CN105046316B - A kind of two-way pedestrian counting method of laser returned based on Gaussian process - Google Patents
A kind of two-way pedestrian counting method of laser returned based on Gaussian process Download PDFInfo
- Publication number
- CN105046316B CN105046316B CN201510240482.2A CN201510240482A CN105046316B CN 105046316 B CN105046316 B CN 105046316B CN 201510240482 A CN201510240482 A CN 201510240482A CN 105046316 B CN105046316 B CN 105046316B
- Authority
- CN
- China
- Prior art keywords
- height
- agglomerate
- pedestrian
- test point
- point
- 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.)
- Active
Links
Landscapes
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of two-way pedestrian counting methods of laser returned based on Gaussian process:Laser range finder detects layback information, and range data is converted to altitude information;By the operation with background model, the foreground height of test point is obtained;Generate sequential height map;Pedestrian level agglomerate is extracted in sequential height map;The geometry and shape feature of height agglomerate are chosen, Gaussian process homing method is chosen, pedestrian level is calculated and rolls into a ball pedestrian's quantity in the block;Using the sliding window of dynamic size, pedestrian level agglomerate is applied in the projection of time shaft;By calculating the maximum point of Mean curve, the head zone of pedestrian is positioned;After completing pedestrian head positioning, using voting method, judge pedestrian into outgoing direction.The laser two-way pedestrian counting method provided by the invention returned based on Gaussian process can realize fast and accurately indoor and outdoor pedestrian counting and walking direction in such a way that top set laser range finder casts oblique rays on scanning under the premise of not with pedestrian contact.
Description
Technical field
The present invention relates to a kind of two-way pedestrian counting methods of laser returned based on Gaussian process.
Background technology
With the development of the social economy, the quantity of city commercial real estate increasingly increases, scale is also gradually expanded.Pedestrian counting
Important decision-making foundation can be provided for the operator of commercial real estate and municipal public safety manager, be an existing business valence
Value and the work for having security implications.Currently, there is the method using contact gate pedestrian counting in public place, also have non-contact
The method of the infrared two-value laser of formula or video pedestrian counting.These method of counting have respective advantage under special scenes, but all
Shortcomings, it is difficult to reach the desired counting effect of user.
In special scenes, the pedestrian counting method based on common monocular-camera may be implemented preferably to count effect.
But change under violent scene, the scene of half-light or pedestrian block apparent scene in light, it is based on common monocular-camera
Method of counting also cannot achieve accurate counting.Have to research and propose at present and carries out pedestrian counting using Infrared LASER Ranger
Method, this method count the quantity for calculating pedestrian by detecting height change, cannot achieve standard under the intensive scene of pedestrian
True pedestrian counting, and may be unable to judge accurately pedestrian into outgoing direction.
Therefore, market needs a kind of new method of counting, to improve the accuracy of the counting under the intensive scene of pedestrian.
Invention content
In order to overcome the existing pedestrian counting method based on laser not counted to pedestrian side by side effectively, pedestrian is judged
The problem of into outgoing direction, the present invention provide a kind of two-way pedestrian counting method of the laser returned based on Gaussian process.In order to reach
Above-mentioned purpose, the present invention adopt the following technical scheme that:
A kind of two-way pedestrian counting method of laser returned based on Gaussian process, is included the following steps:
Range data is converted to altitude information by laser range finder detection layback information after triangulo operation;It is logical
The operation with background model is crossed, the foreground height of test point is obtained;Through accumulation after a period of time, with the foreground of each test point
Height generates sequential height map;It is detected a cluster in sequential height map, extracts pedestrian level agglomerate;Choose height agglomerate
Geometry and shape feature, choose Gaussian process homing method, calculate pedestrian level and roll into a ball pedestrian's quantity in the block;
Using the sliding window of a dynamic size, pedestrian level agglomerate is applied in the projection of time shaft;Sliding window
The size of mouth is codetermined by the width and pedestrian's regression result of pedestrian level agglomerate;The operation that sliding window executes includes calculating
The mean value and variance of detection point height in window;By calculating the maximum point of Mean curve, the head zone of pedestrian is positioned;It is complete
After being positioned at pedestrian head, using voting method, judge pedestrian into outgoing direction.
The laser two-way pedestrian counting method provided by the invention returned based on Gaussian process passes through top set laser range finder
The mode of scanning is casted oblique rays on, can realize fast and accurately indoor and outdoor pedestrian counting and direction under the premise of not with pedestrian contact
Judge.
Description of the drawings
Attached drawing described herein is used to provide further understanding of the present invention, and is constituted part of this application, not
Inappropriate limitation of the present invention is constituted, in the accompanying drawings:
Fig. 1 is a kind of flow chart of two-way pedestrian counting method of laser returned based on Gaussian process provided by the invention;
Fig. 2 is a kind of deployed with devices of two-way pedestrian counting method of laser returned based on Gaussian process provided by the invention
Front view;
Fig. 3 is a kind of deployed with devices of two-way pedestrian counting method of laser returned based on Gaussian process provided by the invention
Lateral plan;
Fig. 4 is a kind of sequential height of two-way pedestrian counting method of laser returned based on Gaussian process provided by the invention
Figure;
Fig. 5 is a kind of height cluster of two-way pedestrian counting method of laser returned based on Gaussian process provided by the invention
Flow chart;
Fig. 6 is a kind of pedestrian level of two-way pedestrian counting method of laser returned based on Gaussian process provided by the invention
Agglomerate figure;
Fig. 7,8 are a kind of test point of the two-way pedestrian counting method of laser returned based on Gaussian process provided by the invention
Pass in and out discriminating direction figure.
Specific implementation mode
Below in conjunction with attached drawing and specific embodiment, the present invention will be described in detail, herein illustrative examples of the invention
And explanation is used for explaining the present invention, but it is not as a limitation of the invention.
Embodiment:
A kind of two-way pedestrian counting method of laser returned based on Gaussian process, is included the following steps:
Range data is converted to altitude information by laser range finder detection layback information after triangulo operation;It is logical
The operation with background model is crossed, the foreground height of test point is obtained;Through accumulation after a period of time, with the foreground of each test point
Height generates sequential height map;It is detected a cluster in sequential height map, extracts pedestrian level agglomerate;Choose height agglomerate
Geometry and shape feature, choose Gaussian process homing method, calculate pedestrian level and roll into a ball pedestrian's quantity in the block;
Using the sliding window of a dynamic size, pedestrian level agglomerate is applied in the projection of time shaft;Sliding window
The size of mouth is codetermined by the width and pedestrian's regression result of pedestrian level agglomerate;The operation that sliding window executes includes calculating
The mean value and variance of detection point height in window;By calculating the maximum point of Mean curve, the head zone of pedestrian is positioned;It is complete
After being positioned at pedestrian head, using voting method, judge pedestrian into outgoing direction.
Fig. 1 is the flow of this method.Method is divided into three parts, 11-13.
Step 11, equipment is disposed in the way of Fig. 2 and Fig. 3.Wherein, the visual angle of equipment be δ, test point i with detection in
The angle of the heart is θ, measured distance D, and device inclined angle is, deployment height H.By acquisition testing point distance, be converted to
The height of test point, height calculation formula are
In conjunction with background height, the foreground height of test point is converted to.The computational methods of foreground height are foreground [i]
=(| h [i]-background [i] |<threshold))background[i]:d[i].In order to avoid shopping cart or other non-
Foreground height is set as 0 by influence of pedestrian's article to counting if foreground height is less than threshold value.
By the accumulation of a period of time, the sequential height map (such as Fig. 4) of test point is constituted.Realize that the often row of height map is every
The height of a test point is successively arranged according to sequential.
Step 12 is that height clusters, and detailed process is as follows:
Step 41, sequential height map is traversed;
Step 42, judge whether the height of current detection point is more than threshold value;
Step 43, after finding height more than the test point of threshold value, whether the test point that decision height is more than threshold value belongs to
Some height agglomerates, from this test point, start breadth first traversal algorithm if being not belonging to known altitude agglomerate,
Height agglomerate is added in the test point that all adjacent height are more than to threshold value.
Step 13, selection height rolls into a ball feature in the block, and pedestrian in the block is rolled into a ball using Gaussian process homing method computed altitude
Quantity.
It is geometric properties to select test point quantity in height agglomerate.
Select the Zhou Changwei geometric properties of height agglomerate.The perimeter of height agglomerate is decided to be the quantity of agglomerate boundary point.Side
Boundary's point back of the body is defined asWherein, C indicates a height agglomerate, N4(x) point x is indicated
Four neighborhoods.
Select the width of the boundary rectangle of height agglomerate for shape feature.
Select the shape feature of a height of agglomerate of the boundary rectangle of height agglomerate.
Select the depth-width ratio of the boundary rectangle of height agglomerate for the shape feature of agglomerate.
Select the circularity of height agglomerate for the shape feature of agglomerate.Circularity is defined as
Using the geometric characteristic of height agglomerate, using Gaussian process homing method, decision height rolls into a ball pedestrian in the block
Quantity.
With reference to《Chinese adult human dimension》(GB 10000-1988), is approximately 1 by human head and shoulder ratio:2.By height
The wide W and pedestrian quantity N of agglomerate are spent, a sliding window is set, size is
By height agglomerate along time-projection, then sliding window is acted in projection.Test point in calculation window
Average height and variance.Utilize maximum point positioning pedestrian head region.
By height agglomerate along time-projection, that is, take the maximum height of each test point as this test point in projection
Highly;
The width of height agglomerate is w, and pedestrian's quantity is N, and the size that sliding window is arranged is
There are two types of the operations of sliding window:Ask the mean value and variance of the detection point height in sliding window;
After the Mean curve for obtaining the test point of sliding window, by finding out the maximum point in Mean curve, home row
The head zone of people.
Judge the disengaging walking direction of each test point in head zone as a result, judgment method is:In computed altitude agglomerate
The time t that the maximum height of test point i occurstop, the height of test point i is calculated for the first time more than the time t of height thresholdf, meter
The height last time for calculating test point i is more than the time t of height thresholdl, compare ttop-tfWith tl-ttopMagnitude relationship, judge
The voting results of test point i:
By counting each test point voting results in height agglomerate, compare the number into and out of poll, judge pedestrian into
Outgoing direction enables
Then the direction of pedestrian can be judged with following formula:
Wherein, n is equal to the width that height rolls into a ball boundary rectangle in the block.
Said program combination pedestrian passes through the test point height change rule (Fig. 7 and Fig. 8) of detection zone in different directions,
The voting results for providing test point in sliding window judge pedestrian into outgoing direction.It can be seen from Fig. 7 that pedestrian is separate backwards to equipment
When (going out), test point maximum height appears in the front half section of height map;As can be seen from Figure 8, when pedestrian's equipment oriented traveling (entering), inspection
Measuring point maximum height appears in the second half section of height map.As a result, using highly roll into a ball each test point in the block be pedestrian disengaging
Direction is voted, and is finally compared into outgoing direction poll, you can judge pedestrian into outgoing direction.
It is provided for the embodiments of the invention technical solution above to be described in detail, specific case used herein
The principle and embodiment of the embodiment of the present invention are expounded, the explanation of above example is only applicable to help to understand this
The principle of inventive embodiments;Meanwhile for those of ordinary skill in the art, embodiment according to the present invention, in specific embodiment party
There will be changes in formula and application range, in conclusion the content of the present specification should not be construed as limiting the invention.
Claims (7)
1. a kind of two-way pedestrian counting method of laser returned based on Gaussian process, it is characterised in that include the following steps:
Range data is converted to altitude information by laser range finder detection layback information after triangulo operation;By with
The operation of background model obtains the foreground height of test point;Through accumulation after a period of time, with the foreground height of each test point
Generate sequential height map;It is detected a cluster in sequential height map, extracts pedestrian level agglomerate;Choose the several of height agglomerate
What and shape feature choose Gaussian process homing method, calculate pedestrian level and roll into a ball pedestrian's quantity in the block;
Using the sliding window of a dynamic size, pedestrian level agglomerate is applied in the projection of time shaft;Sliding window
Size is codetermined by the width and pedestrian's regression result of pedestrian level agglomerate;The operation that sliding window executes includes calculation window
The mean value and variance of interior detection point height;By calculating the maximum point of Mean curve, the head zone of pedestrian is positioned;Complete row
After head part's positioning, using voting method, judge pedestrian into outgoing direction.
2. the two-way pedestrian counting method of laser returned as described in claim 1 based on Gaussian process, it is characterised in that:
After the range information for obtaining laser range finder feedback, the visual angle with equipment is δ, the angle of test point i and inspection center
For θ, measured distance D, device inclined angle isDeployment height is H, by acquisition testing point distance, is converted to test point
Highly, height calculation formula is
3. the two-way pedestrian counting method of laser returned as described in claim 1 based on Gaussian process, it is characterised in that:
Detection point height is subjected to operation with background height model, the foreground height of test point is obtained, if in freshly harvested height
In degrees of data, the height of test point i is less than threshold value with the difference of background height, then it is high to be equal to background for the foreground height of test point i
Degree;If in freshly harvested altitude information, the height of test point i is more than threshold value with the difference of background height, then will newly acquire
Altitude information in the height of test point i be set as the foreground height of test point i, if foreground height is less than threshold value, will before
Scape height is set to 0.
4. the two-way pedestrian counting method of laser returned as claimed in claim 3 based on Gaussian process, it is characterised in that:
After obtaining the new height of test point every time, highly it regard these as row vector, is added in the matrix of sequential height map, detection
The height order of addition of point is exactly the acquisition sequence of altitude information, the height of the foreground detection point by accumulating a period of time, shape
At sequential height map.
5. the two-way pedestrian counting method of laser returned as claimed in claim 4 based on Gaussian process, it is characterised in that:
Sequential height map is traversed, judges whether the height of current detection point is more than threshold value, finds the test point that height is more than threshold value
Afterwards, whether decision height belongs to existing height agglomerate more than the test point of threshold value, if being not belonging to known altitude agglomerate, from
This test point sets out, and using breadth first traversal algorithm, height is added in the test point that all adjacent height are more than to threshold values
After completing traversal sequential height map, height cluster is completed to extract pedestrian level agglomerate for agglomerate.
6. the two-way pedestrian counting method of laser returned as claimed in claim 5 based on Gaussian process, it is characterised in that:
Choose the geometry and shape feature of height agglomerate:
The maximum height of height agglomerate is defined as the maximum height of test point in agglomerate;
The area of height agglomerate is defined as test point quantity in agglomerate;
The perimeter of height agglomerate, is defined as the border detection point quantity of agglomerate, and border detection point is defined as:
Wherein, x is border detection point, and C is a height agglomerate, N4(x) effect is to find out four neighborhoods of test point x;
The width of height agglomerate is defined as the width of agglomerate boundary rectangle;
The height of height agglomerate is defined as the height of agglomerate external world rectangle;
The length-width ratio of height agglomerate is defined as the ratio between the Gao Yukuan of agglomerate boundary rectangle;
The circularity of height agglomerate, is defined as the nearly circle degree of agglomerate, and calculation formula isWherein, S is height agglomerate
Area, P refers to the perimeter of agglomerate;
The shape complexity of height agglomerate, is defined as the complexity of agglomerate, and calculation formula is
Wherein, S is the area of height agglomerate, and P refers to the perimeter of agglomerate;
Using Gaussian process homing method, pedestrian counting is carried out.
7. the two-way pedestrian counting method of laser returned as claimed in claim 5 based on Gaussian process, it is characterised in that:
By height agglomerate along time-projection, that is, take height of the maximum height of each test point as this test point in projection
Degree;
The width of height agglomerate is w, and pedestrian's quantity is N, and the size that sliding window is arranged is
There are two types of the operations of sliding window:Ask the mean value and variance of the detection point height in sliding window;
After the Mean curve for obtaining the test point of sliding window, by finding out the maximum point in Mean curve, position pedestrian's
Head zone;
Judge the disengaging walking direction of each test point in head zone as a result, judgment method is:It is detected in computed altitude agglomerate
The time t that the maximum height of point i occurstop, the height of test point i is calculated for the first time more than the time t of height thresholdf, calculate inspection
The height last time of measuring point i is more than the time t of height thresholdl, compare ttop-tfWith tl-ttopMagnitude relationship, judge to detect
The voting results of point i:
By counting each test point voting results in height agglomerate, compares the number into and out of poll, judge the disengaging side of pedestrian
To order
Then the direction of pedestrian can be judged with following formula:
Wherein, n is equal to the width of height agglomerate boundary rectangle.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510240482.2A CN105046316B (en) | 2015-05-12 | 2015-05-12 | A kind of two-way pedestrian counting method of laser returned based on Gaussian process |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510240482.2A CN105046316B (en) | 2015-05-12 | 2015-05-12 | A kind of two-way pedestrian counting method of laser returned based on Gaussian process |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105046316A CN105046316A (en) | 2015-11-11 |
CN105046316B true CN105046316B (en) | 2018-08-03 |
Family
ID=54452845
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510240482.2A Active CN105046316B (en) | 2015-05-12 | 2015-05-12 | A kind of two-way pedestrian counting method of laser returned based on Gaussian process |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105046316B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105678268B (en) * | 2016-01-11 | 2020-06-30 | 华东理工大学 | Subway station scene pedestrian counting implementation method based on double-region learning |
CN106846297A (en) * | 2016-12-21 | 2017-06-13 | 深圳市镭神智能系统有限公司 | Pedestrian's flow quantity detecting system and method based on laser radar |
US10070259B1 (en) | 2017-02-24 | 2018-09-04 | Here Global B.V. | Altitude map for indoor positioning services |
CN110806588A (en) * | 2019-10-17 | 2020-02-18 | 北醒(北京)光子科技有限公司 | Pedestrian flow detection system based on laser radar |
CN110929636A (en) * | 2019-11-20 | 2020-03-27 | 上海融军实业有限公司 | Passenger flow size and direction detection method and system |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102622578A (en) * | 2012-02-06 | 2012-08-01 | 中山大学 | Passenger counting system and passenger counting method |
CN102772211A (en) * | 2012-08-08 | 2012-11-14 | 中山大学 | Human movement state detection system and detection method |
CN102930287A (en) * | 2012-09-26 | 2013-02-13 | 上海理工大学 | Overlook-based detection and counting system and method for pedestrians |
CN104112305A (en) * | 2013-10-17 | 2014-10-22 | 北京竞业达数码科技有限公司 | Passenger flow monitoring method and apparatus based on dual laser ranging |
-
2015
- 2015-05-12 CN CN201510240482.2A patent/CN105046316B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102622578A (en) * | 2012-02-06 | 2012-08-01 | 中山大学 | Passenger counting system and passenger counting method |
CN102772211A (en) * | 2012-08-08 | 2012-11-14 | 中山大学 | Human movement state detection system and detection method |
CN102930287A (en) * | 2012-09-26 | 2013-02-13 | 上海理工大学 | Overlook-based detection and counting system and method for pedestrians |
CN104112305A (en) * | 2013-10-17 | 2014-10-22 | 北京竞业达数码科技有限公司 | Passenger flow monitoring method and apparatus based on dual laser ranging |
Also Published As
Publication number | Publication date |
---|---|
CN105046316A (en) | 2015-11-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105046316B (en) | A kind of two-way pedestrian counting method of laser returned based on Gaussian process | |
JP6549797B2 (en) | Method and system for identifying head of passerby | |
CN103473571B (en) | Human detection method | |
CN105139425B (en) | A kind of demographic method and device | |
CN105940430B (en) | Personnel's method of counting and its device | |
CN103824070B (en) | A kind of rapid pedestrian detection method based on computer vision | |
CN105631852B (en) | Indoor human body detection method based on depth image contour | |
CN106372570A (en) | Visitor flowrate statistic method | |
CN105550670A (en) | Target object dynamic tracking and measurement positioning method | |
CN103735269B (en) | A kind of height measurement method followed the tracks of based on video multi-target | |
CN102750527A (en) | Long-time stable human face detection and tracking method in bank scene and long-time stable human face detection and tracking device in bank scene | |
CN102982334B (en) | The sparse disparities acquisition methods of based target edge feature and grey similarity | |
CN102592288B (en) | Method for matching pursuit of pedestrian target under illumination environment change condition | |
CN103150572A (en) | On-line type visual tracking method | |
CN104376334B (en) | A kind of pedestrian comparison method of multi-scale feature fusion | |
CN106682641A (en) | Pedestrian identification method based on image with FHOG- LBPH feature | |
CN106875421A (en) | A kind of multi-object tracking method and device | |
CN105160649A (en) | Multi-target tracking method and system based on kernel function unsupervised clustering | |
CN104517095A (en) | Head division method based on depth image | |
CN102496016A (en) | Infrared target detection method based on space-time cooperation framework | |
CN105279769A (en) | Hierarchical particle filtering tracking method combined with multiple features | |
CN106327536A (en) | Neck line measuring method based on sectional point clouds | |
CN106295657A (en) | A kind of method extracting human height's feature during video data structure | |
CN105741326B (en) | A kind of method for tracking target of the video sequence based on Cluster-Fusion | |
CN105913464A (en) | Multi-body target online measurement method based on videos |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant |