positive sample of various sizes of number of people image and the negative sample of correspondingly-sized are selected, by carrying out disaggregated model off-line training, multiple number of people detectors with various sizes of detection block is obtained；Altimetric image to be checked is extracted from video image；Select to carry out the number of people detector of number of people detection for treating detection image from multiple number of people detectors；And treat detection image using selected number of people detector and carry out number of people detection, and quantity to the detected number of people is counted.
Description
Technical field
The application is related to what a kind of number of people in video image was detected
Method and apparatus, can select the HOG of the detection window with appropriate size to examine
Device is surveyed to improve the precision that the number of people in picture is detected and counted.
Background technology
With the continuous progress of society, the range of application of video monitoring system is got over
Come wider.In places such as supermarket, market, gymnasium and airport stations
Gateway is often provided with CCTV camera, so as to security personnel and manager couple
The gateway in these places is monitored.On the other hand, supermarket, market,
The flow of the people of the place such as gymnasium and airport station turnover is for above-mentioned field
Operator or manager for have great significance, wherein, the stream of people
Amount refers to the number for flowing in certain direction, refer in particular to herein by entrance/from
Open the number of both direction flowing.
In existing video monitoring, people flow rate statistical is mainly by monitor
Member manually checks to realize.The method of this artificial statistics flow of the people is in monitoring
Time is short, flow of the people it is sparse in the case of it is reliable, but because people looks unfamiliar
The limitation of thing characteristic, when monitoring period is more long, when flow of the people is intensive, statistics
Accuracy will be greatly reduced, and the mode for manually counting needs to expend big
The human cost of amount.
People flow rate statistical method based on video analysis can realize flow of the people
Programming count, solution manually counts the various problems brought.At present, it is based on
The flow statistical method of video analysis mainly has three classes：
One is the method for distinguished point based tracking, and the method is tracked first
Motion characteristics point, then the track to characteristic point carry out cluster analysis, from
And obtain flow of the people information；The method of distinguished point based tracking needs tracking one
A little motion characteristics points, then the track to characteristic point carry out cluster analysis,
So as to obtain flow of the people information, the shortcoming of the method is that characteristic point is difficult in itself
Stably track, counting precision is poor.
Two be based on human body segmentation and tracking method, the method firstly the need of
Moving target block is extracted, then moving target block split obtaining list
Individual human body target, finally tracks the statistics that each human body target realizes flow of the people；
Method based on human body segmentation and tracking is firstly the need of moving target at extraction
Block, then to moving target block split and obtains single human body target, most
Tracking afterwards obtains the track of each human body, so as to realize the statistics of flow of the people.
The shortcoming of the method is the accuracy of human body segmentation when human body exists and blocks
It is difficult to be guaranteed, influences statistical accuracy.
Three is the method based on the number of people or head and shoulder detect and track, and the method exists
The number of people or head and shoulder are detected in video, is carried out by the tracking to the number of people or head and shoulder
The statistics of flow of the people.Method based on number of people number detect and track is in video
The middle detection number of people, the statistics of flow of the people is carried out by the tracking to the number of people, when
Camera angle is where appropriate, the number of people situation that occurs blocking is less, therefore base
Two methods accuracy increases the method detected in number of people number earlier above.
The number of people is being detected in the prior art, it is proposed that one kind is used
Gradient orientation histogram (HOG) feature of image carries out the side of number of people detection
Method, carrying out number of people detection using HOG features has extremely strong robustness.HOG
The gradient direction of mainly localized region is calculated, and then uses histogram
To describe, i.e. it is a kind of feature descriptor of image local overlapping region.
Compared to further feature description, such as color characteristic, quasi-Haar wavelet feature,
HOG has its unique advantage.First, the description of HOG features is in less space
The optical deformation and geometric deformation of image can be effectively overcome on domain, because HOG
It is the statistics with histogram to image local area unit.Secondly, in region weight
Under the conditions of folded calculating and the normalization of regional area etc., even if similar image exists
Shape can also occur trickle change, and these fine distinctions can't be serious
Influence Detection results.
