CN106886994A - A kind of flow of the people intelligent detection device and detection method based on depth camera - Google Patents
A kind of flow of the people intelligent detection device and detection method based on depth camera Download PDFInfo
- Publication number
- CN106886994A CN106886994A CN201710069241.5A CN201710069241A CN106886994A CN 106886994 A CN106886994 A CN 106886994A CN 201710069241 A CN201710069241 A CN 201710069241A CN 106886994 A CN106886994 A CN 106886994A
- Authority
- CN
- China
- Prior art keywords
- people
- stream
- stable extremal
- flow
- circularity
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06M—COUNTING MECHANISMS; COUNTING OF OBJECTS NOT OTHERWISE PROVIDED FOR
- G06M1/00—Design features of general application
- G06M1/27—Design features of general application for representing the result of count in the form of electric signals, e.g. by sensing markings on the counter drum
- G06M1/272—Design features of general application for representing the result of count in the form of electric signals, e.g. by sensing markings on the counter drum using photoelectric means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30242—Counting objects in image
Abstract
Depth camera is based on the invention provides one kind, with reference to library's room for individual study flow of the people intelligent detecting method of computer vision algorithms make.The method can in real time, accurate statistics each rooms for individual study present number and room number, statistics is illustrated on the large display screen in library or library's wechat public number, each room for individual study is present number, room number is very clear.The classmate for facilitating Xiang Qu libraries to learn grasps situation in real time, it is to avoid lose time.Meanwhile, the present invention can lift the service quality in library, improve the effective rate of utilization of limited educational resource.
Description
Technical field
The invention belongs to computer vision, mode identification technology, and in particular to a kind of stream of people based on depth camera
Amount intelligent detection device and detection method.
Background technology
With the fast development of national higher education, University students number is sharply increased, and causes school's hardware resource shape
Condition is nervous, is a problem for student provides self-study classroom.And Most students pour in library and review one's lessons, there is queuing and account for seat,
One phenomenon hard to find, whenever appendix the stage more so.
The collected books of current Library in China, resource of catalog all realize digital management, have room for individual study seat resources only also
In the labor management stage.Some colleges and universities also independent research or have purchased seat Management System.But existing online booking system
System need to coordinate swiping card equipment to use, and quantities is big, safeguard cumbersome.If counted using all-in-one campus card, whenever open people in morning
During stream peak, student need to for a long time queue up and swipe the card, and easily occur replacing superior eigenvalue of swiping the card, without real-time and accurately
Rate is low.
The content of the invention
For the problem for overcoming existing online Ticket booking system to exist, the present invention provides a kind of based on depth camera, with reference to meter
Library's room for individual study flow of the people intelligent detecting method of calculation machine vision algorithm.The method can in real time, accurate statistics each rooms for individual study
Present number and room number, statistics is illustrated on the large display screen in library or library's wechat public number, is made each
Individual room for individual study is present number, and room number is very clear.The classmate for facilitating Xiang Qu libraries to learn grasps situation in real time, it is to avoid waste
Time.Meanwhile, the present invention can lift the service quality in library, improve the effective rate of utilization of limited educational resource.
The present invention provides a kind of flow of the people intelligent detection device, and it includes depth camera, working host, server, storage
Database in server, storage working host flow of the people intelligent checking system and display platform.
Wherein, the depth camera is arranged on library room for individual study porch, shoots vertically downward, by installed in work
Flow of the people intelligent checking system on main frame produces flow of the people data, and flow of the people data are sent into server data by network
Storehouse, the present number of each room for individual study, room number data Real-time Feedback is aobvious to wechat public number platform and library's giant-screen etc.
Show on platform to help reader.
Wherein, the stream of people's detection and track algorithm that the flow of the people intelligent checking system is used include:
The first step, stream of people's Target Segmentation;
Second step, stream of people's target detection;
3rd step, stream of people's target following and counting.
Wherein, the first step further specifically,
A is walked, and reads depth image;
B is walked, Morphological scale-space;
C is walked, and is extracted using MSER (Maximally Stable Extremal Regions) maximum stable extremal region
Stable extremal region.
