CN107563347A - A kind of passenger flow counting method and apparatus based on TOF camera - Google Patents
A kind of passenger flow counting method and apparatus based on TOF camera Download PDFInfo
- Publication number
- CN107563347A CN107563347A CN201710850191.4A CN201710850191A CN107563347A CN 107563347 A CN107563347 A CN 107563347A CN 201710850191 A CN201710850191 A CN 201710850191A CN 107563347 A CN107563347 A CN 107563347A
- Authority
- CN
- China
- Prior art keywords
- passenger
- image
- head
- tracking
- tof camera
- 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
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
- Closed-Circuit Television Systems (AREA)
Abstract
The invention discloses a kind of passenger flow counting method and apparatus based on TOF camera, including step 1, obtain depth image；Step 2, by height threshold, selective mechanisms region, a width mask image is obtained；Step 3, to mask morphological image process；Step 4, the number of people obtained according to mask figure in coloured image detects candidate region, and step 5, deep neural network model positions head of passenger in candidate region, and step 6, KCF combination Kalman filterings are tracked, step 7, passenger flow counting.The present invention solves the problems, such as to lack pattern recognition function using only depth image, the inactive area in image is filtered out by depth image, so as to accelerate detection speed and Detection accuracy, in summary, the present invention can reach the purpose of fast and accurately passenger flow counting.
Description
Technical field
The present invention relates to computer vision field, more particularly to a kind of passenger flow counting method and apparatus based on TOF camera.
Background technology
Traditional bus passenger flow method of counting, including infrared facility and pressure sensor, these method statistic results are missed
Difference is larger, is gradually abandoned.The direction of present bus passenger flow method of counting is developed toward image processing field, based on common
Colored or black white image carries out bus passenger flow statistics, and general with deep neural network, advantage is, knows due to combining pattern
Not, this method can handle the complex scenes such as crowded, the luggage knapsack of more people, and shortcoming is that computation complexity is high and pattern-recognition is lost
More meters and leakage meter can be caused by losing；Bus passenger flow statistics is carried out based on depth image, advantage is that computation complexity is low, due to depth
Degree information is not influenceed by light, detects one or simple scenario, and accuracy rate is high, the disadvantage is that, because depth image is lost too much
Texture information, so as to be identified to the direct use pattern of depth image, easier meter and leakage meter under complex scene, and have
Depth camera is to detect depth, surface smooth object according to infrared external reflection（Such as the hair or mobile phone of people）Can be to infrared production
Raw mirror-reflection, can not measure distance, so as to influence count results.
The content of the invention
For the deficiency of the above method, it is an object of the invention to provide a kind of passenger flow counting method based on TOF camera and
Device, this method carry out passenger flow counting, the party using the depth image of TOF camera and the recognition methods of 2D coloured image binding patterns
The accuracy rate of method is high, computation complexity is low, calculating speed is fast.
In order to solve the above technical problems, the present invention takes following technical scheme：A kind of passenger flow counting side based on TOF camera
Method, it the described method comprises the following steps:
（1）Obtain depth image：TOF camera is arranged on to the surface of bus front/rear door, shooting direction is perpendicular to bus
Railway carriage or compartment bottom surface, and switch for vehicle gate signal is accessed, switch on power, when bus door is opened, TOF camera shoots in-car depth map
Picture, and coloured image is obtained by the CMOS modules in camera；
（2）The pixel for being less than height threshold H in depth image is all filtered, its value is set to 0, height H above pixel retains,
And its value is set to 255, so obtain a width mask image；
（3）Morphological scale-space, including corrosion and expansive working are carried out to mask image；
（4）Contours extract operation is carried out to mask image, and boundary rectangle is taken out to each profile, it is colored corresponding to depth image
Image-region corresponding to circumscribed rectangular region is referred to as detection candidate region in image；
（5）Using deep neural network head of passenger detection model, each piece of detection candidate region of coloured image is examined
Survey, orient the position of head of passenger, the head of passenger detection model is the mould gone out by deep neural network Algorithm for Training
Type, training sample include it is a number of by by camera be arranged on bus front/rear door surface, shooting direction perpendicular to
Public transport compartment bottom surface, shoot resulting colour picture；
（6）To each head of passenger detected, it is tracked, for current color frame, if using KCF track algorithms
Near the position traced into, deep neural network detection model does not detect the head of passenger, then using KCF tracking result
As the position of present frame head of passenger, if near the position traced into using KCF track algorithms, deep neural network inspection
Model inspection is surveyed to the head of passenger, then using the position obtained by KCF track algorithms as predicted value, position that model inspection arrives
As observed value, the position of present frame head of passenger is determined using Kalman filtering algorithm combination predicted value and observed value；
（7）Statistics of getting on or off the bus is carried out, realizes passenger flow counting, during tracking, for a head of passenger target, if even
There is not observed value in continuous x frames, then abandon the tracking of this target；The head of passenger terminated for tracking, if its duration
More than y frames, the position offset of beginning and end exceedes threshold value set in advance, then forms an effective counting；According to
The position of beginning and end, determine that this passenger gets on the bus or got off, realize passenger flow counting.
