CN108563998A - Vivo identification model training method, biopsy method and device - Google Patents
Vivo identification model training method, biopsy method and device Download PDFInfo
- Publication number
- CN108563998A CN108563998A CN201810219006.6A CN201810219006A CN108563998A CN 108563998 A CN108563998 A CN 108563998A CN 201810219006 A CN201810219006 A CN 201810219006A CN 108563998 A CN108563998 A CN 108563998A
- Authority
- CN
- China
- Prior art keywords
- network model
- neural network
- positive sample
- shoulder feature
- shoulder
- 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
- 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
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06V40/40—Spoof detection, e.g. liveness detection
- G06V40/45—Detection of the body part being alive
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
A kind of vivo identification model training method of present invention offer, biopsy method and device, wherein the vivo identification model training method includes:The multiple images comprising human head and shoulder feature are obtained as positive sample, and the multiple images not comprising human head and shoulder feature are as negative sample;Neural network model is trained using the positive sample and negative sample, until the neural network model is more than predetermined threshold value to the recognition correct rate of human head and shoulder feature.
Description
Technical field
The present invention relates to field of image detection, and in particular to a kind of vivo identification model training method, biopsy method
And device.
Background technology
Currently, the video monitoring system of various scales is widely used to all trades and professions, in addition to public security, finance, bank, friendship
Further include community, office building, hotel, public place, factory, market, cell, even family outside the fields such as logical, army and port
Etc. occasions.
By being detected video image and judging, the content of image can be identified to realize automatic identification
Certain target or the purpose of behavior.Have many human body detecting methods at present, such as has the method detected based on conventional motion such as
Based on foreground detection and image block etc., the also method based on rote learning, such as the pedestrian based on " AdaBoost+Haar "
The methods of detection and " histograms of oriented gradients (HOG)+support vector machines (SVM) ".But these methods all exist in various degree
The problem of, such as method based on motion detection, foreground detection is highly susceptible to the influence of the environment such as illumination, causes unstable, goes out
The problems such as existing ghost, piecemeal, missing inspection;And the method for these rote learnings, place such as subway station, vapour due to crowd than comparatively dense
Station, railway station etc., it is easy to occur, the case where individual is blocked.And when individual occurs in the case of blocking, " AdaBoost+
The methods of Haar " and " HOG+SVM " can have different degrees of missing inspection, meanwhile, big scene is measured in personnel's appearance, is calculated
It spends into the case that magnitude rises, the problem of needing to ensure real-time.
Invention content
Therefore, the present invention is to solve the existing higher problem of human testing scheme omission factor.
In view of this, the present invention provides a kind of vivo identification model training methods, including:
The multiple images comprising human head and shoulder feature are obtained as positive sample, and not comprising the multiple of human head and shoulder feature
Image is as negative sample;
Neural network model is trained using the positive sample and negative sample, until the neural network model is to people
The recognition correct rate of body head and shoulder feature is more than predetermined threshold value.
Preferably, it in described the step of being trained to neural network model using the positive sample and negative sample, adopts
The neural network model is trained with converging channels characteristics algorithm.
Preferably, the positive sample includes different sexes, different age group, different postures and wears different garment
Human body image.
Preferably, the quantity of the positive sample is more than the quantity of the negative sample.
The present invention also provides a kind of biopsy methods, including:
It obtains by the collected video image to be identified of access control system;
Each frame image in the video image to be identified is carried out using the neural network model of above method training
Identification;
The human head and shoulder feature that marker recognition arrives.
Correspondingly, the present invention also provides a kind of vivo identification model training apparatus, including:
Acquiring unit, for obtaining the multiple images comprising human head and shoulder feature as positive sample, and not comprising human body
The multiple images of head and shoulder feature are as negative sample;
Training unit, for being trained to neural network model using the positive sample and negative sample, until the god
Predetermined threshold value is more than to the recognition correct rate of human head and shoulder feature through network model.
Preferably, the training unit is trained the neural network model using converging channels characteristics algorithm.
Preferably, the positive sample includes different sexes, different age group, different postures and wears different garment
Human body image.
Preferably, the quantity of the positive sample is more than the quantity of the negative sample.