When the number of people for example to passenger on public transport is detected, can be advance
From capture bus passenger flow vedio data in manually gather it is various not
With the image for only including passenger people's header of size as positive sample, collection
Correspondingly-sized not including people's header image as negative sample, by positive sample
Originally with negative sample composing training sample set, divided using the training sample set
Class model off-line training, is available for detecting the HOG detectors of the number of people.
The number of people of the passenger on public transport in image can be entered using the HOG detectors
Row detection.Finally, the number of people for detecting can be counted using SURF.
Here, SURF is a kind of feature that computer vision field is used for target following
Extracting mode, compared to other feature extraction modes, SURF have discrimination it is high,
The advantage of fast operation, has rotation and the characteristics of scale invariability.
In such people's head inspecting method, using various sizes of number of people figure
The positive sample of picture and the negative sample of correspondingly-sized are offline to carry out disaggregated model
Training, can obtain the HOG detectors with various sizes of detection window.
When detection has the picture of less number, if using with large-size
Detection window HOG detectors, due to the detection window with large-size
HOG detectors include more information, therefore can more accurately detect
To the number of people.But, the picture more for number, because the number of people is due to meeting
Suffer relatively near and clash, therefore use the detection window with large-size
HOG detectors error detection can occur, it is therefore desirable to using have smaller chi
The HOG detectors of very little detection window are above-mentioned to prevent to be detected to the number of people
The generation of error detection.On the other hand, if adopted for picture fewer in number
The number of people is examined with the HOG detectors of the detection window with reduced size
Survey, then by the information that the HOG detectors are included is minimum, therefore relatively
In the HOG detectors using the detection window with large-size, accuracy of detection
Can be at a fairly low.It is, therefore, desirable to provide one kind can be selected with appropriate size
The HOG detectors of detection window the number of people in picture is detected to improve
With the precision for counting.
The content of the invention
In order to the drawbacks described above for overcoming prior art proposes the present invention.Therefore,
An object of the present invention is to provide a kind of number of people in video image and enters
The method and number of people number detection means of row detection, can select with appropriate chi
The HOG detectors of very little detection window are examined to improve to the number of people in picture
The precision surveyed and count.
To achieve these goals, according to the invention it is proposed that one kind is to regarding
The method that the number of people in frequency image is detected, including：Selection different size
Number of people image positive sample and the negative sample of correspondingly-sized, by being divided
Class model off-line training, obtains multiple people with various sizes of detection block
Head detector；Altimetric image to be checked is extracted from video image；From described in multiple
Select to carry out number of people detection for treating detection image in number of people detector
Number of people detector；And using selected number of people detector to figure to be detected
As carrying out number of people detection, and quantity to the detected number of people is counted.
Preferably, the quantity according to the number of people counted out using rough detection is accounted for
The ratio of the at most detectable number of the altimetric image to be checked, has from multiple
Selection has specific in the number of people detector of various sizes of detection block
The number of people detector of the detection block of size.
Preferably, the rough detection is utilized from multiple number of people detectors
In an optional number of people detector carry out.
Preferably, become equivalent to previous frame image according to altimetric image to be checked
The number of pixels of change accounts for the ratio of all pixels number of altimetric image to be checked, from many
Tool is selected in the individual number of people detector with various sizes of detection block
There is the number of people detector of the detection block of specific dimensions.
In addition, it is also proposed that a kind of number of people in video image is detected
Device, including：Select the positive sample of various sizes of number of people image and right
The negative sample of size is answered, by carrying out disaggregated model off-line training, obtains many
The unit of the individual number of people detector with various sizes of detection block；From video
The unit of altimetric image to be checked is extracted in image；From multiple number of people detectors
It is middle to select to carry out the number of people detector of number of people detection for treating detection image
Unit；And treat detection image using selected number of people detector and enter
Pedestrian's head detection, and the list counted to the quantity of the detected number of people
Unit.
The method according to the invention and device, can select with appropriate size
The HOG detectors of detection window the number of people in picture is detected to improve
With the precision for counting.
Brief description of the drawings
Fig. 1 show it is of the invention HOG detectors are trained and
The schematic diagram of generation.
Fig. 2 is showed and adopted when being detected to the picture with different numbers
With the schematic diagram of the HOG detectors of the detection window with appropriate size.
Fig. 3 shows the number of people in video image of the invention and enters
The flow chart of the method for row detection.
Specific embodiment
The preferred embodiments of the present invention are described below with reference to the accompanying drawings.In accompanying drawing
In, identical element will be represented by identical reference symbol or numeral.Additionally,
In description below of the invention, the tool to known function and configuration will be omitted
Body is described, to avoid making subject of the present invention unclear.
Fig. 1 show it is of the invention HOG detectors are trained and
The schematic diagram of generation.
As shown in figure 1, gathering many by artificial from the vedio data of capture
The various sizes of image for only including people's header is planted as positive sample, collection
Correspondingly-sized not including people's header image as negative sample, by positive sample
Originally with negative sample composing training sample set, divided using the training sample set
Class model off-line training, is available for detecting the HOG detectors of the number of people.
The number of people of personnel present in image can be carried out using the HOG detectors
Detection.Certainly, the invention is not limited in HOG detectors, and can be
Any types that the number of people can be detected obtained by off-line training
Detector.
Fig. 2 is showed and adopted when being detected to the picture with different numbers
With the schematic diagram of the HOG detectors of the detection window with appropriate size.
As has been mentioned previously, in common people's head inspecting method, profit
With the positive sample and the negative sample of correspondingly-sized of various sizes of number of people image
To carry out disaggregated model off-line training, can obtain with various sizes of inspection
Survey the HOG detectors of window.When detection has the picture of less number, such as
Fruit uses the HOG detectors of the detection window with large-size, due to having
The HOG detectors of the detection window of large-size include more information, therefore
The number of people can be more accurately detected.But, the figure more for number
Piece, because the number of people is clashed due to that can suffer relatively near, therefore uses tool
The HOG detectors for having the detection window of large-size can occur error detection, therefore
The HOG detectors of the detection window with reduced size are needed to use to come to the number of people
Detected to prevent the generation of above-mentioned error detection.On the other hand, if right
Examined using the HOG of the detection window with reduced size in picture fewer in number
Survey device to detect the number of people, then included by the HOG detectors
Information is minimum, therefore the detection window with large-size relative to use
HOG detectors, accuracy of detection can be at a fairly low.
As shown in Fig. 2 as an example, to picture A fewer in number, using
The number of people detector of the detection window with large-size carries out inspection time difference method
Can be preferable；The picture A medium to number, using the inspection with medium size
Surveying the number of people detector of window carries out the meeting of inspection time difference method preferably；And to number compared with
Many picture A, using the number of people detector of the detection window with reduced size
Carrying out inspection time difference method can be preferable.
Fig. 3 shows the number of people in video image of the invention and enters
The flow chart of the method for row detection.
In step 301, using the positive sample of various sizes of number of people image and right
Answer the negative sample of size to carry out HOG off-line trainings, obtain with different chis
The HOG detectors of very little detection window.
Here, for the purpose of simplifying the description, it is assumed that obtained have M kinds (M is big
In 1 natural number) the HOG detectors of various sizes of detection window.Every kind of chi
The width w of very little detection window is identical, and height is different, is set to h1、h2、
hi、hM.Also, it can be assumed that h1＞ h2＞ ... ＞ hMAnd hi-hi+1=definite value.
It is, to detecting that the size of window is sorted from big to small.
Then, in step 303, the Video Image Capturing shot from camera is treated
Detection image.
In step 305, from the above-mentioned HOG inspections with various sizes of detection block
Device is surveyed to select to carry out the appropriate HOG of number of people detection for treating detection image
Detector, the specifically chosen method of appropriate HOG detectors will be retouched later
State.
Finally, in step 307, using selected HOG detectors to be checked
The number of people in altimetric image detected, and to the quantity of the detected number of people
Counted.