Wherein, depth map, is converted to gray-scale map by the Morphological scale-space in b step, and Morphological scale-space is exactly by depth
The gray-scale map that the conversion of degree figure comes first corrodes and expands afterwards, then medium filtering again.
Wherein, be specially using MSER algorithms extraction feature point set in step c carries out stable extremal using MSER algorithms
Extracted region, then calculates the girth and area of stable extremal region.
Wherein, the target detection of the stream of people is specifically divided into training stage and detection-phase.
Wherein, the second step further specifically,
A is walked, and calculates the circularity of stable extremal region profile, and stabilization is obtained by area and the girth of stable extremal region
The circularity of extremal region, circularity formula is:
Wherein C is the circularity of stable extremal region, and A is the area of stable extremal region, and P is the week of stable extremal region
It is long.
B is walked, and is contrasted using training the circularity threshold value for obtaining and the circularity that stable extremal region is drawn, is being met
On the premise of circularity threshold value, stable extremal region is judged as the number of people, that is, complete stream of people's detection.
The second step further includes training stage and detection-phase.
Wherein, concretely comprised the following steps in the training stage,
1st step, prepares training sample;
2nd step, calculates the circularity of the sample number of people;
3rd step, determines the threshold value of number of people circularity;
4th step, stable extremal region is extracted using MSER.
Wherein, the detection-phase is concretely comprised the following steps,
1st ' step, calculates the circularity of stable extremal region;
2nd ' step, is contrasted, in symbol using training the circularity threshold value for obtaining and the circularity that stable extremal region is drawn
On the premise of closing circularity threshold value, stable extremal region center-of-mass coordinate is calculated, stable extremal region is judged as the number of people, you can real
Existing stream of people's detection;
3rd ' step, is contrasted, not using training the circularity threshold value for obtaining and the circularity that stable extremal region is drawn
On the premise of meeting circularity threshold value, the 1st ' step is re-started.
Stream of people's target following is specially with counting in 3rd step:
Human body head feature has consistency during motion, it is possible to which the motion of user the is nose heave heart replaces whole
The motion of the individual number of people.The present invention uses the center of minimum enclosed rectangle as the center of gravity of the number of people.Determine that the minimum of the number of people is outer first
Rectangle is connect, the central point of boundary rectangle is then obtained, now the central point of boundary rectangle is exactly the center of gravity of the number of people.Due to consecutive frame
Same number of people center of gravity it is closest, it is possible to according to the nearest principle of the same number of people center of gravity of consecutive frame carry out matching with
Track.
The change of same student's head position of centre of gravity in front and rear two field picture, judges whether student enters setting area.Once
Into the region, then according to the direction of motion of the stream of people, judge that student is to enter or walk out reading room, such that it is able to count this
Number of the period into and out of reading room.
Beneficial technique effect
The present invention provide flow of the people intelligent detecting method compared with traditional statistical method, with following advantage:
1. depth image is generated using depth camera, depth image is converted into the gray level image of layering, improved the number of people and know
Not other rate.
Compared with colour imagery shot, depth camera can obtain the three-dimensional data of scene, meanwhile, according to depth image not
Represented with different gray values with depth, the depth image that will be obtained is converted into the gray level image of different layers, so as to reduce
Head recognition difficulty, improves Head recognition precision.
2. number of people counting is carried out using double-counting line, it is to avoid be repeatedly detected the repeat count for causing
Generally, when crowded, easily occur same student when blocking and occur being repeatedly detected in cog region repeatedly, make
The problems such as into repeat count.Traditional list counting line mode is difficult to such issues that solve.Using one kind by two counting lines, three kinds of meters
Number state composition counting mode, can solve it is crowded, block, same personnel repeatedly be in cog region, repeatedly triggering meter
The problems such as number line produces repeat count.