Correspondingly, the present invention also provides a kind of living body detection devices, including:
Acquiring unit, for obtaining by the collected video image to be identified of access control system;
Recognition unit, for the neural network model using above method training to every in the video image to be identified
One frame image is identified;
Marking unit, the human head and shoulder feature arrived for marker recognition.
Vivo identification model training method and device provided by the invention, pass through positive sample and negative sample and head and shoulder feature
Neural network model is trained, the model trained is enable accurately to be recognised that from image comprising head and shoulder spy
Sign, the model can be detected, due to gate inhibition by identifying that head and shoulder feature whether there is live body to determine in image compared to whole body
Video camera in system is typically to be placed in the position of 45 degree of overhead, therefore comparatively the case where head and shoulder is blocked completely compares
Less appearance, thus the model can be very good to solve conventional pedestrian's detection the crowd is dense block when, be easy missing inspection
Problem.
Biopsy method and device provided by the invention, using the head and shoulder feature in head and shoulder identification model detection image,
Head and shoulder feature is marked, and regards it as human body, the human body recognized can subsequently be managed into line trace etc..Compared to complete
Body detects, because the video camera in access control system is typically to be placed in the position of 45 degree of overhead, therefore head and shoulder is blocked completely
The case where comparatively fewer appearance, therefore can be very good solve conventional pedestrian detection the crowd is dense block when,
The problem of being easy missing inspection.
Description of the drawings
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art are briefly described, it should be apparent that, in being described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, other drawings may also be obtained based on these drawings.
Fig. 1 is the flow chart of the vivo identification model training method in the embodiment of the present invention;
Fig. 2 is the flow chart of the biopsy method in the embodiment of the present invention;
Fig. 3 is the flow chart of the people flow rate statistical method in the embodiment of the present invention;
Fig. 4 is the structural schematic diagram of the vivo identification model training apparatus in the embodiment of the present invention;
Fig. 5 is the structural schematic diagram of the living body detection device in the embodiment of the present invention;
Fig. 6 is the structural schematic diagram of the people flow rate statistical device in the embodiment of the present invention.
Specific implementation mode
Technical scheme of the present invention is clearly and completely described below in conjunction with attached drawing, it is clear that described implementation
Example is a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill
The every other embodiment that personnel are obtained without making creative work, shall fall within the protection scope of the present invention.
First embodiment of the invention provides a kind of vivo identification model training method, includes the following steps as shown in Figure 1:
S11 obtains the multiple images comprising human head and shoulder feature as positive sample, and not comprising human head and shoulder feature
Multiple images are as negative sample, such as can choose 20000 positive samples and 10000 negative samples.In order to make the mould trained
Type performance is better, and the type of positive sample can more comprehensively, such as include men and women from character attribute, old man, child etc.,
Comprising front, side and back side etc. from posture, the sample in each season is contained from the angle of environment, and negative sample includes
The big picture 10000 without pedestrian in each classification is opened.
S12 is trained neural network model using the positive sample and negative sample, until the neural network model
Predetermined threshold value is more than to the recognition correct rate of human head and shoulder feature.Certain characteristic attributes in image can specifically be chosen as god
Input layer through network model is trained neural network model using the position of human head and shoulder in the picture as output layer.
Training algorithm includes a variety of, and the present embodiment preferably uses converging channels feature (ACF, Aggregate Channel Features)
Algorithm is trained the neural network model, extracts input of the feature in 10 channels of image as neural network model
Layer data.Reason mainly considers requirement of the algorithm to hardware memory, robustness, real-time, and ACF algorithms are meeting in real time
Property require while can have preferable robustness, it is more stable to the testing result of various environment.
Vivo identification model training method provided by the invention, by positive sample and negative sample and head and shoulder feature to nerve
Network model is trained, and so that the model trained is accurately recognised that from image comprising head and shoulder feature, the mould
Type can be detected, due in access control system by identifying that head and shoulder feature whether there is live body to determine in image compared to whole body
Video camera be typically be placed in the position of 45 degree of overhead, therefore the case where head and shoulder is blocked completely comparatively it is fewer go out
It is existing, thus the model can be very good to solve the problem of conventional pedestrian's detection the crowd is dense block when easy missing inspection.
Second embodiment of the invention provides a kind of biopsy method, as shown in Fig. 2, this method includes:
S21 is obtained by the collected video image to be identified of access control system;
S22, the neural network model trained using the method that above-mentioned first embodiment provides, to the video to be identified
Each frame image in image is identified;
S23, the human head and shoulder feature that marker recognition arrives.
The embodiment of the present invention utilizes the head and shoulder feature in head and shoulder identification model detection image, and head and shoulder feature is marked,
And human body is regarded it as, the human body recognized can subsequently be managed into line trace etc..It is detected compared to whole body, because of access control system
In video camera be typically to be placed in the position of 45 degree of overhead, therefore the case where head and shoulder is blocked completely is comparatively fewer
Occur, thus can be very good to solve the problem of conventional pedestrian's detection the crowd is dense block when easy missing inspection.
Third embodiment of the invention provides a kind of people flow rate statistical method, as shown in figure 3, this method includes:
S31 is obtained by the collected video image to be identified of access control system;
S32 is identified each frame image in the video image to be identified, to obtain human head and shoulder feature.
The model for the method training that above-mentioned first embodiment provides and the detection side that second embodiment provides may be used in this step
Image is identified in method.
S33 determines tracking target according to the human head and shoulder feature, namely extracts all targets of every frame image;
Whether S34 passes through precalculated position respectively to the tracking target into line trace with the determination tracking target, with
And the moving direction when tracking Target Traversing precalculated position, precalculated position is, for example, the bayonet in image, direction be into
Enter direction and departure direction;
S35 passes in and out the pre-determined bit according to the destination number for passing through the precalculated position and moving direction statistics
The number set.It is tracked since entering video monitoring range for each target, until target leaves monitoring area, by
This can count whithin a period of time, into and/or the number left.
According to people flow rate statistical method provided by the invention, this method includes target detection, target following and behavioural analysis
Three steps;The target detection step carries out head and shoulder detection to the sequence of frames of video of acquisition;The target following is to target
Detection as a result, carrying out track of the track record target in picture;The behavioural analysis is analyzed the track of target,
Judge the direction of motion of target, realizes judgement and the record of stream of people's inward and outward card channel.Through the above steps the present invention realize in real time,
In high precision, resist the effect blocked, there is very high practical value.
As a preferred embodiment, in above-mentioned steps S34, the present embodiment, which uses, is based on coring correlation filter
(KCF, Kernelized Correlation Filter) algorithm is into line trace, the reason of using KCF track algorithms, the fortune of KCF
Scanning frequency degree is very fast, while having certain robustness to the deformation of illumination and target.Quickening KCF in order to be walked simultaneously
The speed of service, target may be used fixed size search, this is because the stream of people statistics application scenarios target have with camera
The dimensional variation of certain distance, target is smaller.
As a preferred embodiment, above-mentioned steps S35 can specifically include following steps:
S351 is pre-positioned described in the tracking Target Traversing when precalculated position according to described in the tracking Target Traversing
Front and back two field pictures when setting judge whether the direction of displacement of the tracking target is predetermined direction.When the position of the tracking target
When shifting direction is predetermined direction, step S352, no to then follow the steps S53, the execution step S354 when needing to export result are executed;
S352 makes preset first counter values add 1;
S353 makes preset second counter values add 1;
S354 is determined according to the predetermined party respectively according to first counter values and second counter values
The precalculated position is passed through to the quantity for the pedestrian for passing through the precalculated position and according to the opposite direction of the predetermined direction
Pedestrian quantity.
Fourth embodiment of the invention provides a kind of vivo identification model training apparatus, as shown in figure 4, including:
Acquiring unit 41, for obtaining the multiple images comprising human head and shoulder feature as positive sample, and not comprising people
The multiple images of body head and shoulder feature are as negative sample;
Training unit 42, for being trained to neural network model using the positive sample and negative sample, until described
Neural network model is more than predetermined threshold value to the recognition correct rate of human head and shoulder feature.