Will be described below carrying out HOG detectors the two of selection in step 305
Individual example.
Example 1
Selection such as detection block height hminHOG detectors Dmin, to be detected
The number of people in image carries out rough detection, and the number of people number measured to Rough Inspection is carried out
Count, it is assumed that by rough detection and count after number of people number be n.Here,
hminIt is h1、h2、hi、hMIn minimum constructive height, according to above-mentioned it is assumed that hmin
As hM.However, it is desirable to, it is noted that the invention is not restricted to using detection
Frame height degree hminHOG detectors DminTo carry out rough detection, and can be use
With any detection block height hiHOG detectors DiTo carry out rough detection, this
Can be specified by operator.
Next, the size of the number n obtained according to rough detection is selected for right
The detailed process that altimetric image to be checked carries out the HOG detectors of number of people detection is as follows：
■ is according to HOG detectors DminCorresponding sample size w*hmin, calculate
Most detectable personnel's number N in altimetric image A to be checked：
N=[WA*HA/w*hmin]
Wherein, WARepresent the height of altimetric image A to be checked, HARepresent mapping to be checked
As the width of A, [] represents bracket function, hminRepresent minimum detection frame height
Degree.
Then ■, counts the number of people step-length according to below equation：
Nstep=N/M；
■ next, select to number according to below equation be J HOG detection
Device, i.e. selection HOG detectors DJ。
J=[n/Nstep],
Wherein, [] represents bracket function.
In example 1, according to the number of people for being detected using rough detection and count out
The quantity at most detectable number that accounts for altimetric image to be checked ratios come from by
Selected in detector after being sorted according to detection window size mode from big to small
Detector at corresponding appropriate ranking come carry out accurate number of people detection and
Count.
Example 2
Calculate what altimetric image A to be checked changed relative to previous frame image
Number of pixels n；
Then, the size according to n selects to carry out the number of people for treating detection image
The detailed process of the HOG detectors of detection is as follows：
■ calculates all pixels number in altimetric image A to be checked：
N '=WA*HA
Pixel step length is calculated according to below equation：
Nstep'=N '/M；
Then, select to number according to below equation be J HOG detectors,
That is, selection HOG detectors DJ。
J=[n/Nstep],
Wherein, [] represents bracket function.
In example 2, according to utilization altimetric image to be checked relative to previous frame image
The number of pixels for changing accounts for the ratio of all pixels number of altimetric image to be checked
Example is come the detection after being sorted from the mode according to detection window size from big to small
The detector at corresponding appropriate ranking is selected to carry out accurate people in device
Head detection and counting.
Some specific embodiments are enumerated above to elaborate the present invention, this
A few examples are merely to illustrate principle of the invention and its implementation, rather than right
Limitation of the invention, without departing from the spirit and scope of the present invention,
Those skilled in the art can also make various modifications and improvement.Therefore,
The present invention should not be limited by above-described embodiment, and should be by appended claims
And its equivalent is limited.
Claims (5)
1. a kind of method that number of people in video image is detected, including：
Select the positive sample of various sizes of number of people image and bearing for correspondingly-sized
Sample, by carrying out disaggregated model off-line training, obtains multiple with difference
The number of people detector of the detection block of size；
Altimetric image to be checked is extracted from video image；
Select to enter for treating detection image from multiple number of people detectors
The number of people detector of pedestrian's head detection；And
Treating detection image using selected number of people detector carries out number of people inspection
Survey, and quantity to the detected number of people is counted.
2. method according to claim 1, wherein,
Quantity according to the number of people counted out using rough detection accounts for described to be checked
The ratio of the at most detectable number of altimetric image, has different size from multiple
Detection block the number of people detector in selection with specific dimensions detection
The number of people detector of frame.
3. method according to claim 1, wherein,
The rough detection is to utilize arbitrarily to be selected from multiple number of people detectors
A number of people detector selecting is carried out.
4. method according to claim 1, wherein,
According to altimetric image to be checked equivalent to the pixel that previous frame image changes
Number accounts for the ratio of all pixels number of altimetric image to be checked, has not from multiple
There are specific dimensions with selection in the number of people detector of the detection block of size
Detection block number of people detector.
5. the device that a kind of number of people in video image is detected, including：
Select the positive sample of various sizes of number of people image and bearing for correspondingly-sized
Sample, by carrying out disaggregated model off-line training, obtains multiple with difference
The unit of the number of people detector of the detection block of size；
The unit of altimetric image to be checked is extracted from video image；
Select to enter for treating detection image from multiple number of people detectors
The unit of the number of people detector of pedestrian's head detection；And
Treating detection image using selected number of people detector carries out number of people inspection
Survey, and the unit counted to the quantity of the detected number of people.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510993978.7A CN106919889B (en) | 2015-12-25 | 2015-12-25 | Method and device for detecting human head in video image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510993978.7A CN106919889B (en) | 2015-12-25 | 2015-12-25 | Method and device for detecting human head in video image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106919889A true CN106919889A (en) | 2017-07-04 |
| CN106919889B CN106919889B (en) | 2021-04-13 |
Family
ID=59455546
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510993978.7A Active CN106919889B (en) | 2015-12-25 | 2015-12-25 | Method and device for detecting human head in video image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106919889B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109858308A (en) * | 2017-11-30 | 2019-06-07 | 株式会社日立制作所 | Video frequency searching device, video retrieval method and storage medium |
| CN110717891A (en) * | 2019-09-17 | 2020-01-21 | 平安科技（深圳）有限公司 | Picture detection method and device based on grouping batch and storage medium |
| CN113553586A (en) * | 2021-06-16 | 2021-10-26 | 深信服科技股份有限公司 | Virus detection method, model training method, device, equipment and storage medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030072470A1 (en) * | 2001-10-15 | 2003-04-17 | Lee Henry C. | Two dimensional autonomous isotropic detection technique |
| CN101477641A (en) * | 2009-01-07 | 2009-07-08 | 北京中星微电子有限公司 | Demographic method and system based on video monitoring |
| CN101872422A (en) * | 2010-02-10 | 2010-10-27 | 杭州海康威视软件有限公司 | People flow rate statistical method and system capable of precisely identifying targets |
| CN102542286A (en) * | 2010-10-12 | 2012-07-04 | 索尼公司 | Learning device, learning method, identification device, identification method, and program |
| CN103136534A (en) * | 2011-11-29 | 2013-06-05 | 汉王科技股份有限公司 | Method and device of self-adapting regional pedestrian counting |
| CN103729649A (en) * | 2014-01-14 | 2014-04-16 | 三星电子（中国）研发中心 | Image rotating angle detection method and device |
| CN104008404A (en) * | 2014-06-16 | 2014-08-27 | 武汉大学 | Pedestrian detection method and system based on significant histogram features |
| CN104281843A (en) * | 2014-10-20 | 2015-01-14 | 上海电机学院 | Image recognition method and system based on self-adaptive features and classification model selection |
| CN104299005A (en) * | 2013-07-15 | 2015-01-21 | 深圳市振邦实业有限公司 | Head detection method and system |
| CN105160292A (en) * | 2015-07-07 | 2015-12-16 | 深圳市捷顺科技实业股份有限公司 | Vehicle identification recognition method and system |
-
2015
- 2015-12-25 CN CN201510993978.7A patent/CN106919889B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030072470A1 (en) * | 2001-10-15 | 2003-04-17 | Lee Henry C. | Two dimensional autonomous isotropic detection technique |
| CN101477641A (en) * | 2009-01-07 | 2009-07-08 | 北京中星微电子有限公司 | Demographic method and system based on video monitoring |
| CN101872422A (en) * | 2010-02-10 | 2010-10-27 | 杭州海康威视软件有限公司 | People flow rate statistical method and system capable of precisely identifying targets |
| CN102542286A (en) * | 2010-10-12 | 2012-07-04 | 索尼公司 | Learning device, learning method, identification device, identification method, and program |
| CN103136534A (en) * | 2011-11-29 | 2013-06-05 | 汉王科技股份有限公司 | Method and device of self-adapting regional pedestrian counting |
