CN102096813A - Blob-based pedestrian counting method under vertical visual angle of monocular camera - Google Patents
Blob-based pedestrian counting method under vertical visual angle of monocular camera Download PDFInfo
- Publication number
- CN102096813A CN102096813A CN2011100352740A CN201110035274A CN102096813A CN 102096813 A CN102096813 A CN 102096813A CN 2011100352740 A CN2011100352740 A CN 2011100352740A CN 201110035274 A CN201110035274 A CN 201110035274A CN 102096813 A CN102096813 A CN 102096813A
- Authority
- CN
- China
- Prior art keywords
- blob
- pedestrian
- counting
- judge
- tracks
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
- Closed-Circuit Television Systems (AREA)
Abstract
The invention discloses a Blob-based pedestrian counting method under a vertical visual angle of a monocular camera. Based on tracking and counting of a pedestrian Blob, the difficulties such as the technical difficulty caused by separation and combination of the pedestrian Blob are overcome by using a Kalman filter to track the pedestrian Blob and judging a movement trace of the pedestrian Blob. By the Blob-based pedestrian counting method, the technical difficulty of partitioning of the pedestrians is avoided and the counting precision is enhanced; moreover, the method is applicable to fields of market monitoring, emergency event detection and the like.
Description
Technical field
The invention belongs to the technology of image processing field and computer vision field, mainly be meant the pedestrian's method of counting under the monocular-camera vertical angle of view.
Background technology
Existing pedestrian's method of counting generally can be divided into two kinds: 1. based on pedestrian's method of counting of video camera angle of squint; 2. based on pedestrian's method of counting of video camera vertical angle of view.In the pedestrian's method of counting based on the video camera angle of squint, the video camera shooting angle generally places the position about 45 °, detects each pedestrian earlier by the information such as body shape feature of obtaining the pedestrian, and then follows the tracks of each pedestrian and realize counting.But when the situation of blocking mutually between the pedestrian was relatively more serious, this method just was difficult to correct differentiation pedestrian, to such an extent as to the pedestrian detection inefficacy, thus can't realize effectively that the pedestrian counts.And in the applied environment of reality, the occlusion issue between the pedestrian is ubiquitous, so in order to solve occlusion issue, obtained using widely based on pedestrian's method of counting of video camera vertical angle of view.Existing pedestrian's method of counting based on the video camera vertical angle of view generally all requires to cut apart (detection) accurately and goes out the pedestrian, and the accurate cutting techniques of present a plurality of pedestrian also is an insurmountable difficult problem.A plurality of pedestrians are accurately cut apart will definitely not increase the difficulty that the pedestrian follows the tracks of to a great extent, finally can obviously reduce the accuracy rate of pedestrian's counting.
Summary of the invention
The object of the present invention is to provide under a kind of monocular-camera vertical angle of view based on pedestrian's method of counting of Blob, this method and technology is accurate, and error is little.For achieving the above object, the technical solution used in the present invention is: based on pedestrian's method of counting of Blob, may further comprise the steps under the monocular-camera vertical angle of view:
(1) adopt the background subtraction point-score to carry out the Blob foreground detection to the pedestrian under the monocular-camera vertical angle of view;
(2) on the basis of step (1), judge the pedestrian Blob that newly enters in the visual angle;
(3) the pedestrian Blob that newly enters in the visual angle is followed the tracks of, it is added tracking queue;
(4) the pedestrian Blob that step (3) is followed the tracks of counts, and judges the number that enters and go out according to direction.
The pedestrian Blob that obtains in the step (1) adopts the method for cluster, finds out each Blob connected region and confines, and obtains the centroid position of each connected region of confining; Then, suppose that barycenter of rectangle frame that any two connected regions are confined is c1, coordinate is (x1, y1), another is c2, and coordinate is (x2, y2), barycenter is that the long limit of the rectangle frame of c1 is p1, and minor face is q1, and barycenter is that the long limit of the rectangle frame of c2 is p2, minor face is q2, if satisfy following formula, then poly-is a class, belongs to same prospect Blob.
In the step (3), set a Blob area threshold, will be less than the Blob deletion of this threshold value.