According to vivo identification model training apparatus provided by the invention, pass through positive sample and negative sample and head and shoulder feature pair
Neural network model is trained, and the model trained is enable accurately to be recognised that from image comprising head and shoulder feature,
The model can be detected, since gate inhibition is by identifying that head and shoulder feature whether there is live body to determine in image compared to whole body
Video camera in system is typically to be placed in the position of 45 degree of overhead, therefore comparatively the case where head and shoulder is blocked completely compares
It is few to occur, thus the model can be very good to solve conventional pedestrian's detection the crowd is dense block when, be easy asking for missing inspection
Topic.
Preferably, the training unit is trained the neural network model using converging channels characteristics algorithm.
Preferably, the positive sample includes different sexes, different age group, different postures and wears different garment
Human body image.
Preferably, the quantity of the positive sample is more than the quantity of the negative sample.
Fifth embodiment of the invention provides a kind of living body detection device, as shown in figure 5, including:
Acquiring unit 51, for obtaining by the collected video image to be identified of access control system;
The neural network model of recognition unit 52, the method training for being provided using above-mentioned first embodiment is waited for described
Each frame image in identification video image is identified;
Marking unit 53, the human head and shoulder feature arrived for marker recognition.
The embodiment of the present invention utilizes the head and shoulder device in head and shoulder identification model detection image, and head and shoulder feature is marked,
And human body is regarded it as, the human body recognized can subsequently be managed into line trace etc..It is detected compared to whole body, because of access control system
In video camera be typically to be placed in the position of 45 degree of overhead, therefore the case where head and shoulder is blocked completely is comparatively fewer
Occur, thus can be very good to solve the problem of conventional pedestrian's detection the crowd is dense block when easy missing inspection.
Sixth embodiment of the invention provides a kind of people flow rate statistical device, as shown in fig. 6, including:
Acquiring unit 61, for obtaining by the collected video image to be identified of access control system;
Recognition unit 62, for each frame image in the video image to be identified to be identified, to obtain human body
Head and shoulder feature;
Determination unit 63, for determining tracking target according to the human head and shoulder feature;
Tracking cell 64, for respectively to the tracking target into line trace, whether being passed through with the determination tracking target
Moving direction when precalculated position and the tracking Target Traversing precalculated position;
Statistic unit 65, for according to the destination number for passing through the precalculated position and moving direction statistics disengaging
The number in the precalculated position.
According to people flow rate statistical device provided by the invention, head and shoulder detection is carried out by the sequence of frames of video to acquisition, and
Target to detecting records track of the target in picture, then analyzes the track of target, judge mesh into line trace
The target direction of motion realizes judgement and the record of stream of people's inward and outward card channel.By above-mentioned processing, the present invention realizes real-time, high-precision
Degree, the anti-effect blocked, have very high practical value.
Preferably, the tracking cell, which uses, is based on coring correlation filter algorithm into line trace.
Preferably, the statistic unit includes:
Walking direction unit is worn when for the precalculated position described in the tracking Target Traversing according to the tracking target
Front and back two field pictures when the precalculated position judge whether the direction of displacement of the tracking target is predetermined direction;
First direction counting unit, for when the direction of displacement of the tracking target is predetermined direction, making preset the
One counter values add 1;
Second direction counting unit, for when the direction of displacement of the tracking target is not predetermined direction, making preset
Second counter values add 1;
Twocouese statistic unit, for being determined respectively according to first counter values and second counter values
It the quantity for the pedestrian that the precalculated position is passed through according to the predetermined direction and is worn according to the opposite direction of the predetermined direction
The quantity of the pedestrian in the precalculated position.
Obviously, the above embodiments are merely examples for clarifying the description, and does not limit the embodiments.It is right
For those of ordinary skill in the art, can also make on the basis of the above description it is other it is various forms of variation or
It changes.There is no necessity and possibility to exhaust all the enbodiments.And it is extended from this it is obvious variation or
It changes still within the protection scope of the invention.
Claims (10)
1. a kind of vivo identification model training method, which is characterized in that including:
The multiple images comprising human head and shoulder feature are obtained as positive sample, and the multiple images not comprising human head and shoulder feature
As negative sample;
Neural network model is trained using the positive sample and negative sample, until the neural network model is to human body head
The recognition correct rate of shoulder feature is more than predetermined threshold value.
2. according to the method described in claim 1, it is characterized in that, utilizing the positive sample and negative sample to nerve net described
In the step of network model is trained, the neural network model is trained using converging channels characteristics algorithm.
3. method according to claim 1 or 2, which is characterized in that the positive sample includes different sexes, all ages and classes
Section, different postures and the human body image for wearing different garment.
4. method according to claim 1 or 2, which is characterized in that the quantity of the positive sample is more than the negative sample
Quantity.
5. a kind of biopsy method, which is characterized in that including:
It obtains by the collected video image to be identified of access control system;
Using the neural network model of the method training described in any one of claim 1-4 in the video image to be identified
Each frame image be identified;
The human head and shoulder feature that marker recognition arrives.
6. a kind of vivo identification model training apparatus, which is characterized in that including:
Acquiring unit, for obtaining the multiple images comprising human head and shoulder feature as positive sample, and not comprising human head and shoulder
The multiple images of feature are as negative sample;
Training unit, for being trained to neural network model using the positive sample and negative sample, until the nerve net
Network model is more than predetermined threshold value to the recognition correct rate of human head and shoulder feature.
7. device according to claim 6, which is characterized in that the training unit is using converging channels characteristics algorithm to institute
Neural network model is stated to be trained.
8. device according to claim 5 or 6, which is characterized in that the positive sample includes different sexes, all ages and classes
Section, different postures and the human body image for wearing different garment.
9. device according to claim 5 or 6, which is characterized in that the quantity of the positive sample is more than the negative sample
Quantity.
10. a kind of living body detection device, which is characterized in that including:
Acquiring unit, for obtaining by the collected video image to be identified of access control system;
Recognition unit, for being waited for described using the neural network model of the method training described in any one of claim 1-4
Each frame image in identification video image is identified;
Marking unit, the human head and shoulder feature arrived for marker recognition.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810219006.6A CN108563998A (en) | 2018-03-16 | 2018-03-16 | Vivo identification model training method, biopsy method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810219006.6A CN108563998A (en) | 2018-03-16 | 2018-03-16 | Vivo identification model training method, biopsy method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108563998A true CN108563998A (en) | 2018-09-21 |
Family
ID=63532794
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810219006.6A Pending CN108563998A (en) | 2018-03-16 | 2018-03-16 | Vivo identification model training method, biopsy method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108563998A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109902576A (en) * | 2019-01-25 | 2019-06-18 | 华中科技大学 | A kind of training method and application of head shoulder images classifier |
CN110186167A (en) * | 2019-05-31 | 2019-08-30 | 广东美的制冷设备有限公司 | Control method, device, air conditioner and the storage medium of air conditioner |
CN110569819A (en) * | 2019-09-16 | 2019-12-13 | 天津通卡智能网络科技股份有限公司 | Bus passenger re-identification method |
CN111553202A (en) * | 2020-04-08 | 2020-08-18 | 浙江大华技术股份有限公司 | Training method, detection method and device of neural network for detecting living body |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100278420A1 (en) * | 2009-04-02 | 2010-11-04 | Siemens Corporation | Predicate Logic based Image Grammars for Complex Visual Pattern Recognition |
CN104156983A (en) * | 2014-08-05 | 2014-11-19 | 天津大学 | Public transport passenger flow statistical method based on video image processing |
CN105844234A (en) * | 2016-03-21 | 2016-08-10 | 商汤集团有限公司 | People counting method and device based on head shoulder detection |
CN106803083A (en) * | 2017-02-04 | 2017-06-06 | 北京旷视科技有限公司 | The method and device of pedestrian detection |
CN106845406A (en) * | 2017-01-20 | 2017-06-13 | 深圳英飞拓科技股份有限公司 | Head and shoulder detection method and device based on multitask concatenated convolutional neutral net |
CN106874855A (en) * | 2017-01-19 | 2017-06-20 | 博康智能信息技术有限公司北京海淀分公司 | Head and shoulder area positioning method and device |
CN107330390A (en) * | 2017-06-26 | 2017-11-07 | 上海远洲核信软件科技股份有限公司 | A kind of demographic method based on graphical analysis and deep learning |
-
2018
- 2018-03-16 CN CN201810219006.6A patent/CN108563998A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100278420A1 (en) * | 2009-04-02 | 2010-11-04 | Siemens Corporation | Predicate Logic based Image Grammars for Complex Visual Pattern Recognition |
CN104156983A (en) * | 2014-08-05 | 2014-11-19 | 天津大学 | Public transport passenger flow statistical method based on video image processing |
CN105844234A (en) * | 2016-03-21 | 2016-08-10 | 商汤集团有限公司 | People counting method and device based on head shoulder detection |
CN106874855A (en) * | 2017-01-19 | 2017-06-20 | 博康智能信息技术有限公司北京海淀分公司 | Head and shoulder area positioning method and device |
CN106845406A (en) * | 2017-01-20 | 2017-06-13 | 深圳英飞拓科技股份有限公司 | Head and shoulder detection method and device based on multitask concatenated convolutional neutral net |
CN106803083A (en) * | 2017-02-04 | 2017-06-06 | 北京旷视科技有限公司 | The method and device of pedestrian detection |
CN107330390A (en) * | 2017-06-26 | 2017-11-07 | 上海远洲核信软件科技股份有限公司 | A kind of demographic method based on graphical analysis and deep learning |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109902576A (en) * | 2019-01-25 | 2019-06-18 | 华中科技大学 | A kind of training method and application of head shoulder images classifier |
CN110186167A (en) * | 2019-05-31 | 2019-08-30 | 广东美的制冷设备有限公司 | Control method, device, air conditioner and the storage medium of air conditioner |
CN110569819A (en) * | 2019-09-16 | 2019-12-13 | 天津通卡智能网络科技股份有限公司 | Bus passenger re-identification method |
CN111553202A (en) * | 2020-04-08 | 2020-08-18 | 浙江大华技术股份有限公司 | Training method, detection method and device of neural network for detecting living body |
CN111553202B (en) * | 2020-04-08 | 2023-05-16 | 浙江大华技术股份有限公司 | Training method, detection method and device for neural network for living body detection |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109522854A (en) | A kind of pedestrian traffic statistical method based on deep learning and multiple target tracking | |
CN108563998A (en) | Vivo identification model training method, biopsy method and device | |
CN102542289B (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
CN104091176B (en) | Portrait comparison application technology in video | |
CN109145708B (en) | Pedestrian flow statistical method based on RGB and D information fusion | |
CN109271554A (en) | A kind of intelligent video identifying system and its application | |
CN104036236B (en) | A kind of face gender identification method based on multiparameter exponential weighting | |
CN108765394A (en) | Target identification method based on quality evaluation | |
CN109800624A (en) | A kind of multi-object tracking method identified again based on pedestrian | |
TW202018594A (en) | People flow condition estimation method and device for designated area | |
CN110443210A (en) | A kind of pedestrian tracting method, device and terminal | |
CN108549846A (en) | A kind of movement character combined and head and shoulder structure pedestrian detection and statistical method | |
CN104966070A (en) | Face recognition based living body detection method and apparatus | |
Hashemzadeh et al. | Counting moving people in crowds using motion statistics of feature-points | |
CN109583373B (en) | Pedestrian re-identification implementation method | |
CN106709438A (en) | Method for collecting statistics of number of people based on video conference | |
CN104408406A (en) | Staff off-post detection method based on frame difference method and background subtraction method | |
CN103150552B (en) | A kind of driving training management method based on number of people counting | |
CN107480607A (en) | A kind of method that standing Face datection positions in intelligent recording and broadcasting system | |
CN111598132A (en) | Portrait recognition algorithm performance evaluation method and device | |
WO2021217764A1 (en) | Human face liveness detection method based on polarization imaging | |
CN107358163A (en) | Visitor's line trace statistical method, electronic equipment and storage medium based on recognition of face | |
CN109409250A (en) | A kind of across the video camera pedestrian of no overlap ken recognition methods again based on deep learning | |
CN108764338A (en) | A kind of pedestrian tracking algorithm applied to video analysis | |
Ardeshir et al. | Integrating egocentric videos in top-view surveillance videos: Joint identification and temporal alignment |
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 |
Application publication date: 20180921 |
|
RJ01 | Rejection of invention patent application after publication |