Brief description of the drawings
Fig. 1 is wechat public number platform design sketch of the present invention;
Fig. 2 builds flow framework for the present inventor's flow quantity intelligent detecting system;
Fig. 3 is that the stream of people of the present invention detects and track algorithm framework;
Fig. 4 is the image that camera of the present invention is obtained at entry and exit, (a) original color image, (b) depth image;
Fig. 5 is Morphological scale-space of the present invention and median filter process original gradation figure, schemes (a) original-gray image, and (b) is rotten
Image after erosion and expansion process, the image after (c) median filter process;
Fig. 6 is the stable extremal region of reservation after MSER treatment of the present invention;
Fig. 7 is stream of people's target detection flow chart of the present invention;
Fig. 8 is the number of people image after the present invention is detected through the stream of people;
Fig. 9 is that the stream of people of the present invention detects double-counting line.
Specific embodiment
The present invention provides a kind of flow of the people intelligent detection device, and it includes depth camera, working host, server, storage
Database in server, storage working host flow of the people intelligent checking system and display platform.
The depth camera is arranged on library room for individual study porch, shoots vertically downward, by installed in working host
On flow of the people intelligent checking system produce flow of the people data, flow of the people data are sent to by server database by network,
The present number of each room for individual study, room number data Real-time Feedback is flat to the display such as wechat public number platform and library's giant-screen
Helping reader on platform.
Library leader can according to the data statistics of flow of the people, analyze reader's distribution situation of each time period, be
Formulate emergency measure and reference frame is provided in library.Meanwhile, accurately acquire in shop number, manager can be made to be closed according to flow of the people
The resource distributions such as reason adjustment books, reasonably optimizing librarian, security, cleaning, promote the raising of library works efficiency.
The stream of people's detection and track algorithm that the flow of the people intelligent checking system is used include:
The first step, stream of people's Target Segmentation;
Second step, stream of people's target detection;
3rd step, stream of people's target following.
The first step is further specially
A is walked, and reads depth image, specially for the environment in library, the figure that camera is obtained at entry and exit
Picture, is respectively coloured image and depth image;
B is walked, Morphological scale-space, specially using a depth threshold scope obtained according to experiment, preferably 0-3m,
To remove the ambient noise beyond human body, depth image is changed into gray level image again after treatment, expand afterwards using first being corroded and
The method of medium filtering carries out Morphological scale-space to gray level image;
C is walked, and is extracted using MSER (Maximally Stable Extremal Regions) maximum stable extremal region
Stable extremal region.
The Morphological scale-space in b step, the depth image that collection comes can not be used directly because depth bounds is excessive,
Need to choose suitable depth value and depth map is converted to gray-scale map.General bus gateway top is far from about 2.4 meters of ground
To between 2.8 meters, therefore depth value of the depth value within 0 meter to 2.8 meters is only chosen when depth image is converted to gray level image,
Remaining depth information is given up, and 0 to 2.7 meter of gray value is normalized to 0 to 255.On the one hand do so can eliminate a large amount of
Noise, on the other hand can refine depth value, the extraction of the human body head feature for being easy to be converted to after gray-scale map.At morphology
Reason is exactly that the gray-scale map come by depth map conversion is first corroded and expanded afterwards, then medium filtering again.Corrosion is in order to less than knot
The part removal of constitutive element, and it is then the cavity filled up after image procossing to expand.Morphological scale-space method can effectively remove figure
Burr as in.Human body contour outline edge is smoothened during medium filtering can make image.
Using MSER algorithms extracting feature point set in step c and being specially carries out stable extremal region and carries using MSER algorithms
Take, then calculate the girth and area of stable extremal region.
The method for asking for girth is a lot, but in view of the factor of real-time, stabilization is calculated using quick algorithm of convex hull
The approximate girth of extremal region profile, it is a point set that MSER extracts stable extremal region, and the size of point set is stable extremal
The area in region.The circularity of stable extremal region, circularity formula are obtained with girth by the area of stable extremal region:
Wherein C is the circularity of stable extremal region, and A is the area of stable extremal region, and P is the week of stable extremal region
It is long.
The top of doorway entry and exit is fixed on due to camera value, human body head can be as the stream of people in this case
The foundation of detection, the target detection of the stream of people is specifically divided into training stage and detection-phase.
The second step further specifically,
A is walked, and calculates the circularity of stable extremal region profile,
B is walked, and is contrasted using training the circularity threshold value for obtaining and the circularity that stable extremal region is drawn, is being met
On the premise of circularity threshold value, stable extremal region is judged as the number of people, that is, complete stream of people's detection.
The second step further includes training stage and detection-phase, in the training stage, collection height at 1.4 meters extremely
2.3 meters of head features of the stream of people, these features include the circularity of the standard number of people and the area of the standard number of people.By reality repeatedly
Test, it was demonstrated that the threshold value of standard number of people circularity is 0.65, and the circularity of the standard number of people is between 0.65 to 1.Detection-phase is exactly
Using circularity and the circularity threshold comparison of stable extremal region, regarded if circularity threshold value is met the stable extremal region as
The number of people.Same extreme value stability region in image can repeatedly be extracted when extracting stable extremal region using MSER, in order to remove weight
The multiple number of people for extracting, is carried out at duplicate removal using the close principle of the stable extremal region barycentric coodinates repeatedly extracted to the same number of people
Reason.
Concretely comprised the following steps in the training stage,
1st step, prepares training sample;
2nd step, calculates the circularity of the sample number of people;
3rd step, determines the threshold value of number of people circularity;
4th step, stable extremal region is extracted using MSER;
The detection-phase is concretely comprised the following steps,
1st ' step, calculates the circularity of stable extremal region;
2nd ' step, is contrasted, in symbol using training the circularity threshold value for obtaining and the circularity that stable extremal region is drawn
On the premise of closing circularity threshold value, stable extremal region center-of-mass coordinate is calculated, stable extremal region is judged as the number of people, you can enter
Pedestrian stream is detected;
3rd ' step, is contrasted, not using training the circularity threshold value for obtaining and the circularity that stable extremal region is drawn
On the premise of meeting circularity threshold value, the 1st ' step is re-started.
Motion tracking is carried out based on number of people center of gravity in 3rd step, specially delimit specific region enter pedestrian stream with
Track, specially sets a region, and the region is the region that must pass through when the stream of people gets on or off the bus, when the stream of people gets on or off the bus, human body head
Portion can be persistently detected as the number of people inside this region, for the stable extremal region of non-human head, even if in video
In some frames by false retrieval be the number of people, as human motion and cause the unstable of testing conditions, so as to exclude by false retrieval
May, complete the tracking of the stream of people according to the direction of motion of the center of gravity of the number of people in the region has been detected.
The human body head feature of the stream of people has consistency during motion, it is possible to the motion of user the is nose heave heart
Instead of the motion of the whole number of people.The method that tradition asks for object center of gravity is each pixel by traveling through stable extremal region
Value, and the barycentric coodinates of stable extremal region are calculated using center of gravity formula, although this method can obtain the center of gravity of standard,
But due to being related to the calculating of each pixel value in stable extremal region, influence the real-time of whole system.In order to quickly ask
Go out the center of gravity of stable extremal region, meet the requirement of real-time, herein using minimum enclosed rectangle center as the number of people weight
The heart.The minimum enclosed rectangle of the number of people is determined first, the central point of boundary rectangle is then obtained, and now the central point of boundary rectangle is just
It is the center of gravity of the number of people.It is closest due to the same number of people center of gravity of consecutive frame, it is possible to according to the same number of people center of gravity of consecutive frame
Nearest principle is matched and tracked.
Describe embodiments of the present invention in detail using embodiment and accompanying drawing below, how skill is applied to the present invention whereby
Art means solve technical problem, and reach the implementation process of technique effect and can fully understand and implement according to this.
Embodiment of the present invention is as follows:
1. flow of the people intelligent checking system builds framework
Video at library room for individual study entry and exit is gathered by multiple depth cameras, library works main frame is sent to, led to
The image processing program crossed on working host produces flow of the people data, and is preserved in real time to server data by network
Storehouse, by database data Real-time Feedback to library's giant-screen and wechat public number platform, as shown in Figure 1.System building frame
Frame is as shown in Figure 2.
2. the stream of people detects and track algorithm
The present invention proposes a kind of stream of people's detection based on depth image and track algorithm, and this algorithm is divided into stream of people's Target Segmentation
With stream of people object detecting and tracking two large divisions, algorithm frame is as shown in Figure 3.
For the environment in library, the image that camera is obtained at entry and exit, as shown in Figure 4.It is respectively coloured image
And depth image.Ambient noise beyond human body is removed using a depth threshold scope, after treatment depth image again
It is changed into gray level image.Use first to corrode to expand afterwards herein and Morphological scale-space is carried out to gray level image with the method for medium filtering, connect
And extract stable extremal region using MSER (Maximally Stable Extremal Regions) maximum stable extremal region
And the circularity of stable extremal region profile is calculated, the circularity threshold value and stable extremal region gained for then being obtained using training
The circularity contrast for going out, on the premise of circularity threshold value is met, stable extremal region is judged as the number of people, that is, complete number of people inspection
Survey.
After completing the segmentation and the detection of stream of people's number of people of the stream of people, the tracking that pedestrian stream is entered in specific region delimited.Set first
One region, the region is the region that must pass through when the stream of people gets on or off the bus, and when the stream of people gets on or off the bus, human body head is in this region
Face can be persistently detected as the number of people.For the stable extremal region of non-human head, even if by false retrieval in some frames of video
Be the number of people, as human motion and cause the unstable of testing conditions, so as to exclude by the possibility of false retrieval.According to being detected
The direction of motion of the center of gravity of the number of people in the region is surveyed to complete the tracking of the stream of people.
2.1 stream of people's Target Segmentations
2.1.1 Morphological scale-space
The depth image that collection comes can not be used directly, it is necessary to choose suitable depth value simultaneously because depth bounds is excessive
Depth map is converted to gray-scale map.General bus gateway top between about 2.4 meters to 2.8 meters of ground, therefore in depth map
As only choosing depth value of the depth value within 0 meter to 2.8 meters when being converted to gray level image, remaining depth information is given up, and 0
Gray value to 2.7 meters is normalized to 0 to 255.On the one hand do so can eliminate substantial amounts of noise, on the other hand can refine
Depth value, the extraction of the human body head feature for being easy to be converted to after gray-scale map.Morphological scale-space is exactly next by depth map conversion
Gray-scale map first corrode and expand afterwards, then medium filtering again.Corrosion is to the part removal less than structural element, and expand
It is then the cavity filled up after image procossing.Morphological scale-space method can effectively remove the burr in image.Medium filtering can be with
Human body contour outline edge is smoothened in making image, as a result as shown in Figure 5.
2.1.2 feature point set is extracted using MSER
Stable extremal region extraction, Ran Houji are carried out using Maximally Stable Extremal Regions algorithms
The girth and area of stable extremal region are calculated, the image of extraction is as shown in Figure 6.
The method for asking for girth is a lot, but in view of the factor of real-time, stabilization is calculated using quick algorithm of convex hull
The approximate girth of extremal region profile, it is a point set that MSER extracts stable extremal region, and the size of point set is stable extremal
The area in region.The circularity of stable extremal region, circularity formula are obtained with girth by the area of stable extremal region:
Wherein C is the circularity of stable extremal region, and A is the area of stable extremal region, and P is the week of stable extremal region
It is long.
2.2 stream of people's object detecting and trackings
2.2.1 the number of people based on number of people circularity is detected
Camera value is fixed on the top of doorway entry and exit, and human body head can be detected as the stream of people in this case
Foundation, the target detection of the stream of people is divided into training and two stages of detection, and flow is as shown in Figure 7.
In the training stage, collection height includes the standard number of people in 1.4 meters to the 2.3 meters head features of the stream of people, these features
Circularity and the standard number of people area.By experiment repeatedly, it was demonstrated that the threshold value of standard number of people circularity is 0.65, standard people
The circularity of head is between 0.65 to 1.Detection-phase is exactly the circularity and circularity threshold comparison for using stable extremal region,
It is the number of people that the stable extremal region is regarded if circularity threshold value is met.Can be in image during using MSER extraction stable extremal regions
Same extreme value stability region is repeatedly extracted, in order to remove the number of people of repetition extraction, using the stabilization repeatedly extracted to the same number of people
The close principle of extremal region barycentric coodinates carries out duplicate removal treatment, and Fig. 8 is the stable extremal region stayed after being detected through passerby.
2.2.2 the stream of people based on number of people center of gravity tracks
The human body head feature of the stream of people has consistency during motion, it is possible to the motion of user the is nose heave heart
Instead of the motion of the whole number of people.The method that tradition asks for object center of gravity is each pixel by traveling through stable extremal region
Value, and the barycentric coodinates of stable extremal region are calculated using center of gravity formula, although this method can obtain the center of gravity of standard,
But due to being related to the calculating of each pixel value in stable extremal region, influence the real-time of whole system.In order to quickly ask
Go out the center of gravity of stable extremal region, meet the requirement of real-time, herein using minimum enclosed rectangle center as the number of people weight
The heart.The minimum enclosed rectangle of the number of people is determined first, the central point of boundary rectangle is then obtained, and now the central point of boundary rectangle is just
It is the center of gravity of the number of people.It is closest due to the same number of people center of gravity of consecutive frame, it is possible to according to the same number of people center of gravity of consecutive frame
Nearest principle is matched and tracked.
All above-mentioned this intellectual properties of primarily implementation, the not this new product of implementation of setting limitation other forms
And/or new method.Those skilled in the art will be using this important information, the above modification, to realize similar execution feelings
Condition.But, all modifications or transformation are based on the right that new product of the present invention belongs to reservation.
The above, is only presently preferred embodiments of the present invention, is not the limitation for making other forms to the present invention, is appointed
What those skilled in the art changed possibly also with the technology contents of the disclosure above or be modified as equivalent variations etc.
Effect embodiment.But it is every without departing from technical solution of the present invention content, according to technical spirit of the invention to above example institute
Any simple modification, equivalent variations and the remodeling made, still fall within the protection domain of technical solution of the present invention.
Claims (10)
1. a kind of flow of the people intelligent detection device, it is characterised in that including:Depth camera, working host, server, it is stored in
The database of server, storage working host flow of the people intelligent checking system and display platform.
2. flow of the people intelligent detection device as claimed in claim 1, it is characterised in that:The depth camera is arranged on library
Room for individual study porch, is shot vertically downward, and flow of the people is produced by the flow of the people intelligent checking system on working host
Flow of the people data are sent to server database by data by network, and the present number of each room for individual study, room number data are real-time
Feed back on the display platform such as wechat public number platform and library's giant-screen to help reader.
3. flow of the people intelligent detection device as claimed in claim 1 or 2, it is characterised in that:The flow of the people Intelligent Measurement system
The used stream of people's detection of system and track algorithm include:
The first step, stream of people's Target Segmentation;
Second step, stream of people's target detection;
3rd step, stream of people's target following.
4. a kind of flow of the people intelligent checking system for any one of claims 1 to 3 the stream of people detection and track algorithm, its
Including:
The first step, stream of people's Target Segmentation;
Second step, stream of people's target detection;
3rd step, stream of people's target following.
5. the stream of people described in claim 4 detects and track algorithm, it is characterised in that:The first step further specifically,
A is walked, and reads depth image;
B is walked, Morphological scale-space;
C is walked, and stabilization is extracted using MSER (Maximally Stable Extremal Regions) maximum stable extremal region
Extremal region.
6. the stream of people described in claim 4 or 5 detects and track algorithm, it is characterised in that:The Morphological scale-space in b step,
Depth map is converted to gray-scale map, Morphological scale-space is exactly that the gray-scale map come by depth map conversion is first corroded to expand afterwards, then
Medium filtering again.
7. the stream of people described in claim 4 to 6 detects and track algorithm, it is characterised in that:Carried using MSER algorithms in step c
Taking feature point set and being specially carries out stable extremal region extraction using MSER algorithms, then calculate the girth of stable extremal region with
Area.
8. the stream of people described in claim 4 to 7 detects and track algorithm, it is characterised in that:The target detection of the stream of people is specifically divided into
Training stage and detection-phase.
9. the stream of people described in claim 4 to 8 detects and track algorithm, it is characterised in that:The second step further specifically,
A is walked, and calculates the circularity of stable extremal region profile,
B is walked, and is contrasted using training the circularity threshold value for obtaining and the circularity that stable extremal region is drawn, is meeting circle
On the premise of degree threshold value, stable extremal region is judged as the number of people, that is, complete stream of people's detection.
10. the stream of people described in claim 4 to 9 detects and track algorithm, it is characterised in that:
Wherein, concretely comprised the following steps in the training stage,
1st step, prepares training sample;
2nd step, calculates the circularity of the sample number of people;
3rd step, determines the threshold value of number of people circularity;
4th step, stable extremal region is extracted using MSER.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710069241.5A CN106886994A (en) | 2017-02-08 | 2017-02-08 | A kind of flow of the people intelligent detection device and detection method based on depth camera |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710069241.5A CN106886994A (en) | 2017-02-08 | 2017-02-08 | A kind of flow of the people intelligent detection device and detection method based on depth camera |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106886994A true CN106886994A (en) | 2017-06-23 |
Family
ID=59179260
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710069241.5A Pending CN106886994A (en) | 2017-02-08 | 2017-02-08 | A kind of flow of the people intelligent detection device and detection method based on depth camera |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106886994A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107563347A (en) * | 2017-09-20 | 2018-01-09 | 南京行者易智能交通科技有限公司 | A kind of passenger flow counting method and apparatus based on TOF camera |
CN108985265A (en) * | 2018-08-10 | 2018-12-11 | 南京华捷艾米软件科技有限公司 | Volume of the flow of passengers monitoring system and volume of the flow of passengers monitoring method |
CN109250616A (en) * | 2018-09-28 | 2019-01-22 | 绍兴市特种设备检测院 | A kind of staircase entrance congestion detection system and method |
CN109271847A (en) * | 2018-08-01 | 2019-01-25 | 阿里巴巴集团控股有限公司 | Method for detecting abnormality, device and equipment in unmanned clearing scene |
CN109446895A (en) * | 2018-09-18 | 2019-03-08 | 中国汽车技术研究中心有限公司 | A kind of pedestrian recognition method based on human body head feature |
CN110349217A (en) * | 2019-07-19 | 2019-10-18 | 四川长虹电器股份有限公司 | A kind of target candidate location estimation method and its device based on depth image |
CN110619285A (en) * | 2019-08-29 | 2019-12-27 | 福建天晴数码有限公司 | Human skeleton key point extracting method and computer readable storage medium |
CN112418064A (en) * | 2020-11-19 | 2021-02-26 | 上海交通大学 | Real-time automatic detection method for number of people in library reading room |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101727691A (en) * | 2009-12-10 | 2010-06-09 | 浙江工业大学 | Remote detection system for indoor visitor flow rate of school place |
US20130259306A1 (en) * | 2012-03-29 | 2013-10-03 | Hon Hai Precision Industry Co., Ltd. | Automatic revolving door and automatic revolving door control method |
CN105005764A (en) * | 2015-06-29 | 2015-10-28 | 东南大学 | Multi-direction text detection method of natural scene |
-
2017
- 2017-02-08 CN CN201710069241.5A patent/CN106886994A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101727691A (en) * | 2009-12-10 | 2010-06-09 | 浙江工业大学 | Remote detection system for indoor visitor flow rate of school place |
US20130259306A1 (en) * | 2012-03-29 | 2013-10-03 | Hon Hai Precision Industry Co., Ltd. | Automatic revolving door and automatic revolving door control method |
CN105005764A (en) * | 2015-06-29 | 2015-10-28 | 东南大学 | Multi-direction text detection method of natural scene |
Non-Patent Citations (3)
Title |
---|
柏森 等: "《信息隐藏算法及应用》", 31 July 2015, 国防工业出版社 * |
栾悉道 等: "《多媒体情报处理技术》", 31 May 2016, 国防工业出版社 * |
郭秀杰: "基于Kinect的人流量统计系统研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107563347A (en) * | 2017-09-20 | 2018-01-09 | 南京行者易智能交通科技有限公司 | A kind of passenger flow counting method and apparatus based on TOF camera |
CN107563347B (en) * | 2017-09-20 | 2019-12-13 | 南京行者易智能交通科技有限公司 | Passenger flow counting method and device based on TOF camera |
CN109271847A (en) * | 2018-08-01 | 2019-01-25 | 阿里巴巴集团控股有限公司 | Method for detecting abnormality, device and equipment in unmanned clearing scene |
US11132559B2 (en) | 2018-08-01 | 2021-09-28 | Advanced New Technologies Co., Ltd. | Abnormality detection method, apparatus, and device for unmanned checkout |
CN108985265A (en) * | 2018-08-10 | 2018-12-11 | 南京华捷艾米软件科技有限公司 | Volume of the flow of passengers monitoring system and volume of the flow of passengers monitoring method |
CN109446895A (en) * | 2018-09-18 | 2019-03-08 | 中国汽车技术研究中心有限公司 | A kind of pedestrian recognition method based on human body head feature |
CN109446895B (en) * | 2018-09-18 | 2022-04-08 | 中国汽车技术研究中心有限公司 | Pedestrian identification method based on human head features |
CN109250616A (en) * | 2018-09-28 | 2019-01-22 | 绍兴市特种设备检测院 | A kind of staircase entrance congestion detection system and method |
CN109250616B (en) * | 2018-09-28 | 2020-05-19 | 绍兴市特种设备检测院 | Escalator entrance and exit congestion detection system and method |
CN110349217A (en) * | 2019-07-19 | 2019-10-18 | 四川长虹电器股份有限公司 | A kind of target candidate location estimation method and its device based on depth image |
CN110619285A (en) * | 2019-08-29 | 2019-12-27 | 福建天晴数码有限公司 | Human skeleton key point extracting method and computer readable storage medium |
CN110619285B (en) * | 2019-08-29 | 2022-02-11 | 福建天晴数码有限公司 | Human skeleton key point extracting method and computer readable storage medium |
CN112418064A (en) * | 2020-11-19 | 2021-02-26 | 上海交通大学 | Real-time automatic detection method for number of people in library reading room |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106886994A (en) | A kind of flow of the people intelligent detection device and detection method based on depth camera | |
CN102542289B (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
CN104166841B (en) | The quick detection recognition methods of pedestrian or vehicle is specified in a kind of video surveillance network | |
CN110378259A (en) | A kind of multiple target Activity recognition method and system towards monitor video | |
CN109241871A (en) | A kind of public domain stream of people's tracking based on video data | |
CN107742099A (en) | A kind of crowd density estimation based on full convolutional network, the method for demographics | |
CN103310183B (en) | The method and apparatus of crowd massing detection | |
CN103632158B (en) | Forest fire prevention monitor method and forest fire prevention monitor system | |
CN103577875B (en) | A kind of area of computer aided CAD demographic method based on FAST | |
CN105844234A (en) | People counting method and device based on head shoulder detection | |
CN103310444B (en) | A kind of method of the monitoring people counting based on overhead camera head | |
CN105303191A (en) | Method and apparatus for counting pedestrians in foresight monitoring scene | |
CN105893946A (en) | Front face image detection method | |
CN104616006B (en) | A kind of beard method for detecting human face towards monitor video | |
CN104715238A (en) | Pedestrian detection method based on multi-feature fusion | |
CN110347870A (en) | The video frequency abstract generation method of view-based access control model conspicuousness detection and hierarchical clustering method | |
CN106127205A (en) | A kind of recognition methods of the digital instrument image being applicable to indoor track machine people | |
CN103390151B (en) | Method for detecting human face and device | |
CN107358163A (en) | Visitor's line trace statistical method, electronic equipment and storage medium based on recognition of face | |
CN106096553A (en) | A kind of pedestrian traffic statistical method based on multiple features | |
CN106709438A (en) | Method for collecting statistics of number of people based on video conference | |
Song et al. | Real-time monitoring for crowd counting using video surveillance and GIS | |
CN107767416A (en) | The recognition methods of pedestrian's direction in a kind of low-resolution image | |
CN109684986A (en) | A kind of vehicle analysis method and system based on automobile detecting following | |
CN107808524A (en) | A kind of intersection vehicle checking method based on unmanned plane |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170623 |