Step (3) judges whether to be the pedestrian Blob that newly enters, and adopts following steps: a. to judge whether Blob has been present in the tracking queue, if be present in the tracking queue, has judged with regard to not needing, directly with its deletion again; B. on the result that step a draws, Blob in five frames judges to continued presence, determining step is: Blob keeps under the situation of certain distance apart from the border, judge by the movement velocity of color development and Blob whether Blob belongs to the pedestrian, will be judged to be the pedestrian Blob that newly enters then and join in the tracking queue.
Adopt Kalman filter that Blob is followed the tracks of in the step (3), give unique ID value to the pedestrian Blob that newly enters the visual angle, utilize Kalman filter that a few frame movement positions below the Blob that follows the tracks of are predicted, if following a few frame two or more pedestrian Blob occur in the position of prediction, also give unique ID value to additional pedestrian Blob piece.
The step of pedestrian's counting is as follows in the step (4):
A. determine the counting region: the zone of selected pedestrian's counting in guarded region;
B. judge Blob direction of motion: utilize the movement locus of Blob in continuous several frames to judge the direction of motion of Blob, judge that promptly it enters or leaves.
C. the pedestrian counts
According to the direction of judging previously, when the Blob that follows the tracks of enters the counting region, calculate the area of this Blob, according to the reasonable area of the single Blob that has added up, calculate the number that comprises among this Blob then, it is added on the global statistics number finish counting.
Beneficial effect of the present invention: pedestrian's method of counting of the present invention, adopt Kalman filter tracking pedestrians Blob, and, judge division and the merging of pedestrian Blob by predicting the position of following a few frame pedestrian Blob, accurately calculate the number of turnover direction.
Pedestrian's method of counting based on the monocular-camera vertical angle of view is an important component part of intelligent video monitoring, can be widely used in following occasion:
1) people flow rate statistical: than higher place, carry out quantitative statistics into and out of the stream of people as the movable active place of the stream of peoples such as bank, parking lot in demand for security.
2) anomalous event detects: in video monitoring some anomalous events are detected as run, pace up and down, delay etc. and warning automatically.
3) the public place crowded state detects: in some public places, carry out guest flow statistics as park, museum, market, the personnel that assist management effectively dredge and control flow of the people.
Description of drawings
Fig. 1 .a-Fig. 1 .b adopts the background subtraction point-score to carry out the result of foreground detection;
Fig. 2 .a-Fig. 2 .d is that the present invention adopts the result schematic diagram of clustering method to the foreground detection Flame Image Process;
Fig. 3 .a-Fig. 3 .d is the synoptic diagram of pedestrian Blob division among the present invention;
Fig. 4 .a-Fig. 4 .d is the synoptic diagram that pedestrian Blob merges among the present invention;
Fig. 5 is pedestrian's method of counting total algorithm schematic flow sheet of the present invention;
Fig. 6 is that pedestrian of the present invention side enters synoptic diagram;
Fig. 7 is pedestrian counting region and direction of motion synoptic diagram;
Fig. 8 pedestrian's direction of motion is judged synoptic diagram.
Specific embodiment
The base unit that the present invention handles is pedestrian Blob.Specific embodiment is divided into 4 stages: pedestrian Blob foreground detection, the pedestrian Blob that newly enters guarded region detect, pedestrian Blob follows the tracks of, pedestrian's counting among the Blob.
1. adopt the background subtraction point-score to carry out the Blob foreground detection to the pedestrian under the monocular-camera vertical angle of view;
Because video camera is fixing ornaments, so can obtain the sport foreground zone by the method for background difference.Handling what obtain through the method for background difference is exactly pedestrian Blob and since in our system main statistics be the motion pedestrian, so to actionless pedestrian counting not in guarded region always.In fact, actionless pedestrian is few in guarded region fully, and so long as not long-time transfixion, the method for background difference just can obtain pedestrian Blob.The method of background difference is comparative maturity, and what we adopted is that Maddalena2008 is published in the method that the article " A Self-Organizing Approach to Background Subtraction for Visual Surveillance Applications " on the IEEE Trans on Image Processing is mentioned.The effect of background difference is as scheming: Fig. 1 .a is an original image, and Fig. 1 .b is the differentiated two-value foreground picture of background, and the part of confining among Fig. 1 .b is detected pedestrian.