Present invention also offers a kind of passenger flow counting device based on TOF camera, described device includes what is be sequentially connected electrically
Image collecting device, image processing apparatus, target locating set and count tracking device, it is characterised in that：
Described image harvester is TOF camera, and TOF camera is arranged on to the surface of bus front/rear door, and shooting direction is hung down
Directly in public transport compartment bottom surface, and switch for vehicle gate signal is accessed, switched on power, when bus door is opened, TOF camera shooting car
Interior depth image, and coloured image is obtained by the CMOS modules in camera；
Described image processing unit, one end electrically connect with image collecting device, and one end electrically connects with target locating set；The figure
As the height screening module, morphological images processing module, detection candidate region that processing unit includes being sequentially connected electrically screen mould
Block；
The height screening module, the pixel in depth image highly less than threshold value H is all filtered, and its value is set into 0, height
More than H pixels retain, and its value is set into 255, obtain a width mask image；
The Morphological scale-space module, the mask image that the height screening module is obtained carry out Morphological scale-space, including corruption
Erosion processing and expansion process；
The detection candidate region screening module, the number of people detection candidate region in coloured image corresponding to depth image is extracted,
Mask image after Morphological scale-space resume module is carried out contours extract operation by the detection candidate region screening module first,
And boundary rectangle is gone out to each contours extract, then obtain colored corresponding to the depth image that described image harvester is got
Image, region corresponding to boundary rectangle in coloured image, referred to as detects candidate region；
The target locating set, one end electrically connect with image processing apparatus, and one end electrically connects with count tracking device；Using depth
Degree neutral net head of passenger detection model detects to each detection candidate region in coloured image, orients passenger
Head position；The head of passenger detection model show that training sample includes a fixed number by deep neural network Algorithm for Training
The surface by the way that camera to be arranged on to bus front/rear door of amount, shooting direction is perpendicular to public transport compartment bottom surface, shooting gained
The colour picture arrived；
The count tracking device, one end electrically connect with target locating set, including tracking module and counting module；The tracking
Module is tracked, for current color picture frame, if using KCF track algorithms to each head of passenger detected
Near the position traced into, deep neural network detection model does not detect the head of passenger, then using KCF tracking result
Position as present frame head of passenger；If near the position traced into using KCF track algorithms, deep neural network inspection
Model inspection is surveyed to the head of passenger, then using the position obtained by KCF track algorithms as predicted value, position that model inspection arrives
As observed value, the position of present frame head of passenger is determined using Kalman filtering algorithm combination predicted value and observed value；
The counting module, judge getting on the bus and getting off for passenger, realize passenger flow counting, during tracking, for one
Head of passenger target, if observed value do not occur in continuous x frames, abandon the tracking of this target；The passenger terminated for tracking
Head, if its duration exceed threshold value set in advance more than y frames, the position offset of beginning and end, then form
One effective counting；According to the position of beginning and end, determine that this passenger gets on the bus or got off, finally realize passenger flow
Tally function.
The present invention has following technique effect and advantage：
1st, solve the problems, such as to lack mode capabilities using only depth image：Because depth image is not common gray level image,
It lost too many texture information, so as to be identified to the direct use pattern of depth image, and the methods of use image segmentation
Depth image is directly handled to carry out passenger flow statisticses, it is not high to count accuracy rate.
2nd, inactive area in image has been filtered out by depth image, has greatly reduced the regional extent for needing to detect, have
Effect alleviates the problem of deep neural network detection speed is slow：The computation complexity of deep neural network detection algorithm is high, directly
Highly inappropriate region and other inactive areas are filtered out, not only the relative flase drop for avoiding deep neural network model but also quickening
Detection speed.
3rd, asking for deep neural network leakage meter that may be present is alleviated by KCF combination Kalman filter tracking algorithms
Topic, and KCF calculating speeds are fast, but the shortcomings that lack dimensional variation, and after being combined with Kalman filtering algorithm, track algorithm can be with
Accomplish both quick and accurate.
Brief description of the drawings
Fig. 1 is the inventive method schematic flow sheet.
Fig. 2 is the schematic diagram of apparatus of the present invention.
Fig. 3 is the depth image that TOF depth cameras obtain.
Fig. 4 is coloured image corresponding to depth image that the CMOS modules of TOF depth cameras are got.
Fig. 5 is the mask figure after the screening of the inventive method height threshold.
Fig. 6 is the mask figure after the inventive method Morphological scale-space.
Fig. 7 is boundary rectangle of the inventive method to each contours extract in the mask image after processing.
Fig. 8 is coloured image corresponding to circumscribed rectangular region of the present invention.
Fig. 9 is the testing result figure of the inventive method.
Embodiment
Technical scheme is described in detail below in conjunction with the accompanying drawings.
Such as Fig. 1, a kind of passenger flow counting method based on TOF camera of the present invention, comprise the following steps：
Step 1, depth image is obtained：TOF camera is arranged on to the surface of bus front/rear door, shooting direction is perpendicular to public transport
Compartment bottom surface, and switch for vehicle gate signal is accessed, switch on power, when bus door is opened, TOF camera shoots in-car depth map
Picture, and coloured image is obtained by the CMOS modules in camera.Preferably, the depth image resolution ratio is 320*240.
Step 2, by height threshold, region more than height threshold in depth image is extracted：
Photographed scene can be converted into deep image information by TOF camera, the value record of each pixel in obtained depth image
The relative distance of this pixel distance camera, passes through the actual distance measured during installation, can obtain each in depth image
Pixel is watched and calculated, the numerical value of each pixel in depth image arrives in 0 to the actual distance of camera for convenience
Between 255,0 represents infinity, 255 represent it is infinitely near；
Height threshold H is set, the pixel that H is less than in depth image is all filtered, its value is set to 0, height H above pixel is protected
Stay, and its value is set to 255, so obtain a width mask image, screen the time that head of passenger is there may be in depth image
Favored area, the mask image is referring to Fig. 5；
Preferably, the height threshold H is 1.3 meters.
Step 3, Morphological scale-space, including corrosion and expansive working are carried out to mask image, is concretely comprised the following steps：
Step 31, etching operation is carried out to image, centered on each pixel, calculates the minimum value in the windows of 5 * 5, formula
It is as follows：
；
Etching operation can remove the noise spot isolated in mask image, and the window sizes of 5 * 5 can change according to actual conditions；
Step 32, expansive working is carried out to image, centered on each pixel, calculates the maximum in the windows of 5 * 5, formula
It is as follows：
；
Expansive working is connected to element adjacent in mask image, and the window sizes of 5 * 5 can change according to actual conditions, form
Mask image after processing is referring to Fig. 6.
Step 4, the number of people detection candidate region after the screening in coloured image is extracted：Concretely comprise the following steps：
Step 41, contours extract operation is carried out to the mask image after step 3 Morphological scale-space, and each contours extract is gone out
Boundary rectangle, the mask image of boundary rectangle is decorated with referring to Fig. 7；
Step 42, cromogram corresponding to the depth image same time got by the CMOS modules of TOF depth cameras is obtained
Picture, the region of coloured image corresponding to the boundary rectangle obtained in step 41 are referred to as number of people detection candidate region, colored original graph
As referring to Fig. 4, comprising only the coloured image example of number of people detection candidate region referring to Fig. 8.
It should be noted that the number of people may be will be obviously not present in some rectangle frames, such as region very little obviously can not accommodate
The number of people, therefore can be excluded to will be obviously not present the rectangular area of the number of people in advance according to threshold value set in advance, it is also necessary to explanation
It is that the coloured image example that number of people detection candidate region is comprised only described in step 42 detects referring to Fig. 8, the number of people that comprises only
The example that the coloured image of candidate region is intended merely to be best understood from inventing and made, does not represent the present invention and includes this operation.
By step 2-4 operation, number of people detection candidate region can be tentatively extracted by depth image, it is so notable
Reduce the region that deep neural network model needs detect, not only dramatically speeded up detection speed, and reduce depth god
Through the potential false drop rate of network model.
Step 5, the deep neural network head of passenger detection model gone out using training in advance, to each in coloured image
Block detection candidate region is detected, and orients the position of head of passenger；
The deep neural network model that the head of passenger detection model comes out for training in advance, the head of passenger detection model
Training sample include it is a number of by by camera be arranged on bus front/rear door surface, shooting direction is perpendicular to public affairs
Compartment bottom surface is handed over, shoots resulting colour picture.For sample set, each head of passenger is manually marked out using rectangle frame
Position, can be trained by deep neural network algorithm and draw the head of passenger detection model.Preferably, the number of training sample
Measure as more than 10,000.
Step 6, to each head of passenger detected, it is tracked, concretely comprises the following steps：
Step 61, for current color picture frame, if near the position traced into using KCF track algorithms, depth nerve
Network detection model does not detect the head of passenger, then is used as the position of present frame head of passenger using KCF tracking result；
Step 62, if near the position traced into using KCF track algorithms, deep neural network detection model, which detects, to be multiplied
The head of visitor, then using the position obtained by KCF track algorithms as predicted value, the position that model inspection arrives makes as observed value
The position of present frame head of passenger is determined with Kalman filtering algorithm combination predicted value and observed value.
Because deep neural network there may be missing inspection in itself, above-mentioned steps 61 can solve the problems, such as missing inspection, in step 6
Using KCF carry out number of people tracking, be because KCF tracking velocity quickly, but KCF lack target scale estimation, to target chi
It is bad to spend the video tracking effect of significant changes, then in step 62, with reference to Kalman filtering algorithm, judges number of people position,
So the tracking of step 6 can be accomplished both quick and accurate, and compensate for deep neural network in step 5 and the problem of missing inspection be present.
Step 7, statistics of getting on or off the bus is carried out, realizes passenger flow counting：Concretely comprise the following steps：
Step 71, during tracking, for a head of passenger target, if observed value do not occur in continuous x frames, that is, connect
Continuous x frame deep neural networks detection model does not detect the head of passenger near the position of KCF track algorithms prediction,
Then abandon the tracking of this target；
Step 72, the head of passenger terminated for tracking, if its duration is more than y frames, the position skew of beginning and end
Amount exceedes threshold value set in advance, then forms an effective counting；
Step 73, according to the position of beginning and end, determine that this passenger gets on the bus or got off, finally realize passenger flow counting
Function.
The x and y are set according to video actual conditions, it is preferable that x 5, y 10.
Present invention additionally comprises a kind of passenger flow counting device based on TOF camera, as shown in Fig. 2 it includes electrical connection successively
Image collecting device, image processing apparatus, target locating set and count tracking device.
Described image harvester is TOF camera, and TOF camera is arranged on to the surface of bus front/rear door, shooting side
To perpendicular to public transport compartment bottom surface, and switch for vehicle gate signal is accessed, switched on power, when bus door is opened, TOF camera is clapped
In-car depth image is taken the photograph, and coloured image is obtained by the CMOS modules in camera.Preferably, the depth image resolution ratio is
320*240。
Described image processing unit, one end electrically connect with image collecting device, and one end electrically connects with target locating set.Institute
State height screening module, morphological images processing module, detection candidate region sieve that image processing apparatus includes being sequentially connected electrically
Modeling block.
The depth image that described image harvester collects, the value of each pixel of the image have recorded this pixel away from
From the relative distance of camera, pass through the actual distance measured during installation, each pixel can be obtained in depth image to camera
Actual distance, watch and calculate for convenience, the numerical value of each pixel in depth image is between 0 to 255, and 0 represents
Infinity, 255 represent it is infinitely near.
The height screening module, threshold value H will be highly less than in depth image（H size is set according to actual conditions
It is fixed）Pixel all filter, its value is set to 0, height H above pixel retains, and its value is set into 255, finally gives a width
It there may be the mask image of the candidate region of head of passenger.Preferably, the value of the height threshold H is 1.3 meters.
The Morphological scale-space module, the mask image that the height screening module is obtained carry out Morphological scale-space, bag
Corrosion treatment and expansion process are included, the corrosion treatment, centered on each pixel, calculates the minimum value in the windows of 5 * 5,
Formula is as follows：
;
The expansion process, centered on each pixel, the maximum in the windows of 5 * 5 is calculated, formula is as follows：
;
It should be known that the window sizes of 5 * 5 can change according to actual conditions, after finally giving a width Morphological scale-space
Mask image.
The detection candidate region screening module, extract the number of people detection candidate regions in coloured image corresponding to depth image
Mask image after Morphological scale-space resume module is carried out contours extract behaviour by domain, the detection candidate region screening module first
Make, and boundary rectangle is gone out to each contours extract, then obtain corresponding to the depth image that described image harvester is got
Coloured image, region corresponding to boundary rectangle is to detect candidate region in coloured image.
The target locating set, one end electrically connect with image processing apparatus, and one end electrically connects with count tracking device.Adopt
The deep neural network head of passenger detection model obtained with training in advance is to each detection candidate region in coloured image
Detected respectively, orient head of passenger（That is target）Position.The head of passenger detection model passes through depth nerve net
Network Algorithm for Training show that training sample includes a number of surface by the way that camera to be arranged on to bus front/rear door, claps
Direction is taken the photograph perpendicular to public transport compartment bottom surface, shoots resulting colour picture.For sample set, manually marked out using rectangle frame
The position of each head of passenger.Preferably, the quantity of training sample is more than 10,000.
The count tracking device, one end electrically connect with target locating set, including tracking module and counting module.It is described
Tracking module is tracked to each head of passenger detected, is specially：For current color picture frame, if used
Near the position that KCF track algorithms trace into, deep neural network detection model does not detect the head of passenger, then uses
Position of the KCF tracking result as present frame head of passenger；If near the position traced into using KCF track algorithms,
Deep neural network detection model detects the head of passenger, then using the position obtained by KCF track algorithms as predicted value, mould
Present frame passenger is determined using Kalman filtering algorithm combination predicted value and observed value in the position that type detects as observed value
The position on head.
The counting module, judge getting on the bus and getting off for passenger, realize passenger flow counting, be specially：In the process of tracking
In, for a head of passenger target, if there is not observed value in continuous x frames, i.e., continuous x frames deep neural network detection mould
Type does not detect the head of passenger near the position of KCF track algorithms prediction, then abandons the tracking of this target；For
The head of passenger terminated is tracked, if its duration, more than y frames, the position offset of beginning and end is more than set in advance
Threshold value, then form an effective counting；According to the position of beginning and end, determine this passenger get on the bus or under
Car, finally realize passenger flow counting function.The x and y are set according to video actual conditions, it is preferable that x 5, y 10.
A kind of passenger flow counting method and apparatus based on TOF camera of the present invention, have the following advantages that：
1st, solve the problems, such as to lack mode capabilities using only depth image：Because depth image is not common gray level image,
It lost too many texture information, so as to be identified to the direct use pattern of depth image, and the methods of use image segmentation
Depth image is directly handled to carry out passenger flow statisticses, it is not high to count accuracy rate.
2nd, inactive area in image has been filtered out by depth image, has greatly reduced the regional extent for needing to detect, have
Effect alleviates the problem of deep neural network detection speed is slow：The computation complexity of deep neural network detection algorithm is high, directly
Highly inappropriate region and other inactive areas are filtered out, not only the relative flase drop for avoiding deep neural network model but also quickening
Detection speed.
3rd, asking for deep neural network leakage meter that may be present is alleviated by KCF combination Kalman filter tracking algorithms
Topic, and KCF calculating speeds are fast, but the shortcomings that lack dimensional variation, and after being combined with Kalman filtering algorithm, track algorithm can be with
Accomplish both quick and accurate.
Claims (7)
- A kind of 1. passenger flow counting method based on TOF camera, it is characterised in that comprise the following steps：（1）Obtain depth image：TOF camera is arranged on to the surface of bus front/rear door, shooting direction is perpendicular to bus Railway carriage or compartment bottom surface, and switch for vehicle gate signal is accessed, switch on power, when bus door is opened, TOF camera shoots in-car depth map Picture, and coloured image is obtained by the CMOS modules in camera；（2）The pixel for being less than height threshold H in depth image is all filtered, its value is set to 0, height H above pixel retains, And its value is set to 255, so obtain a width mask image；（3）Morphological scale-space, including corrosion and expansive working are carried out to mask image；（4）Contours extract operation is carried out to mask image, and boundary rectangle is taken out to each profile, it is colored corresponding to depth image Image-region corresponding to circumscribed rectangular region is referred to as detection candidate region in image；（5）Using deep neural network head of passenger detection model, each piece of detection candidate region of coloured image is examined Survey, orient the position of head of passenger, the head of passenger detection model is the mould gone out by deep neural network Algorithm for Training Type, training sample include it is a number of by by camera be arranged on bus front/rear door surface, shooting direction perpendicular to Public transport compartment bottom surface, shoot resulting colour picture；（6）To each head of passenger detected, it is tracked, for current color frame, if using KCF track algorithms Near the position traced into, deep neural network detection model does not detect the head of passenger, then using KCF tracking result As the position of present frame head of passenger, if near the position traced into using KCF track algorithms, deep neural network inspection Model inspection is surveyed to the head of passenger, then using the position obtained by KCF track algorithms as predicted value, position that model inspection arrives As observed value, the position of present frame head of passenger is determined using Kalman filtering algorithm combination predicted value and observed value；（7）Statistics of getting on or off the bus is carried out, realizes passenger flow counting, during tracking, for a head of passenger target, if even There is not observed value in continuous x frames, then abandon the tracking of this target；The head of passenger terminated for tracking, if its duration More than y frames, the position offset of beginning and end exceedes threshold value set in advance, then forms an effective counting；According to The position of beginning and end, determine that this passenger gets on the bus or got off, realize passenger flow counting.
- 2. a kind of passenger flow counting method based on TOF camera as claimed in claim 1, the etching operation, with each picture Centered on element, the minimum value in the windows of 5 * 5 is calculated, formula is：；The expansive working, centered on each pixel, the maximum in the windows of 5* 5 is calculated, formula is：。
- 3. a kind of passenger flow counting method based on TOF camera as claimed in claim 1, the x is 5, and the y is 10.
- 4. a kind of passenger flow counting method based on TOF camera as described in claim any one of 1-3, the step（4）In it is right The boundary rectangle that each profile takes out is screened first according to rectangle size.
- 5. a kind of passenger flow counting device based on TOF camera, including image collecting device, the image procossing dress being sequentially connected electrically Put, target locating set and count tracking device, it is characterised in that：Described image harvester is TOF camera, and TOF camera is arranged on to the surface of bus front/rear door, and shooting direction is hung down Directly in public transport compartment bottom surface, and switch for vehicle gate signal is accessed, switched on power, when bus door is opened, TOF camera shooting car Interior depth image, and coloured image is obtained by the CMOS modules in camera；Described image processing unit, one end electrically connect with image collecting device, and one end electrically connects with target locating set；The figure As the height screening module, morphological images processing module, detection candidate region that processing unit includes being sequentially connected electrically screen mould Block；The height screening module, the pixel in depth image highly less than threshold value H is all filtered, and its value is set into 0, height More than H pixels retain, and its value is set into 255, obtain a width mask image；The Morphological scale-space module, the mask image that the height screening module is obtained carry out Morphological scale-space, including corruption Erosion processing and expansion process；The detection candidate region screening module, the number of people detection candidate region in coloured image corresponding to depth image is extracted, Mask image after Morphological scale-space resume module is carried out contours extract operation by the detection candidate region screening module first, And boundary rectangle is gone out to each contours extract, then obtain colored corresponding to the depth image that described image harvester is got Image, region corresponding to boundary rectangle in coloured image, referred to as detects candidate region；The target locating set, one end electrically connect with image processing apparatus, and one end electrically connects with count tracking device；Using depth Degree neutral net head of passenger detection model detects to each detection candidate region in coloured image, orients passenger Head position；The head of passenger detection model show that training sample includes a fixed number by deep neural network Algorithm for Training The surface by the way that camera to be arranged on to bus front/rear door of amount, shooting direction is perpendicular to public transport compartment bottom surface, shooting gained The colour picture arrived；The count tracking device, one end electrically connect with target locating set, including tracking module and counting module；The tracking Module is tracked, for current color picture frame, if using KCF track algorithms to each head of passenger detected Near the position traced into, deep neural network detection model does not detect the head of passenger, then using KCF tracking result Position as present frame head of passenger；If near the position traced into using KCF track algorithms, deep neural network inspection Model inspection is surveyed to the head of passenger, then using the position obtained by KCF track algorithms as predicted value, position that model inspection arrives As observed value, the position of present frame head of passenger is determined using Kalman filtering algorithm combination predicted value and observed value；The counting module, judge getting on the bus and getting off for passenger, realize passenger flow counting, during tracking, for one Head of passenger target, if observed value do not occur in continuous x frames, abandon the tracking of this target；The passenger terminated for tracking Head, if its duration exceed threshold value set in advance more than y frames, the position offset of beginning and end, then form One effective counting；According to the position of beginning and end, determine that this passenger gets on the bus or got off, finally realize passenger flow Tally function.
- 6. a kind of passenger flow counting device based on TOF camera as claimed in claim 5, the etching operation, with each picture Centered on element, the minimum value in the windows of 5 * 5 is calculated, formula is：；The expansive working, centered on each pixel, the maximum in the windows of 5 * 5 is calculated, formula is：。
- 7. a kind of passenger flow counting device based on TOF camera as described in claim 5 or 6, the x is 5, and the y is 10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710850191.4A CN107563347B (en) | 2017-09-20 | 2017-09-20 | Passenger flow counting method and device based on TOF camera |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710850191.4A CN107563347B (en) | 2017-09-20 | 2017-09-20 | Passenger flow counting method and device based on TOF camera |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107563347A true CN107563347A (en) | 2018-01-09 |
| CN107563347B CN107563347B (en) | 2019-12-13 |
Family
ID=60981675
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710850191.4A Active CN107563347B (en) | 2017-09-20 | 2017-09-20 | Passenger flow counting method and device based on TOF camera |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107563347B (en) |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108279421A (en) * | 2018-01-28 | 2018-07-13 | 深圳新亮智能技术有限公司 | Time-of-flight camera with high-resolution colour picture |
| CN108509914A (en) * | 2018-04-03 | 2018-09-07 | 华录智达科技有限公司 | Bus passenger flow statistical analysis system based on TOF camera and method |
| CN109035841A (en) * | 2018-09-30 | 2018-12-18 | 上海交通大学 | Parking lot vehicle positioning system and method |
| CN109059797A (en) * | 2018-08-22 | 2018-12-21 | Oppo广东移动通信有限公司 | Flight time mould group and its control method, controller and electronic device |
| CN109285376A (en) * | 2018-08-09 | 2019-01-29 | 同济大学 | A kind of bus passenger flow statistical analysis system based on deep learning |
| CN109325963A (en) * | 2018-08-07 | 2019-02-12 | 长安大学 | A kind of bus passenger three-dimensional track classification method based on SVM |
| CN109448027A (en) * | 2018-10-19 | 2019-03-08 | 成都睿码科技有限责任公司 | A kind of adaptive, lasting motion estimate method based on algorithm fusion |
| CN109584297A (en) * | 2018-10-24 | 2019-04-05 | 北京升哲科技有限公司 | Object detection method and device |
| CN109615647A (en) * | 2018-10-24 | 2019-04-12 | 北京升哲科技有限公司 | Object detection method and device |
| CN110516602A (en) * | 2019-08-28 | 2019-11-29 | 杭州律橙电子科技有限公司 | A kind of public traffice passenger flow statistical method based on monocular camera and depth learning technology |
| CN110633671A (en) * | 2019-09-16 | 2019-12-31 | 天津通卡智能网络科技股份有限公司 | Bus passenger flow real-time statistical method based on depth image |
| CN110717408A (en) * | 2019-09-20 | 2020-01-21 | 台州智必安科技有限责任公司 | People flow counting method based on TOF camera |
| CN110728227A (en) * | 2019-10-09 | 2020-01-24 | 北京百度网讯科技有限公司 | Image processing method and device |
| CN110837769A (en) * | 2019-08-13 | 2020-02-25 | 广州三木智能科技有限公司 | Embedded far infrared pedestrian detection method based on image processing and deep learning |
| CN111640135A (en) * | 2020-05-25 | 2020-09-08 | 台州智必安科技有限责任公司 | TOF camera pedestrian counting method based on hardware front end |
| CN111797661A (en) * | 2019-05-07 | 2020-10-20 | 王青雷 | Wireless notification method based on big data analysis |
| CN112085767A (en) * | 2020-08-28 | 2020-12-15 | 安徽清新互联信息科技有限公司 | Passenger flow statistical method and system based on deep optical flow tracking |
| CN112417939A (en) * | 2019-08-21 | 2021-02-26 | 南京行者易智能交通科技有限公司 | Passenger flow OD data acquisition method and device based on image recognition, mobile terminal equipment, server and model training method |
| CN112767375A (en) * | 2021-01-27 | 2021-05-07 | 深圳技术大学 | OCT image classification method, system and equipment based on computer vision characteristics |
| CN113012136A (en) * | 2021-03-24 | 2021-06-22 | 中国民航大学 | Airport luggage counting method and counting system based on target detection |
| CN113034544A (en) * | 2021-03-19 | 2021-06-25 | 奥比中光科技集团股份有限公司 | People flow analysis method and device based on depth camera |
| CN113963318A (en) * | 2021-12-22 | 2022-01-21 | 北京的卢深视科技有限公司 | People flow statistical method and device, electronic equipment and storage medium |
| CN114066956A (en) * | 2020-07-27 | 2022-02-18 | 南京行者易智能交通科技有限公司 | Model training method, detection method and device for detecting opening and closing states of bus door and mobile terminal equipment |
| TWI813198B (en) * | 2021-03-25 | 2023-08-21 | 日商矢崎總業股份有限公司 | The system for counting the number of people getting on and off the bus, the method for counting the number of people getting on and off the bus, and the program for counting the number of people getting on and off the bus |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101777188A (en) * | 2010-03-12 | 2010-07-14 | 华中科技大学 | Real-time bus passenger flow volume statistical method |
| CN106548163A (en) * | 2016-11-25 | 2017-03-29 | 青岛大学 | Method based on TOF depth camera passenger flow countings |
| CN106886994A (en) * | 2017-02-08 | 2017-06-23 | 青岛大学 | A kind of flow of the people intelligent detection device and detection method based on depth camera |
-
2017
- 2017-09-20 CN CN201710850191.4A patent/CN107563347B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101777188A (en) * | 2010-03-12 | 2010-07-14 | 华中科技大学 | Real-time bus passenger flow volume statistical method |
| CN106548163A (en) * | 2016-11-25 | 2017-03-29 | 青岛大学 | Method based on TOF depth camera passenger flow countings |
| CN106886994A (en) * | 2017-02-08 | 2017-06-23 | 青岛大学 | A kind of flow of the people intelligent detection device and detection method based on depth camera |
Cited By (35)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108279421A (en) * | 2018-01-28 | 2018-07-13 | 深圳新亮智能技术有限公司 | Time-of-flight camera with high-resolution colour picture |
| CN108279421B (en) * | 2018-01-28 | 2021-09-28 | 深圳新亮智能技术有限公司 | Time-of-flight camera with high resolution color images |
| CN108509914A (en) * | 2018-04-03 | 2018-09-07 | 华录智达科技有限公司 | Bus passenger flow statistical analysis system based on TOF camera and method |
| CN109325963A (en) * | 2018-08-07 | 2019-02-12 | 长安大学 | A kind of bus passenger three-dimensional track classification method based on SVM |
| CN109325963B (en) * | 2018-08-07 | 2021-05-18 | 长安大学 | SVM-based three-dimensional trajectory classification method for bus passengers |
| CN109285376A (en) * | 2018-08-09 | 2019-01-29 | 同济大学 | A kind of bus passenger flow statistical analysis system based on deep learning |
| CN109285376B (en) * | 2018-08-09 | 2022-04-19 | 同济大学 | Bus passenger flow statistical analysis system based on deep learning |
| CN109059797B (en) * | 2018-08-22 | 2020-12-18 | Oppo广东移动通信有限公司 | Time-of-flight module, control method thereof, controller and electronic device |
| CN109059797A (en) * | 2018-08-22 | 2018-12-21 | Oppo广东移动通信有限公司 | Flight time mould group and its control method, controller and electronic device |
| CN109035841A (en) * | 2018-09-30 | 2018-12-18 | 上海交通大学 | Parking lot vehicle positioning system and method |
| CN109448027A (en) * | 2018-10-19 | 2019-03-08 | 成都睿码科技有限责任公司 | A kind of adaptive, lasting motion estimate method based on algorithm fusion |
| CN109448027B (en) * | 2018-10-19 | 2022-03-29 | 成都睿码科技有限责任公司 | Adaptive and persistent moving target identification method based on algorithm fusion |
| CN109615647A (en) * | 2018-10-24 | 2019-04-12 | 北京升哲科技有限公司 | Object detection method and device |
| CN109584297A (en) * | 2018-10-24 | 2019-04-05 | 北京升哲科技有限公司 | Object detection method and device |
| CN111797661B (en) * | 2019-05-07 | 2021-03-30 | 湖南琴海数码股份有限公司 | Wireless notification method based on big data analysis |
| CN111797661A (en) * | 2019-05-07 | 2020-10-20 | 王青雷 | Wireless notification method based on big data analysis |
| CN110837769B (en) * | 2019-08-13 | 2023-08-29 | 中山市三卓智能科技有限公司 | Image processing and deep learning embedded far infrared pedestrian detection method |
| CN110837769A (en) * | 2019-08-13 | 2020-02-25 | 广州三木智能科技有限公司 | Embedded far infrared pedestrian detection method based on image processing and deep learning |
| CN112417939A (en) * | 2019-08-21 | 2021-02-26 | 南京行者易智能交通科技有限公司 | Passenger flow OD data acquisition method and device based on image recognition, mobile terminal equipment, server and model training method |
| CN110516602A (en) * | 2019-08-28 | 2019-11-29 | 杭州律橙电子科技有限公司 | A kind of public traffice passenger flow statistical method based on monocular camera and depth learning technology |
| CN110633671A (en) * | 2019-09-16 | 2019-12-31 | 天津通卡智能网络科技股份有限公司 | Bus passenger flow real-time statistical method based on depth image |
| CN110717408A (en) * | 2019-09-20 | 2020-01-21 | 台州智必安科技有限责任公司 | People flow counting method based on TOF camera |
| CN110717408B (en) * | 2019-09-20 | 2022-04-12 | 台州智必安科技有限责任公司 | People flow counting method based on TOF camera |
| CN110728227A (en) * | 2019-10-09 | 2020-01-24 | 北京百度网讯科技有限公司 | Image processing method and device |
| CN111640135A (en) * | 2020-05-25 | 2020-09-08 | 台州智必安科技有限责任公司 | TOF camera pedestrian counting method based on hardware front end |
| CN114066956A (en) * | 2020-07-27 | 2022-02-18 | 南京行者易智能交通科技有限公司 | Model training method, detection method and device for detecting opening and closing states of bus door and mobile terminal equipment |
| CN114066956B (en) * | 2020-07-27 | 2024-07-12 | 南京行者易智能交通科技有限公司 | Model training method, detection method and device for detecting opening and closing states of bus doors and mobile terminal equipment |
| CN112085767A (en) * | 2020-08-28 | 2020-12-15 | 安徽清新互联信息科技有限公司 | Passenger flow statistical method and system based on deep optical flow tracking |
| CN112085767B (en) * | 2020-08-28 | 2023-04-18 | 安徽清新互联信息科技有限公司 | Passenger flow statistical method and system based on deep optical flow tracking |
| CN112767375A (en) * | 2021-01-27 | 2021-05-07 | 深圳技术大学 | OCT image classification method, system and equipment based on computer vision characteristics |
| CN112767375B (en) * | 2021-01-27 | 2022-03-08 | 深圳技术大学 | OCT image classification method, system and equipment based on computer vision characteristics |
| CN113034544A (en) * | 2021-03-19 | 2021-06-25 | 奥比中光科技集团股份有限公司 | People flow analysis method and device based on depth camera |
| CN113012136A (en) * | 2021-03-24 | 2021-06-22 | 中国民航大学 | Airport luggage counting method and counting system based on target detection |
| TWI813198B (en) * | 2021-03-25 | 2023-08-21 | 日商矢崎總業股份有限公司 | The system for counting the number of people getting on and off the bus, the method for counting the number of people getting on and off the bus, and the program for counting the number of people getting on and off the bus |
| CN113963318A (en) * | 2021-12-22 | 2022-01-21 | 北京的卢深视科技有限公司 | People flow statistical method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107563347B (en) | 2019-12-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107563347A (en) | A kind of passenger flow counting method and apparatus based on TOF camera | |
| CN102542289B (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
| CN102867416B (en) | Vehicle part feature-based vehicle detection and tracking method | |
| CN101739686B (en) | Moving object tracking method and system thereof | |
| CN106575473A (en) | Method and axle-counting device for contact-free axle counting of a vehicle and axle-counting system for road traffic | |
| CN113449606B (en) | Target object identification method and device, computer equipment and storage medium | |
| WO2008020598A1 (en) | Subject number detecting device and subject number detecting method | |
| CN113435336B (en) | Running intelligent timing system and method based on artificial intelligence | |
| CN110189375A (en) | A kind of images steganalysis method based on monocular vision measurement | |
| CN108022249A (en) | A kind of remote sensing video satellite moving vehicle target region of interest extraction method | |
| CN107798688A (en) | Motion estimate method, method for early warning and automobile anti-rear end collision prior-warning device | |
| Siricharoen et al. | Robust outdoor human segmentation based on color-based statistical approach and edge combination | |
| CN103077387A (en) | Method for automatically detecting carriage of freight train in video | |
| CN103021179A (en) | Real-time monitoring video based safety belt detection method | |
| Su et al. | A new local-main-gradient-orientation HOG and contour differences based algorithm for object classification | |
| CN104200455B (en) | A kind of key poses extracting method based on movement statistics signature analysis | |
| JP2002197588A (en) | Method for discriminating tire type of traveling vehicle and method and device for discriminating vehicle model | |
| CN112560557A (en) | People number detection method, face detection device and electronic equipment | |
| Xiang et al. | Vehicle detection and tracking for gas station surveillance based on AdaBoosting and optical flow | |
| Molinier et al. | 3D-connected components analysis for traffic monitoring in image sequences acquired from a helicopter | |
| CN110796035B (en) | People entering and exiting number counting method based on human shape detection and speed calculation | |
| Makawana et al. | Moving vehicle detection and speed measurement in video sequence | |
| Ó Conaire et al. | Fusion of infrared and visible spectrum video for indoor surveillance | |
| Yu et al. | A Video-Based Method for Traffic Flow Detection of Multi-lane Road | |
| Chaiyawatana et al. | Robust object detection on video surveillance |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |