CN103258232A - Method for estimating number of people in public place based on two cameras - Google Patents
Method for estimating number of people in public place based on two cameras Download PDFInfo
- Publication number
- CN103258232A CN103258232A CN2013101258656A CN201310125865A CN103258232A CN 103258232 A CN103258232 A CN 103258232A CN 2013101258656 A CN2013101258656 A CN 2013101258656A CN 201310125865 A CN201310125865 A CN 201310125865A CN 103258232 A CN103258232 A CN 103258232A
- Authority
- CN
- China
- Prior art keywords
- stage
- image
- camera
- foreground image
- pixel count
- 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
Images
Abstract
The invention discloses a method for estimating the number of people in a public place based on two cameras. According to the method, at first a multivariate linear model which is used for estimating is set up, a training sample set of the multivariate linear model is acquired in a scene, and then model parameters are acquired through the analysis of multiple linear regression, so that a complete multiple linear regression equation is acquired. Only corresponding characters are extracted from a surveillance video image to be substituted into the set multiple linear regression equation so that people flow density estimation can be carried out in real time. According to the method for estimating the number of people in the public place, the estimation of the number of people can be carried out according to a multivariate linear relation of prospect pixels, face area pixels, prospect outer/inner edge pixels and people flow density, the method for estimating the number of people in the public place based on the two cameras is suitable for a low-density crowd, and estimation accuracy under the condition of a high-density crowd is improved.
Description
Technical field
The present invention relates to relate to a kind of people from public place number estimation method of dual camera, can be widely used in places such as subway, station, carry out stream of people's measuring density and estimation.
Background technology
Along with the raising of rapid economy development and people's living standard, increasing population pours in the city.Such as market, subway etc., all can welcome the stream of people's peak for the many communal facilitys in the city in the short time.The injures and deaths event of trampling that takes place because the stream of people is too crowded also all took place in various places in recent years, therefore the intensity of passenger flow in these places was carried out real-time statistics and analysis and seemed most important.The crowd density estimation technique has vital role in fields such as the management and control of transport hubs such as public safety, station, airport and services in addition, also can be business decision foundation is provided.
Along with development of computer, utilize computer vision and image processing techniques to monitoring picture real-time analysis processing, can realize the stream of people statistics and management automatically, method commonly used mainly contains three major types: based on the method for pixel analysis, based on the method for texture analysis and the method for analyzing based on individual goal.Wherein, foundation based on the method for pixel analysis is by approximate linear between crowd density and the pixel count, can directly obtain the estimated result of crowd density, but can not solve high density crowd's overlap problem, because crowd density is more big, overlapping more serious, crowd density at this moment and pixel count have not been linear relationship before.Be to have corresponding relation between the thickness of the height of crowd density and texture pattern based on the foundation of the method for texture analysis, can solve the overlapping problem of blocking of high density crowd well, but very big error is but arranged when low-density is monitored.Method based on the individual goal analysis mainly is that target is identified, followed the tracks of, and to distinguish different target individual, still when people's current density was excessive, target was cut apart very difficult, and the method is very complicated, was difficult to requirement of real time.
Summary of the invention
In order to address the above problem, the object of the present invention is to provide a kind of public place crowd estimate's method based on dual camera.
In order to achieve the above object, this method of public place crowd estimate's method based on dual camera provided by the invention adopts the hardware platform that is made of first camera, second camera and main frame, wherein: first camera is connected with main frame respectively with second camera, and first camera and second camera are installed in the two ends of collection site respectively with symmetrical manner; Main frame is arithmetic unit, adopts the universal PC computing machine, is used for gathering image from the two ends at scene respectively by first camera and second camera, and finishes the corresponding computing to crowd estimate in on-the-spot; Described public place crowd estimate's method is made up of training stage and real time phase; The wherein said training stage comprises the following step of carrying out in order:
Whether step 4, judging characteristic sample sequence complete S104 stage: judge whether the feature samples sequence that is used for making up training sample set is complete, if judged result is "Yes" then enters next step, otherwise change the porch that skips to the S101 stage, next step continues to carry out the S101 stage;
Step 5, the S105 stage of setting up equation of linear regression: according to the training sample set that has made up, try to achieve the coefficient value of equation of linear regression by least square method, thereby set up equation of linear regression, so far the training stage flow process finishes;
Described real time phase comprises the following step of carrying out in order:
The S202 stage of step 2, the effective monitored area of setting: in the monitoring picture of two cameras, determine same effective monitored area respectively.
The S205 stage of step 5, edge calculation pixel: the external margin pixel count and the internal edge pixel count that calculate foreground image;
The S206 stage of step 6, comprehensive characteristics information and estimated number: the pixel count of above-mentioned these characteristic elements is carried out overall treatment and be updated in the equation of linear regression that generates in the training stage as characteristic, namely try to achieve crowd estimate's value of movable crowd.
2, the public place crowd estimate's method based on dual camera according to claim 1 is characterized in that: in the stage, described extraction characteristics of image process comprises that mainly foreground image extracts, human face region detects and rim detection at S103; Wherein the foreground image extracting method is at first to utilize gauss hybrid models to set up background model, judge then whether the image slices vegetarian refreshments mates with K Gaussian distribution of current background, if coupling then is judged as background, upgrade the Gaussian distribution weights simultaneously, if all do not match, then be judged as prospect, the Gaussian distribution weights do not upgrade; Generate the foreground mask of a two-value, wherein background is 0, and prospect is 255.Then mask is carried out medium filtering and morphology processing, so just finished the extraction to foreground image; The human face region detection method is based on the Haar sorter and detects human face region, and then counts colour of skin district pixel count in the human face region by the YCrCb color space; Edge detection method is at first the foreground image that extracts to be converted to gray level image, uses the canny operator to carry out rim detection after mean filter.
In stage, the background subtraction separating method of mixed Gauss model is mainly adopted in the extraction of described foreground image at S203.
In stage, the method for described detection face complexion area is at first to adopt based on the method for Haar sorter to determine human face region at S204, then by adding up skin pixel in the human face region based on the skin color detection method of YCrCb color space.
At S205 in the stage, the edge pixel method of described calculating foreground image is to adopt the canny operator that foreground image is carried out rim detection, and the result is carried out noise reduction, filtering handle, remove and disturb bigger zone, count internal edge pixel count and external margin pixel count then.
Public place crowd estimate's method based on dual camera provided by the invention is that to close with the multiple linear of foreground pixel, human face region pixel and the outer/inner edge pixel of prospect and people's current density be to estimate according to carrying out number, be not only applicable to the low-density crowd, improved the estimated accuracy under high density crowd's situation simultaneously.
Description of drawings
Fig. 1 is that the public place crowd estimate's method hardware platform based on dual camera provided by the invention constitutes synoptic diagram.
Fig. 2 is the composition frame chart of the public place crowd estimate's method hardware platform based on dual camera provided by the invention.
Fig. 3 is provided by the invention based on training stage process flow diagram in public place crowd estimate's method of dual camera.
Fig. 4 is provided by the invention based on real time phase process flow diagram in public place crowd estimate's method of dual camera.
Embodiment
Below in conjunction with the drawings and specific embodiments the public place crowd estimate's method based on dual camera provided by the invention is elaborated.
The hardware platform that Fig. 1, Fig. 2 show the public place crowd estimate's method based on dual camera provided by the invention constitutes, wherein: first camera 1 and second camera 2 are image collecting device, it is connected with main frame 3 respectively, be used for the image of collection site, first camera 1 and second camera 2 are installed in the two ends of collection site respectively with symmetrical manner; Main frame 3 is arithmetic unit, adopts the universal PC computing machine, and crowd 4 is the stream of people passing in the monitored scene; Main frame 3 is connected with second camera 2 with first camera 1 respectively, is used for gathering image from the two ends at scene respectively by first camera 1 and second camera 2, and finishes the corresponding computing to crowd estimate in on-the-spot.
The people's number estimation method that is applicable to the public place provided by the invention is not only applicable to the low-density crowd, has improved the estimated accuracy under high density crowd's situation simultaneously.This method is at two monitorings of the symmetrical in opposite directions installation in the two ends of monitored area cameras, and two cameras are monitored the same area simultaneously from both direction, and model of place as shown in Figure 1.
Described public place crowd estimate's method based on dual camera is on the basis that foreground pixel is analyzed, and takes all factors into consideration skin pixel and edge pixel, and main thought is based on following 2 points:
(1) in the scene, along with the increase of crowd's 4 density, the proportion that accounts for whole foreground area based on the area of skin color of people's face is increase tendency;
(2) foreground image is carried out edge extracting, along with the increase of crowd density, the proportion that the internal edge pixel except the prospect profile edge accounts for whole edge pixels is increase tendency.
This method is utilized the stronger multiple linear relation of existence between crowd's number and the above-mentioned two kinds of ratios. can try to achieve multivariate linear model between number and these features by multiple linear regression analysis, thereby can realize more accurate number estimation for revising based on the estimated result of foreground pixel counting method.
Utilize the public place crowd estimate's method based on dual camera provided by the invention to estimate for the number of certain class scene, need at first to set up the multivariate linear model that is used for estimation according to above-mentioned thought, and in such scene, obtain the training sample set of multivariate linear model, and then ask for model parameter by multiple linear regression analysis, thereby obtain complete multiple linear regression equations; Only need just can carry out the number estimation in real time in the multiple linear regression equations that the corresponding feature substitution of extraction is built up from the monitoring video image afterwards.
Therefore, public place crowd estimate's method based on dual camera provided by the invention is made up of training stage and real time phase, in the training stage: main frame 3 is gathered a width of cloth live video image respectively as sample image by first camera 1 and second camera 2, and therefrom extract the sport people foreground image, calculate the foreground image pixel count then, the pixel count that face complexion area is shared, the pixel count that foreground area inside/outside edge is shared and concrete number of the crowd among the image pattern 4, above-mentioned characteristic element is formed a feature samples, repeated acquisition multiframe sample image, obtain the feature samples of sufficient amount, just can obtain a complete training sample set, the training sample set that utilization obtains just can be asked for the parameter of multivariate linear model by multiple linear regression analysis, thereby obtains the multiple linear regression equations for real-time estimation; At real time phase: main frame 2 is gathered a two field picture respectively in real time by first camera 1 and second camera 2, and therefrom calculate foreground image pixel count, shared pixel count, the shared features such as pixel in foreground area inside/outside edge of face complexion area, with among the above-mentioned feature substitution multiple linear regression equations, just can obtain the crowd estimate's value that needs then.
As shown in Figure 3, the described training stage comprises the following step of carrying out in order:
Whether step 4, judging characteristic sample sequence complete S104 stage: judge whether the feature samples sequence that is used for making up training sample set is complete, if judged result is "Yes" then enters next step, otherwise change the porch that skips to the S101 stage, next step continues to carry out the S101 stage;
Step 5, the S105 stage of setting up equation of linear regression: try to achieve the coefficient of equation of linear regression according to the training sample set of obtaining, that is:
If P is the number in the monitoring picture, q is the foreground pixel area, and φ is a weighting coefficient, X
1, X
2Be respectively the proportion that accounts for whole prospect in the surveillance map picture of two cameras based on the area of skin color of people's face, Y
1, Y
2Be respectively internal edge pixel in the surveillance map picture of two cameras and account for the proportion of whole edge pixels, a, b, c, d are regression coefficient.
Make X=X
1+ X
2, Y=Y
1+ Y
2, φ=bX+cY+d, P=aq φ
According to the training sample set that has made up, try to achieve the coefficient value of equation of linear regression by least square method, thereby set up equation of linear regression, so far the training stage flow process finishes.
In the described process of setting up sample sequence, follow following principle: capturing sample image will be selected two field picture that the stream of people is distributed in the monitored area diverse location as sample image as far as possible, should suitably increase the sample image number for people's current density place big or circumstance complication.
In stage, described extraction characteristics of image process mainly comprises calculating processes such as foreground image extraction, human face region detection and rim detection at S103;
When foreground image extracts, at first utilize gauss hybrid models to set up background model, judge then whether the image slices vegetarian refreshments mates with K Gaussian distribution of current background, if coupling then is judged as background, upgrade the Gaussian distribution weights simultaneously, if all do not match, then be judged as prospect, the Gaussian distribution weights do not upgrade.Generate the foreground mask of a two-value, wherein background is 0, and prospect is 255.Then mask is carried out medium filtering and morphology processing, so just finished the extraction to foreground image; Consider that in addition the monitoring video sequence has 25 frame pictures one second, and destination object is the stream of people, the difference of adjacent two frames is very little, and number can not have significant change yet, so the method for the invention is handled to get a frame every 10 frames;
When human face region detects, mainly adopt the method based on the Haar sorter to detect human face region, and then count colour of skin district pixel count in the human face region by the YCrCb color space.At first from the monitoring video recording, choose a large amount of images, use the training of AdaBoost algorithm to distinguish people's face and non-face strong classifier, use the screening type cascade that strong classifier is cascaded to together then, after training is finished it is preserved, carrying out calling position and the approximate size that cascade classifier just can effectively detect human face region when people's face district is detected.Then the human face region image transitions is arrived the YCrCb space, judge that the pixel region that is in 133＜Cr＜173,77＜Cb＜127 is colour of skin district, then just can count the pixel count in colour of skin district in the human face region;
When rim detection, mainly adopt the canny operator, at first the foreground image that extracts is converted to gray level image, after mean filter, use the canny operator to carry out rim detection.The result of rim detection is bianry image, and the pixel of non-zero is the edge, counts the total edge pixel count.Position with each edge pixel point corresponds on the foreground mask then, and the value of 8 positions adjacent as if this position is nonzero value, then this edge pixel is judged as the inward flange pixel, otherwise is not.
As shown in Figure 4, described real time phase comprises the following step of carrying out in order:
The S202 stage of step 2, the effective monitored area of setting: in the monitoring picture of two cameras, determine same effective monitored area respectively.
The S205 stage of step 5, edge calculation pixel: the external margin pixel count and the internal edge pixel count that calculate foreground image;
The S206 stage of step 6, comprehensive characteristics information and estimated number: the pixel count of above-mentioned these characteristic elements is carried out overall treatment and be updated in the equation of linear regression that generates in the training stage as characteristic, just can be worth in the hope of movable crowd's crowd estimate.
In stage, the background subtraction separating method of mixed Gauss model is mainly adopted in the extraction of foreground image at described S203.Before adding up element of interest, need carry out processing such as filtering, denoising to foreground image.
In stage, detect face complexion area at described S204, at first adopt based on the method for Haar sorter and determine human face region, then by adding up skin pixel in the human face region based on the skin color detection method of YCrCb color space.
At described S205 in the stage, when calculating the edge pixel of foreground image, adopt the canny operator that foreground image is carried out rim detection, and the result is carried out noise reduction, filtering processing, remove and disturb bigger zone, count internal edge pixel count and external margin pixel count then.
Claims (5)
1. public place crowd estimate's method based on dual camera, this method adopts the hardware platform that is made of first camera (1), second camera (2) and main frame (3), wherein: first camera (1) is connected with main frame (3) respectively with second camera (2), and first camera (1) and second camera (2) are installed in the two ends of collection site respectively with symmetrical manner; Main frame (3) is arithmetic unit, adopts the universal PC computing machine, is used for gathering image from the two ends at scene respectively by first camera (1) and second camera (2), and finishes the corresponding computing to crowd estimate in on-the-spot; It is characterized in that: described public place crowd estimate's method is made up of training stage and real time phase; The wherein said training stage comprises the following step of carrying out in order:
Step 1, read in S101 stage of video image: gather image synchronously by first camera (1) and second camera (2) respectively, must guarantee that two pending frame sample images are the surveillance map picture of different cameras synchronization;
Step 2, judge the S102 stage whether image meets the demands: by main frame (3) according to people's flow path direction, crowd's distribution occasion in the image that reads in, judge whether the image that reads in satisfies the principle of construction feature sample, if judged result is "Yes" then enters next step, otherwise change the porch that skips to the S101 stage, next step continues to carry out the S101 stage;
Step 3, the S103 stage of setting up the feature samples sequence: the number in the record sample image, and extraction characteristics of image, that is: foreground image pixel count, face complexion area pixel count, foreground area outward flange pixel count and inward flange pixel count, above-mentioned characteristic element is kept in the feature samples sequence as a sample in the feature samples sequence, enters next step then;
Whether step 4, judging characteristic sample sequence complete S104 stage: judge whether the feature samples sequence that is used for making up training sample set is complete, if judged result is "Yes" then enters next step, otherwise change the porch that skips to the S101 stage, next step continues to carry out the S101 stage;
Step 5, the S105 stage of setting up equation of linear regression: according to the training sample set that has made up, try to achieve the coefficient value of equation of linear regression by least square method, thereby set up equation of linear regression, so far the training stage flow process finishes;
Described real time phase comprises the following step of carrying out in order:
Step 1, read in S201 stage of the surveillance map picture of two cameras synchronously: gather realtime graphic synchronously by first camera (1) and second camera (2) respectively, guarantee that handled two the monitoring pictures of main frame (3) are synchronizations;
The S202 stage of step 2, the effective monitored area of setting: in the monitoring picture of two cameras, determine same effective monitored area respectively;
Step 3, the S203 stage of extracting foreground image: in the monitoring picture of two cameras, be partitioned into movable crowd as foreground image respectively;
Step 4, pretreated S204 stage of foreground image: foreground image is carried out filtering, denoising, calculate the foreground image pixel count then, from foreground image, detect face complexion area again, testing result is done denoising, analyze and count the pixel count of face complexion area then;
The S205 stage of step 5, edge calculation pixel: the external margin pixel count and the internal edge pixel count that calculate foreground image;
The S206 stage of step 6, comprehensive characteristics information and estimated number: the pixel count of above-mentioned these characteristic elements is carried out overall treatment and be updated in the equation of linear regression that generates in the training stage as characteristic, namely try to achieve crowd estimate's value of movable crowd.
2. the public place crowd estimate's method based on dual camera according to claim 1 is characterized in that: in the stage, described extraction characteristics of image process comprises that mainly foreground image extracts, human face region detects and rim detection at S103; Wherein the foreground image extracting method is at first to utilize gauss hybrid models to set up background model, judge then whether the image slices vegetarian refreshments mates with K Gaussian distribution of current background, if coupling then is judged as background, upgrade the Gaussian distribution weights simultaneously, if all do not match, then be judged as prospect, the Gaussian distribution weights do not upgrade; Generate the foreground mask of a two-value, wherein background is 0, and prospect is 255.Then mask is carried out medium filtering and morphology processing, so just finished the extraction to foreground image; The human face region detection method is based on the Haar sorter and detects human face region, and then counts colour of skin district pixel count in the human face region by the YCrCb color space; Edge detection method is at first the foreground image that extracts to be converted to gray level image, uses the canny operator to carry out rim detection after mean filter.
3. the public place crowd estimate's method based on dual camera according to claim 1, it is characterized in that: in the stage, the background subtraction separating method of mixed Gauss model is mainly adopted in the extraction of described foreground image at S203.
4. the public place crowd estimate's method based on dual camera according to claim 1, it is characterized in that: at S204 in the stage, the method of described detection face complexion area is at first to adopt based on the method for Haar sorter to determine human face region, then by adding up skin pixel in the human face region based on the skin color detection method of YCrCb color space.
5. the public place crowd estimate's method based on dual camera according to claim 1, it is characterized in that: at S205 in the stage, the edge pixel method of described calculating foreground image is to adopt the canny operator that foreground image is carried out rim detection, and the result is carried out noise reduction, filtering handle, remove and disturb bigger zone, count internal edge pixel count and external margin pixel count then.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310125865.6A CN103258232B (en) | 2013-04-12 | 2013-04-12 | A kind of public place crowd estimate's method based on dual camera |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310125865.6A CN103258232B (en) | 2013-04-12 | 2013-04-12 | A kind of public place crowd estimate's method based on dual camera |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103258232A true CN103258232A (en) | 2013-08-21 |
| CN103258232B CN103258232B (en) | 2015-10-28 |
Family
ID=48962134
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310125865.6A Expired - Fee Related CN103258232B (en) | 2013-04-12 | 2013-04-12 | A kind of public place crowd estimate's method based on dual camera |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103258232B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104298969A (en) * | 2014-09-25 | 2015-01-21 | 电子科技大学 | Crowd scale statistical method based on color and HAAR feature fusion |
| CN105426828A (en) * | 2015-11-10 | 2016-03-23 | 浙江宇视科技有限公司 | Face detection method, face detection device and face detection system |
| CN105740945A (en) * | 2016-02-04 | 2016-07-06 | 中山大学 | People counting method based on video analysis |
| CN108537687A (en) * | 2018-04-10 | 2018-09-14 | 陕西智多搭智能科技有限公司 | A kind of school lunch service's management system |
| CN108596019A (en) * | 2018-03-12 | 2018-09-28 | 杭州英歌智达科技有限公司 | A kind of camera head monitor method and system based on diagonal gesture |
| CN108939230A (en) * | 2018-05-31 | 2018-12-07 | 李洪均 | Gymnasium oxygen generating plant, gymnasium oxygen system, method for supplying oxygen and device |
| US10176381B2 (en) | 2014-04-30 | 2019-01-08 | Carrier Corporation | Video analysis system for energy-consuming building equipment and intelligent building management system |
| CN109791638A (en) * | 2016-09-30 | 2019-05-21 | 三菱电机大楼技术服务株式会社 | Place number prediction meanss, equipment management system and program |
| CN109919064A (en) * | 2019-02-27 | 2019-06-21 | 湖南信达通信息技术有限公司 | Demographic method and device in real time in a kind of rail transit cars |
| CN110210045A (en) * | 2018-03-28 | 2019-09-06 | 腾讯大地通途（北京）科技有限公司 | Number evaluation method, device and the storage medium of target area |
| CN110414301A (en) * | 2018-04-28 | 2019-11-05 | 中山大学 | It is a kind of based on double compartment crowd density estimation methods for taking the photograph head |
| CN110717466A (en) * | 2019-10-15 | 2020-01-21 | 中国电建集团成都勘测设计研究院有限公司 | Method for returning position of safety helmet based on face detection frame |
| CN111881829A (en) * | 2020-07-28 | 2020-11-03 | 广州技诺智能设备有限公司 | Method and device for detecting personnel and calculating personnel number in public place and monitoring equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060269145A1 (en) * | 2003-04-17 | 2006-11-30 | The University Of Dundee | Method and system for determining object pose from images |
| CN102087707A (en) * | 2009-12-03 | 2011-06-08 | 索尼株式会社 | Image processing equipment and image processing method |
-
2013
- 2013-04-12 CN CN201310125865.6A patent/CN103258232B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060269145A1 (en) * | 2003-04-17 | 2006-11-30 | The University Of Dundee | Method and system for determining object pose from images |
| CN102087707A (en) * | 2009-12-03 | 2011-06-08 | 索尼株式会社 | Image processing equipment and image processing method |
Non-Patent Citations (3)
| Title |
|---|
| JAVIER BARANDIARAN等: "Real-Time People Counting Using Multiple Lines", 《NINTH INTERNATIONAL WORKSHOP ON IMAGE ANALYSIS FOR MULTIMEDIA INTERACTIVE SERVICES》, 9 May 2008 (2008-05-09), pages 159 - 162, XP031281847 * |
| 汤石晨等: "基于光流法和行人面积特征人数统计方法研究", 《图学学报》, vol. 34, no. 1, 15 January 2013 (2013-01-15), pages 139 - 144 * |
| 薛陈: "复杂场景下的人数统计", 《电子测量技术》, vol. 35, no. 12, 15 December 2012 (2012-12-15), pages 63 - 66 * |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10176381B2 (en) | 2014-04-30 | 2019-01-08 | Carrier Corporation | Video analysis system for energy-consuming building equipment and intelligent building management system |
| CN104298969A (en) * | 2014-09-25 | 2015-01-21 | 电子科技大学 | Crowd scale statistical method based on color and HAAR feature fusion |
| CN104298969B (en) * | 2014-09-25 | 2018-06-26 | 电子科技大学 | Crowd size's statistical method based on color Yu HAAR Fusion Features |
| CN105426828A (en) * | 2015-11-10 | 2016-03-23 | 浙江宇视科技有限公司 | Face detection method, face detection device and face detection system |
| CN105426828B (en) * | 2015-11-10 | 2019-02-15 | 浙江宇视科技有限公司 | Method for detecting human face, apparatus and system |
| CN105740945B (en) * | 2016-02-04 | 2018-03-16 | 中山大学 | A kind of people counting method based on video analysis |
| CN105740945A (en) * | 2016-02-04 | 2016-07-06 | 中山大学 | People counting method based on video analysis |
| CN109791638B (en) * | 2016-09-30 | 2023-09-15 | 三菱电机楼宇解决方案株式会社 | Device for predicting number of persons, device management system, and recording medium |
| CN109791638A (en) * | 2016-09-30 | 2019-05-21 | 三菱电机大楼技术服务株式会社 | Place number prediction meanss, equipment management system and program |
| CN108596019A (en) * | 2018-03-12 | 2018-09-28 | 杭州英歌智达科技有限公司 | A kind of camera head monitor method and system based on diagonal gesture |
| CN108596019B (en) * | 2018-03-12 | 2022-01-28 | 杭州英歌智达科技有限公司 | Camera monitoring method and system based on diagonal potential |
| CN110210045B (en) * | 2018-03-28 | 2022-03-08 | 腾讯大地通途（北京）科技有限公司 | Method and device for estimating number of people in target area and storage medium |
| CN110210045A (en) * | 2018-03-28 | 2019-09-06 | 腾讯大地通途（北京）科技有限公司 | Number evaluation method, device and the storage medium of target area |
| CN108537687A (en) * | 2018-04-10 | 2018-09-14 | 陕西智多搭智能科技有限公司 | A kind of school lunch service's management system |
| CN110414301A (en) * | 2018-04-28 | 2019-11-05 | 中山大学 | It is a kind of based on double compartment crowd density estimation methods for taking the photograph head |
| CN108939230B (en) * | 2018-05-31 | 2020-10-16 | 李洪均 | Gymnasium oxygen generation equipment, gymnasium oxygen supply system, and oxygen supply method and device |
| CN108939230A (en) * | 2018-05-31 | 2018-12-07 | 李洪均 | Gymnasium oxygen generating plant, gymnasium oxygen system, method for supplying oxygen and device |
| CN109919064B (en) * | 2019-02-27 | 2020-12-22 | 湖南信达通信息技术有限公司 | Real-time people counting method and device in rail transit carriage |
| CN109919064A (en) * | 2019-02-27 | 2019-06-21 | 湖南信达通信息技术有限公司 | Demographic method and device in real time in a kind of rail transit cars |
| CN110717466A (en) * | 2019-10-15 | 2020-01-21 | 中国电建集团成都勘测设计研究院有限公司 | Method for returning position of safety helmet based on face detection frame |
| CN110717466B (en) * | 2019-10-15 | 2023-06-20 | 中国电建集团成都勘测设计研究院有限公司 | Method for returning to position of safety helmet based on face detection frame |
| CN111881829A (en) * | 2020-07-28 | 2020-11-03 | 广州技诺智能设备有限公司 | Method and device for detecting personnel and calculating personnel number in public place and monitoring equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103258232B (en) | 2015-10-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103258232B (en) | A kind of public place crowd estimate's method based on dual camera | |
| CN103226860B (en) | Passage passenger traffic density estimation method | |
| JP6549797B2 (en) | Method and system for identifying head of passerby | |
| CN102542289B (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
| Lai et al. | Image-based vehicle tracking and classification on the highway | |
| CN101303727B (en) | Intelligent management method based on video human number Stat. and system thereof | |
| WO2017122258A1 (en) | Congestion-state-monitoring system | |
| CN107622258A (en) | A kind of rapid pedestrian detection method of combination static state low-level image feature and movable information | |
| CN105187785B (en) | A kind of across bayonet pedestrian's identifying system and method based on choice of dynamical notable feature | |
| US9672434B2 (en) | Video-based system and method for parking occupancy detection | |
| CN109684996B (en) | Real-time vehicle access identification method based on video | |
| CN105160297B (en) | Masked man's event automatic detection method based on features of skin colors | |
| CN104036236B (en) | A kind of face gender identification method based on multiparameter exponential weighting | |
| CN109918971B (en) | Method and device for detecting number of people in monitoring video | |
| CN103729858B (en) | A kind of video monitoring system is left over the detection method of article | |
| CN105404847A (en) | Real-time detection method for object left behind | |
| CN104504377A (en) | Bus passenger crowding degree identification system and method | |
| CN105678213A (en) | Dual-mode masked man event automatic detection method based on video characteristic statistics | |
| CN112287823A (en) | Facial mask identification method based on video monitoring | |
| CN106127292A (en) | Effusion meter counting method and equipment | |
| CN106599880A (en) | Discrimination method of the same person facing examination without monitor | |
| Alzughaibi et al. | Review of human motion detection based on background subtraction techniques | |
| CN105701474B (en) | A kind of video smoke recognition methods of color combining and external physical characteristic | |
| KR102244878B1 (en) | Cctv security system and method based on artificial intelligence | |
| Furuya et al. | Road intersection monitoring from video with large perspective deformation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20151028 Termination date: 20190412 |