But in some cases, when for example light changes relatively acutely, the effect of background difference can be subjected to some influences, bigger cavity can appear in the prospect that extracts, with effectively filtering of morphology methods, so we have proposed a kind ofly simply based on the method for cluster the result of background difference to be carried out aftertreatment.Shown in Fig. 2 .b, the result of background difference is unsatisfactory, and bigger cavity is arranged in the foreground picture that obtains, and also can be identified as a prospect Blob in order to make the prospect that obtains under the situation that the part separation occurs, and we have proposed a kind of simple and effective clustering method.At first, we find each connected region and confine in Fig. 2 .b, shown in Fig. 2 .c; Then, obtain the centroid position of each connected region of confining; At last, suppose that barycenter of rectangle frame that any two connected regions are confined is c1, coordinate be (x1, y1), another is c2, coordinate be (x2, y2).Barycenter is that the long limit of the rectangle frame of c1 is p1, and minor face is q1, and barycenter is that the long limit of the rectangle frame of c2 is p2, and minor face is q2, if satisfy the following formula of formula, then poly-is a class, belongs to same prospect Blob, and last cluster result is shown in Fig. 2 .d.
2. the pedestrian Blob that newly enters guarded region detects
Detect on the basis of the mask that we obtain in foreground detection and newly enter the pedestrian Blob of scene.The cardinal rule that detects fresh target is: comprise this connected region in the continuous multiple frames image, and have consistent rational speed.
In this patent, we set connected region and are present at least and could consider the adding tracking queue in continuous 5 frames, and concrete testing process is as follows:
1) at first, the Blob less than reasonable value filters out with some areas, then remaining Blob is judged whether it has been present in the tracking queue, if be present in the tracking queue, has judged in this module with regard to not needing again, and can directly delete.Criterion is to mate with Blob in the tracking queue and new detected Blob, and matching formula is as follows:
If newly the center of detected a certain Blob is (x1, y1), long is h1, wide is w1; The center of a certain Blob in the tracking queue is that (x2 y2), long is h2, and wide is w2.
(5-2-3)
If formula (5-2-2) or formula (5-2-3) have an establishment among both, show that this Blob is tracked, need not judge in this module, it should be deleted.If traveled through each Blob in the tracking queue, formula (5-2-2) and formula (5-2-3) all are false, and show that this Blob in the new frame is the object of newly coming in, and need be determined further.
2) we judge its rationality again to the Blob of continued presence in five frames.The criterion that we judge is followed successively by: Blob whether comprises the number of people in the distance on border, Blob and whether Blob has rational movement velocity.Specific implementation process is as follows:
From the 6th frame to the 10 frames, we only use the distance of preceding two judgment criterion: Blob from the border, whether include the number of people among the Blob.If can not detect the number of people always in preceding 10 frames, we just need the 3rd judgment criterion of adding, as long as Blob satisfies criterion one, and satisfy among criterion two and three any one, and this Blob just is regarded as rational Blob, need to add to follow the tracks of and tabulate.Such Blob detects strategy can both obtain effect preferably on accuracy and real-time.
Wherein, the specific implementation of three criterions is as follows:
Criterion one: Blob is from the distance on border
The distance of the center of Blob from the guarded region border is the length of Blob or wide 1.5 times at least.
Criterion two: whether comprise the number of people among the Blob
At first find out circle in the prospect by the Hough conversion, and then judge in the circle of finding out whether meet the color development feature, selecting color development here is black.Color development selects black can solve the situation of the overwhelming majority, if meet some special special circumstances as shaven head, band cap, dyeing, the rationality of Blob will have criterion three to judge.
Whether criterion three: Blob has rational speed.
We are safeguarding the formation at the Blob center of continuous 10 frames of record, use the new frame of the 10th frame back to upgrade this formation simultaneously.With the Blob center simulation straight line equation of these records, the slope of straight-line equation is exactly the movement velocity of Blob then.As the straight-line equation that simulates is y=ax+b, and we just judge whether the value of a is reasonable.In this patent, we choose 1/10 of the length of guarded region or width and are threshold values, are considered to rational greater than a of this threshold values.
After final rationality judgement, we will filter the pedestrian that these remaining Blob are considered as newly entering guarded region, and they are added tracking queue, handle in tracking module.
Detection algorithm by above-mentioned strategy can be obtained high processing efficient, can solve run, from complex situations such as the surveyed area side enter.As illustrate shown in Figure 6, when the people when the guarded region lateral edge is come in because its frame at guarded region is less relatively, therefore the efficient to algorithm needs higher requirement, the algorithm that this patent proposes previously can be handled such special circumstances preferably.