| CN104299005A (en) * | 2013-07-15 | 2015-01-21 | 深圳市振邦实业有限公司 | Head detection method and system |
| CN103729649A (en) * | 2014-01-14 | 2014-04-16 | 三星电子（中国）研发中心 | Image rotating angle detection method and device |
| CN104008404A (en) * | 2014-06-16 | 2014-08-27 | 武汉大学 | Pedestrian detection method and system based on significant histogram features |
| CN104281843A (en) * | 2014-10-20 | 2015-01-14 | 上海电机学院 | Image recognition method and system based on self-adaptive features and classification model selection |
| CN105160292A (en) * | 2015-07-07 | 2015-12-16 | 深圳市捷顺科技实业股份有限公司 | Vehicle identification recognition method and system |
Non-Patent Citations (2)
| Title |
|---|
| MI YOUNG NAM AND PHILL KYU: "A Face Detection Using Multiple Detectors for External Environment", 《FUZZY SYSTEM AND KNOWLEDGE DISCOVERY》 * |
| 唐大伟等: "壁画图像分类中的分组多实例学习方法", 《中国图象图形学报》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109858308A (en) * | 2017-11-30 | 2019-06-07 | 株式会社日立制作所 | Video frequency searching device, video retrieval method and storage medium |
| CN110717891A (en) * | 2019-09-17 | 2020-01-21 | 平安科技（深圳）有限公司 | Picture detection method and device based on grouping batch and storage medium |
| WO2021051580A1 (en) * | 2019-09-17 | 2021-03-25 | 平安科技（深圳）有限公司 | Grouping batch-based picture detection method and apparatus, and storage medium |
| CN113553586A (en) * | 2021-06-16 | 2021-10-26 | 深信服科技股份有限公司 | Virus detection method, model training method, device, equipment and storage medium |
| CN113553586B (en) * | 2021-06-16 | 2024-05-28 | 深信服科技股份有限公司 | Virus detection method, model training method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106919889B (en) | 2021-04-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103164706B (en) | Object counting method and device based on video signal analysis | |
| US9002068B2 (en) | Video speed detection system | |
| CN102542289B (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
| CN107741231B (en) | Multi-moving-target rapid ranging method based on machine vision | |
| US10515277B2 (en) | Method for analyzing the distribution of objects in free queues | |
| CN201255897Y (en) | Human flow monitoring device for bus | |
| CN107845264A (en) | A kind of volume of traffic acquisition system and method based on video monitoring | |
| CN110555397A (en) | crowd situation analysis method | |
| CN105608431A (en) | Vehicle number and traffic flow speed based highway congestion detection method | |
| CN105303191A (en) | Method and apparatus for counting pedestrians in foresight monitoring scene | |
| CN106845325B (en) | A kind of information detecting method and device | |
| CN105551264A (en) | Speed detection method based on license plate characteristic matching | |
| CN108229352B (en) | Standing detection method based on deep learning | |
| CN103413149B (en) | Method for detecting and identifying static target in complicated background | |
| WO2009039350A1 (en) | System and method for estimating characteristics of persons or things | |
| CN106919889A (en) | The method and apparatus detected to the number of people in video image | |
| CN109711313B (en) | Real-time video monitoring method for identifying sewage poured into river channel | |
| CN103198323B (en) | A kind of pedestrian detection method based on curve matching | |
| CN112991769A (en) | Traffic volume investigation method and device based on video | |
| Chen et al. | Crowd abnormal behaviour identification based on integral optical flow in video surveillance systems | |
| EP3244344A1 (en) | Ground object tracking system | |
| Gao | Vehicle detection in wide-area aerial imagery: cross-association of detection schemes with post-processings | |
| CN109492616A (en) | A kind of advertisement screen face identification method based on autonomous learning | |
| CN110021174A (en) | A kind of vehicle flowrate calculation method for being applicable in more scenes based on video image | |
| CN111898434B (en) | Video detection and analysis system |
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 |