3. pedestrian Blob follows the tracks of
3.1 Blob label
In tracking, in order to distinguish each Blob, we give unique ID number successively since 1 when each Blob appears at guarded region for the first time.In fact, whole tracing process is exactly to distribute in each frame the correct ID of each Blob number.In continuous video frames we for the first time detected Blob appear at guarded region for the first time, to its distribute one new ID number, the positional information of all Blob that obtain is deposited in a chained list preserves; Afterwards, we deposit the positional information of detected all Blob of next frame in second chained list and preserve, and if between the corresponding Blob position coincidence is arranged in first chained list, illustrate that then this Blob had distributed ID; Otherwise, think that then this Blob is emerging, compose to give one new ID number because same Blob is adjacent certainly in the continuous video frames, uses the same method and handle all frame of video.What need replenish is that we have at first carried out pre-service to all Blob that obtain according to the area size in each frame, deleted those because very little zones of area that picture noise produces.
3.2 the division of Blob and merging
We can simply be divided into two classes to the tracing process of Blob: a class is that Blob to the process of leaving guarded region division and situation about merging takes place never from entering guarded region, no matter how many people are this Blob in fact comprised, for example two people are mutually against the guarded region of passing by, and in fact this Blob has just comprised two people; The another kind of Blob of being divides and merges no matter how many times has taken place through division and situation about merging have taken place in the process of guarded region.
First kind problem is fairly simple, and we can be regarded as the tracking problem of single goal, has all done detailed introduction in a lot of documents, can solve by the Kalman filter prediction or based on mixed informations such as color histogram, area, coordinates.The second class problem more complicated has related to multiple goal division and merging each other, and next we will provide our solution at this situation.
3.2.1 Blob division
If contain many people among the prospect Blob, in guarded region, have because people's the direction of motion and the speed of travel different, this big Blob will resolve into several little Blob so, for example in the represented actual scene of Fig. 3, the detected big Blob of Fig. 3 .a will split into two little Blob, shown in Fig. 3 .b, this will bring difficulty to following the tracks of Blob accurately.
We have detected a big Blob to suppose present frame, if we have detected a plurality of little Blob next frame, the position that we may occur by the big Blob of Kalman filter prediction next frame, prediction at this moment to the Blob position may with detected a plurality of little Blob in one or several coincide.We carried preamble, the process that Blob follows the tracks of is exactly to solve the problem of how to distribute ID number, our method is: the ID that supposes big Blob is x, n altogether of detected all little Blob in division back, ID number of detected all the little Blob in division back is x so, x+1 ... .x+n-1, wherein ID number with original big Blob ID number is identical is among the little Blob that coincides with prediction any one, shown in Fig. 3 .c and Fig. 3 .d.In order to see more clearlyly, we have chosen the 72nd frame and the 78th frame in the experimentation and have illustrated, and do not get two continuous frames.Can see that ID is that to have split into ID be respectively 2 and 3 little Blob for 2 big Blob.
In the process of Blob division, the reader may produce query be if two aspectant divisions of people are come situation how about because each Blob has directional information, before not meeting, they have separately ID number.Kalman filter has been considered direction in prediction, so when its tangible two people meet face-to-face, even in detected same big Blob, do not become for their ID number, the division back still keeps original separately ID number.
3.2.2 Blob merges
It also is the problem that may run in the Blob tracing process that Blob merges.Because pedestrian's direction of motion is arbitrarily, when they are close mutually, the situation that a plurality of little Blob are merged into a big Blob will appear.For example in the represented actual scene of Fig. 4, detected two the little Blob of Fig. 4 .a will be merged into a big Blob, and shown in Fig. 4 .b, this also can bring difficulty to following the tracks of Blob accurately.
We whenever detect an emerging Blob, will compose a unique ID number to it, and go to the position of predicting that each Blob next frame may occur by Kalman filter, and the process of this prediction has been with directional information.Shown in Fig. 4 .c, ID is respectively that two Blob of 5 and 6 are detected at the 25th frame, and to time the 28th frame, two little Blob just have been merged into a big Blob, shown in Fig. 4 .d.The problem that we will solve is how to go to distribute ID number of the big Blob that is merged into.We begin to consider be to this big Blob that is merged into distribute in addition one ID number, the problem of bringing like this is that ID number distribution is very chaotic, the big Blob that is merged into must note the information of each original little Blob.Under the situation that repeatedly divides and merge, just be difficult to control well ID number distribution, also give how to make whole number system bring problem by clear counting.
For ID number distribution clearly in the process that makes whole Blob tracking, little Blob before we wish to merge is after being merged into big Blob, still can keep original ID number, up to leaving guarded region or this big Blob splits into a plurality of little Blob again, just reach the effect shown in Fig. 4 .d.
We are the method that 6 little Blob illustrates us with ID.In case the area of the Blob that the area of the detected Blob of present frame significantly predicts greater than former frame, and both have taken place overlappingly, and we just think has a plurality of little Blob just to be merged into a big Blob.In the scene shown in Fig. 4 .c be exactly ID be 5 and ID be that two little Blob of 6 just have been merged into a big Blob.In this case, we as the position at Blob present frame place, also need not change the position of prediction for ID number.In all frames that subsequent again merging is followed the tracks of, we all be the position of prediction as the position on the coupling, until the area of the Blob of prediction and detected Blob area ratio be near 1 o'clock, we think that division has taken place the big Blob that is merged into.According to this way, we have guaranteed to merge, and each original little Blob has still kept merging preceding ID number when following the tracks of.
4.Blob middle pedestrian's counting
Because our camera height is certain, detected each pedestrian's Blob size basically identical in the prospect, so the core of pedestrian's counting is to use area to add up the number of pedestrian among the turnover Blob.Detailed process is as follows:
1) determines the counting region
The zone of our selected pedestrian's counting in guarded region, when tracking Blob was in the counting region, we just counted it.
In this patent, our mode of selected counting region is as follows: establish the Blob center in that (this border is the border on the Blob direction of motion) distance is d from the border on the Y direction, the height of Blob is h, and the counting region that we select is: 1.05*h/2＜d＜1.8*h/2.As shown in Figure 7.
2) judge Blob direction of motion
Utilize the movement locus of Blob in continuous several frames to judge the direction of motion of Blob, judge that promptly it enters or leaves.Shown in synoptic diagram, we use continuous 3 frames of this Blob, judge the position relation of the 3rd frame and first interframe.The first and the 3rd frame, can judge its direction of motion from Blob, illustrate that wherein direction of motion shown in Figure 8 is in outdoor inlet chamber along the position on the Y direction.
3) counting
According to the direction of judging previously, when the Blob that follows the tracks of enters the counting region, we calculate the area of this Blob, then according to the reasonable area of the single Blob that has added up, calculate the number that comprises among this Blob, it is added on the global statistics number finish counting.From tracking queue, delete the Blob of this tracking at last.
It should be noted that at last: the above only is embodiments of the invention, be not limited to the present invention, although the present invention is had been described in detail with reference to previous embodiment, for a person skilled in the art, it still can be made amendment to the technical scheme that aforementioned each embodiment put down in writing, and perhaps part technical characterictic wherein is equal to replacement.Within the spirit and principles in the present invention all, any modification of being done, be equal to replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (6)
- Under the monocular-camera vertical angle of view based on pedestrian's method of counting of Blob, it is characterized in that, may further comprise the steps:(1) adopt the background subtraction point-score to carry out the Blob foreground detection to the pedestrian under the monocular-camera vertical angle of view;(2) on the basis of step (1), judge the pedestrian Blob that newly enters in the visual angle;(3) the pedestrian Blob that newly enters in the visual angle is followed the tracks of, it is added tracking queue;(4) the pedestrian Blob that step (3) is followed the tracks of counts, and judges the number that enters and go out according to direction.
- 2. pedestrian's method of counting according to claim 1 is characterized in that, to adopted the method for cluster by the pedestrian Blob that obtains in the step (1), finds out each Blob connected region and confines, and obtains the centroid position of each connected region of confining; Then, suppose that barycenter of rectangle frame that any two connected regions are confined is c1, coordinate is (x1, y1), another is c2, and coordinate is (x2, y2), barycenter is that the long limit of the rectangle frame of c1 is p1, and minor face is q1, and barycenter is that the long limit of the rectangle frame of c2 is p2, minor face is q2, if satisfy following formula, then poly-is a class, belongs to same prospect Blob.
- 3. pedestrian's method of counting according to claim 2 is characterized in that, in the step (3), sets a Blob area threshold, will be less than the Blob deletion of this threshold value.
- 4. pedestrian's method of counting according to claim 3, it is characterized in that, step (3) judges whether the pedestrian Blob for newly entering, adopt following steps: a. to judge whether Blob has been present in the tracking queue, if be present in the tracking queue, judged again with regard to not needing, directly with its deletion; B. on the result that step a draws, Blob in five frames judges to continued presence, determining step is: Blob keeps under the situation of certain distance apart from the border, judge by the movement velocity of color development and Blob whether Blob belongs to the pedestrian, will be judged to be the pedestrian Blob that newly enters then and join in the tracking queue.
- 5. pedestrian's method of counting according to claim 4, it is characterized in that, adopt Kalman filter that Blob is followed the tracks of in the step (3), give unique ID value to the pedestrian Blob that newly enters the visual angle, utilize Kalman filter that a few frame movement positions below the Blob that follows the tracks of are predicted, if following a few frame two or more pedestrian Blob occur in the position of prediction, also give unique ID value to additional pedestrian Blob piece.
- 6. pedestrian's method of counting according to claim 1 is characterized in that, the step of pedestrian's counting is as follows in the step (4):A. determine the counting region: the zone of selected pedestrian's counting in guarded region;B. judge Blob direction of motion: utilize the movement locus of Blob in continuous several frames to judge the direction of motion of Blob, judge that promptly it enters or leaves;C. the pedestrian countsAccording to the direction of judging previously, when the Blob that follows the tracks of enters the counting region, calculate the area of this Blob, according to the reasonable area of the single Blob that has added up, calculate the number that comprises among this Blob then, it is added on the global statistics number finish counting.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011100352740A CN102096813A (en) | 2011-02-10 | 2011-02-10 | Blob-based pedestrian counting method under vertical visual angle of monocular camera |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011100352740A CN102096813A (en) | 2011-02-10 | 2011-02-10 | Blob-based pedestrian counting method under vertical visual angle of monocular camera |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102096813A true CN102096813A (en) | 2011-06-15 |
Family
ID=44129901
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2011100352740A Pending CN102096813A (en) | 2011-02-10 | 2011-02-10 | Blob-based pedestrian counting method under vertical visual angle of monocular camera |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102096813A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013189464A3 (en) * | 2012-11-28 | 2014-02-13 | 中兴通讯股份有限公司 | Pedestrian tracking and counting method and device for near-front top-view monitoring video |
| CN103871082A (en) * | 2014-03-31 | 2014-06-18 | 百年金海科技有限公司 | Method for counting people stream based on security and protection video image |
| TWI611378B (en) * | 2016-03-03 | 2018-01-11 | Mitsubishi Electric Corp | Chaos prediction device and chaotic prediction method |
| CN110309717A (en) * | 2019-05-23 | 2019-10-08 | 南京熊猫电子股份有限公司 | A kind of pedestrian counting method based on deep neural network |
| CN110688924A (en) * | 2019-09-19 | 2020-01-14 | 天津天地伟业机器人技术有限公司 | RFCN-based vertical monocular passenger flow volume statistical method |
| CN114913195A (en) * | 2022-07-13 | 2022-08-16 | 苏州魔视智能科技有限公司 | People counting method, device and system for target area |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060177097A1 (en) * | 2002-06-14 | 2006-08-10 | Kikuo Fujimura | Pedestrian detection and tracking with night vision |
| US20080205703A1 (en) * | 2005-05-27 | 2008-08-28 | International Business Machines Corporation | Methods and Apparatus for Automatically Tracking Moving Entities Entering and Exiting a Specified Region |
| CN101847265A (en) * | 2010-04-20 | 2010-09-29 | 上海理工大学 | Method for extracting moving objects and partitioning multiple objects used in bus passenger flow statistical system |
-
2011
- 2011-02-10 CN CN2011100352740A patent/CN102096813A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060177097A1 (en) * | 2002-06-14 | 2006-08-10 | Kikuo Fujimura | Pedestrian detection and tracking with night vision |
| US20080205703A1 (en) * | 2005-05-27 | 2008-08-28 | International Business Machines Corporation | Methods and Apparatus for Automatically Tracking Moving Entities Entering and Exiting a Specified Region |
| CN101847265A (en) * | 2010-04-20 | 2010-09-29 | 上海理工大学 | Method for extracting moving objects and partitioning multiple objects used in bus passenger flow statistical system |
Non-Patent Citations (3)
| Title |
|---|
| 张景元: "基于OpenCV和无线通信的智能视频监控报警系统设计", 《中国优秀硕士学位论文全文数据库信息科技辑2010年》, no. 02, 15 February 2010 (2010-02-15), pages 27 - 29 * |
| 郭凡 等: "基于云台摄像的实时车速检测算法", 《计算机辅助设计与图像学学报》, vol. 18, no. 9, 30 September 2006 (2006-09-30), pages 1341 - 1342 * |
| 靳晶 等: "使用单目摄像头的实时行人计数算法", 《计算机工程》, vol. 35, no. 23, 31 December 2009 (2009-12-31), pages 204 - 206 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013189464A3 (en) * | 2012-11-28 | 2014-02-13 | 中兴通讯股份有限公司 | Pedestrian tracking and counting method and device for near-front top-view monitoring video |
| CN103854273A (en) * | 2012-11-28 | 2014-06-11 | 中兴通讯股份有限公司 | Pedestrian tracking counting method and device based on near forward direction overlooking monitoring video |
| US9773317B2 (en) | 2012-11-28 | 2017-09-26 | Xi'an Zhongxing New Software Co. Ltd. | Pedestrian tracking and counting method and device for near-front top-view monitoring video |
| CN103871082A (en) * | 2014-03-31 | 2014-06-18 | 百年金海科技有限公司 | Method for counting people stream based on security and protection video image |
| TWI611378B (en) * | 2016-03-03 | 2018-01-11 | Mitsubishi Electric Corp | Chaos prediction device and chaotic prediction method |
| CN110309717A (en) * | 2019-05-23 | 2019-10-08 | 南京熊猫电子股份有限公司 | A kind of pedestrian counting method based on deep neural network |
| CN110688924A (en) * | 2019-09-19 | 2020-01-14 | 天津天地伟业机器人技术有限公司 | RFCN-based vertical monocular passenger flow volume statistical method |
| CN114913195A (en) * | 2022-07-13 | 2022-08-16 | 苏州魔视智能科技有限公司 | People counting method, device and system for target area |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104021570B (en) | Directive in a kind of video cross the border and mix line detecting method | |
| CN102096813A (en) | Blob-based pedestrian counting method under vertical visual angle of monocular camera | |
| CN105844234B (en) | Method and equipment for counting people based on head and shoulder detection | |
| Qi et al. | Efficient railway tracks detection and turnouts recognition method using HOG features | |
| CN102243765A (en) | Multi-camera-based multi-objective positioning tracking method and system | |
| CN104091348A (en) | Multi-target tracking method integrating obvious characteristics and block division templates | |
| CN101794385A (en) | Multi-angle multi-target fast human face tracking method used in video sequence | |
| CN101777188A (en) | Real-time bus passenger flow volume statistical method | |
| CN103729861A (en) | Multiple object tracking method | |
| Salvi | An automated vehicle counting system based on blob analysis for traffic surveillance | |
| CN103325115A (en) | Pedestrian counting monitoring method based on head top camera | |
| CN106920253A (en) | It is a kind of based on the multi-object tracking method for blocking layering | |
| CN107659754A (en) | Effective method for concentration of monitor video in the case of a kind of leaf disturbance | |
| Masoud et al. | Robust pedestrian tracking using a model-based approach | |
| CN106529434B (en) | The shoal of fish individual goal tracking of view-based access control model attention model | |
| Liu et al. | Dloam: Real-time and robust lidar slam system based on cnn in dynamic urban environments | |
| CN103677734A (en) | Multi-target data association algorithm based on feature matching matrix | |
| CN102043957A (en) | Vehicle segmentation method based on concave spots of image | |
| Li et al. | Intelligent transportation video tracking technology based on computer and image processing technology | |
| CN106530325A (en) | Multi-target visual detection and tracking method | |
| Ullah | Crowd motion analysis: segmentation, anomaly detection, and behavior classification | |
| Hai-Feng et al. | Vehicle abnormal behavior detection system based on video | |
| CN109410578A (en) | A kind of traffic intersection section traffic flow detecting method based on computer vision | |
| CN108932728A (en) | The tracking of view-based access control model monitoring | |
| Kollek et al. | Real-time traffic counting on resource constrained embedded systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20110615 |
|
| RJ01 | Rejection of invention patent application after publication |