CN110110598A - The pedestrian of a kind of view-based access control model feature and space-time restriction recognition methods and system again - Google Patents
The pedestrian of a kind of view-based access control model feature and space-time restriction recognition methods and system again Download PDFInfo
- Publication number
- CN110110598A CN110110598A CN201910263596.7A CN201910263596A CN110110598A CN 110110598 A CN110110598 A CN 110110598A CN 201910263596 A CN201910263596 A CN 201910263596A CN 110110598 A CN110110598 A CN 110110598A
- Authority
- CN
- China
- Prior art keywords
- space
- pedestrian
- model
- time
- image
- 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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses pedestrian's recognition methods again of a kind of view-based access control model feature and space-time restriction, comprising the following steps: the First look similarity of retrieval image and pedestrian image data set is calculated using trained Image Classifier；Establish the space-time probability that space-time restriction model calculates retrieval image and pedestrian image data set；The First look similarity and the space-time probability are merged, Fusion Model is obtained；The second vision similarity of the retrieval image and the pedestrian image data set is calculated based on the Fusion Model, and is ranked up based on second vision similarity to obtain pedestrian's recognition result.It is also proposed that a kind of pedestrian identifying system again, comprising: vision computing module, space-time calculation module, Fusion Module and pedestrian's identification module.The beneficial effects of the present invention are: by conjunction with the First look similarity and the space-time probability, Fusion Model has great promotion than single vision mode in the accuracy of identification pedestrian image.
Description
Technical field
The present invention relates to image identification technical fields, are the rows of a kind of view-based access control model feature and space-time restriction specifically
People recognition methods and system again.
Background technique
In recent years, video monitoring has played great effect in public safety field.In monitor video, due to camera
The reason of resolution ratio and shooting angle is not typically available the very high face picture of quality.Therefore, when recognition of face failure
In the case of, pedestrian's weight identification technology has played very crucial effect in video monitoring.
But recognition methods is typically based on image itself to current pedestrian again, extracts visual signature, training image classification
Device, to target, pedestrian is matched.The image data that this training method needs largely to have label not only obtains such instruction
It is very time-consuming and laborious to practice data, and this Image Classifier is often performed poor in practical applications.
Summary of the invention
Technical problem to be solved by the invention is to provide the pedestrians of a kind of view-based access control model feature and space-time restriction to identify again
Method and system, to improve the efficiency and accuracy that pedestrian identifies again.
The technical scheme to solve the above technical problems is that
A kind of pedestrian's recognition methods again of view-based access control model feature and space-time restriction, comprising the following steps:
The First look similarity of retrieval image and pedestrian image data set is calculated using trained Image Classifier；
Establish the space-time probability that space-time restriction model calculates the retrieval image and the pedestrian image data set；
The First look similarity and the space-time probability are merged, Fusion Model is obtained；
The second vision similarity of the retrieval image and the pedestrian image data set is calculated based on the Fusion Model,
And it is ranked up based on second vision similarity to obtain pedestrian's recognition result.
The beneficial effects of the present invention are: by conjunction with the First look similarity and the space-time probability, i.e. pedestrian
Visual signature in conjunction with the space-time restriction for photographing the pedestrian in camera network, Fusion Model identification pedestrian image it is accurate
There is great promotion than single vision mode on degree.
Based on the above technical solution, the present invention can also be improved as follows.
Further, described to be ranked up based on second vision similarity to obtain pedestrian's recognition result, later also
Include:
Study sequence is carried out based on pedestrian's recognition result；
Study ranking results are returned to described image classifier, are classified based on the study ranking results to described image
Device is trained with the Image Classifier after being optimized；
The study ranking results are returned to the space-time restriction model, based on the study ranking results to it is described when
Empty restricted model is trained with the space-time restriction model after being optimized；
The Fusion Model is carried out based on the Image Classifier after the optimization and the space-time restriction model after optimization
Optimization, the Fusion Model after being optimized.
Beneficial effect using above-mentioned further scheme is: described by being sent to the ranking results of the Fusion Model
Image Classifier and the space-time restriction model are relearned, can be to original Image Classifier and space-time restriction
Model optimizes, and then research on optimizing information fusion, and Fusion Model is enabled to carry out more accurate identification to pedestrian image.
It is further, described that study sequence is carried out based on pedestrian's recognition result, comprising:
Pedestrian's recognition result is ranked up using List-wise method.
Beneficial effect using above-mentioned further scheme is: by being identified using the method for List-wise to the pedestrian
As a result it is sorted from high to low according to score, so that sequence better effect.
Further, described to establish space-time restriction model, it specifically includes:
The time difference probability distribution that the same pedestrian image that statistics training data is concentrated is taken by different cameras, base
The space-time restriction model is established in the probability distribution.
Beneficial effect using above-mentioned further scheme is: by being based on the time difference probability distribution, can train
One space-time restriction model, so that obtained Temporal And Spatial Distribution Model has better Generalization Capability.
Further, described to merge the First look similarity and the space-time probability, obtain fusion mould
Type specifically includes:
The First look similarity and the space-time probability are merged based on the algorithm of Boosting, it is described to melt
Molding type are as follows:
Based on Adaboost algorithm, the value of α and β is calculated, the Fusion Model is obtained；
Wherein,For First look similarity,For space-time probability.
Beneficial effect using above-mentioned further scheme is: by the algorithm fusion of Boosting by the First look phase
It is merged like degree and the space-time probability, the Fusion Model of available Rational Proportion, so that Fusion Model is more accurate.
Further, described that the first of retrieval image and pedestrian image data set is calculated using trained Image Classifier
Vision similarity, before further include:
Two classification Image Classifiers are trained using the data set with smooth label, obtain the trained figure
As classifier.
Beneficial effect using above-mentioned further scheme is: by utilizing the data set with smooth label to two classification charts
As classifier is trained, the Image Classifier of an available weak typing.
Meanwhile the present invention also proposes a kind of pedestrian of view-based access control model feature and space-time restriction identifying system again, comprising:
Vision computing module, for calculating retrieval image and pedestrian image data set according to trained Image Classifier
First look similarity；
Space-time calculation module calculates the retrieval image and the pedestrian image data set for establishing space-time restriction model
Space-time probability；
Fusion Module obtains fusion mould for merging the First look similarity and the space-time probability
Type；
Pedestrian's identification module, for calculating the retrieval image and the pedestrian image data set based on the Fusion Model
The second vision similarity, and be ranked up to obtain pedestrian's recognition result based on second vision similarity.
Further, the system also includes optimization module, the optimization module includes study sequencing unit, the first optimization
Unit, the second optimization unit and fusion optimization unit, in which:
The study sequencing unit is used to pedestrian's recognition result carrying out study sequence；
The first optimization unit is used to be trained to obtain described image classifier according to the study ranking results
Image Classifier after to optimization；
It is described second optimization unit be used for according to the study ranking results to the space-time restriction model be trained with
Space-time restriction model after being optimized；
The fusion optimization unit is used for based on the Image Classifier after the optimization and the space-time restriction mould after optimization
Type optimizes the Fusion Model, the Fusion Model after being optimized.
Further, the study sequencing unit includes sequence computing unit, and the first sequence computing unit is used for root
Pedestrian's recognition result is ranked up according to List-wise method.
Further, the space-time calculation module includes space-time model creating unit, and the space-time model creating unit is used
In the time difference probability distribution that the same pedestrian image that statistics training data is concentrated is taken by different cameras, and it is based on institute
It states probability distribution and establishes the space-time restriction model.
Detailed description of the invention
Fig. 1 is the logical schematic of pedestrian's recognition methods again of a kind of view-based access control model feature of the present invention and space-time restriction；
Fig. 2 is the structural schematic diagram of pedestrian's identifying system again of a kind of view-based access control model feature of the present invention and space-time restriction.
Specific embodiment
The principle and features of the present invention will be described below with reference to the accompanying drawings, and the given examples are served only to explain the present invention, and
It is non-to be used to limit the scope of the invention.
As shown in Figure 1, pedestrian's recognition methods again of a kind of view-based access control model feature and space-time restriction, comprising the following steps:
The First look similarity of retrieval image and pedestrian image data set is calculated using trained Image Classifier；
Establish the space-time probability that space-time restriction model calculates the retrieval image and the pedestrian image data set；
The First look similarity and the space-time probability are merged, Fusion Model is obtained；
The second vision similarity of the retrieval image and the pedestrian image data set is calculated based on the Fusion Model,
And it is ranked up based on second vision similarity to obtain pedestrian's recognition result.
It should be noted that described calculate retrieval image and pedestrian image data set using trained Image Classifier
First look similarity, including extracting the first pedestrian's feature for retrieving image and the pedestrian using ResNet50 network
Second pedestrian's feature of the image that image data is concentrated, then calculates the Europe of the first pedestrian feature and second pedestrian's feature
Formula distance obtains the First look similarity.Pass through the picture in given retrieval image and pedestrian image data set, institute
It states Image Classifier and exports the probability that this two picture includes the same pedestrian, then carried out according to the First look similarity
Sequence, the high part that will sort is as pedestrian's recognition result, the i.e. higher retrieval image of First look sequencing of similarity and pedestrian
The image of image construction in image data base to be exactly closer to be the same pedestrian image pair.Concrete operations are as follows: first exist
An identity mapping loss is added in CycleGAN, guarantees personages' information loss such as prospect such as color after converting
It is small as far as possible.Then data enhancing is carried out using improved CycleGAN, while is generated using some drawbacks that GAN makes figure
Noise, the data enhancement methods of cooperation overturning and selective erasing, avoids the generation of over-fitting to obtain better pedestrian's feature.
In addition, calculate the First look similarity of retrieval image and pedestrian image data set, calculate the retrieval image and
The space-time probability of the pedestrian image data set is the probability for calculating probe to gallery, and wherein probe is to be checked defeated
Enter, i.e. retrieval portrait；Gallery is candidate pedestrian library, i.e. pedestrian image data set.
Specifically, described to be ranked up based on second vision similarity to obtain pedestrian's recognition result, it also wraps later
It includes:
Study sequence is carried out based on pedestrian's recognition result；
Study ranking results are returned to described image classifier, are classified based on the study ranking results to described image
Device is trained with the Image Classifier after being optimized；
The study ranking results are returned to the space-time restriction model, based on the study ranking results to it is described when
Empty restricted model is trained with the space-time restriction model after being optimized；
The Fusion Model is carried out based on the Image Classifier after the optimization and the space-time restriction model after optimization
Optimization, the Fusion Model after being optimized.
It should be noted that carrying out study sequence to pedestrian's recognition result, i.e. LTR marks training set, after then sorting
Pedestrian's recognition result be fed for Image Classifier and space-time restriction model again to relearn, to Image Classifier
And space-time restriction model optimizes, and is then come by Image Classifier after optimization and space-time restriction model further excellent
Change Fusion Model, to obtain a powerful Fusion Model.Moreover, the step for can recycle repeatedly, make Fusion Model continuous
Optimization, so that obtaining a target scene identifies very powerful Fusion Model.
It is specifically, described that study sequence is carried out based on pedestrian's recognition result, comprising:
Pedestrian's recognition result is ranked up using List-wise method.
It should be noted that the present invention is preferably using List-wise method come to being ranked up, List-wise according to
Training examples training obtains optimal score function F, and corresponding new inquiry, scoring F gives a mark to each document, then according to score by
High to Low sequence, as final ranking results calculate comprehensive score by enumerating all arranging situations, and by highest scoring
A kind of result that sequence is sorted as study.
Specifically, described to establish space-time restriction model, it specifically includes:
The time difference probability distribution that the same pedestrian image that statistics training data is concentrated is taken by different cameras, base
The space-time restriction model is established in the probability distribution.
It should be noted that the training dataset is to obtain the data set of the space-time restriction model for training, lead to
The association between the shooting time and camera number for the monitoring image that statistics training data concentration belongs to the same pedestrian is crossed, it can
To obtain time difference probability distribution, it is then based on time difference probability and constructs a space-time restriction model.
In addition, it is necessary to which explanation, calculates the retrieval image and the pedestrian image data using space-time restriction model
The space-time probability of collection is in camera network, and pedestrian is at the minimum walking time between any video camera n to video camera m
TMinnm, then the time interval of the pedestrian i photographed at video camera 0 and the pedestrian j photographed at video camera m can be with is defined as:IfThe pedestrian i that then explanation takes at video camera 0 is clapped at video camera m
The pedestrian j taken the photograph is not same people.It therefore, further include building filter model, exclusion the case where for being zero by space-time probability,
To reduce the complexity of data query.
Moreover, working asWhen, the matching probability of space-time restriction model increases first, then reaches
Peak value, as time interval is elongated, matching probability is gradually reduced；Thus it can be assumed that the transit time between camera network
Wei Buer distribution is followed, so space-time probability can be with is defined as:
Wherein, k > 0 is form parameter, and λ > 0 is the scale parameter of distribution.
Specifically, described to merge the First look similarity and the space-time probability, Fusion Model is obtained,
It specifically includes:
The First look similarity and the space-time probability are merged based on the algorithm of Boosting, it is described to melt
Molding type are as follows:
Based on Adaboost algorithm, the value of α and β is calculated, the Fusion Model is obtained；
Wherein,For First look similarity,For space-time probability.
It should be noted that by the present invention in that carried out with the algorithm of Boosting the First look similarity and
The fusion of the space-time probability, i.e., merged by average weighted mode, then by Adaboost algorithm, calculates α's and β
Weight obtains the Fusion Model.
Specifically, first view that retrieval image and pedestrian image data set are calculated using trained Image Classifier
Feel similarity, before further include:
Two classification Image Classifiers are trained using the data set with smooth label, obtain the trained figure
As classifier.
Meanwhile as shown in Fig. 2, the present invention also proposes that the pedestrian of a kind of view-based access control model feature and space-time restriction identifies again is
System, comprising:
Vision computing module, for calculating retrieval image and pedestrian image data set according to trained Image Classifier
First look similarity；
Space-time calculation module calculates the retrieval image and the pedestrian image data set for establishing space-time restriction model
Space-time probability；
Fusion Module obtains fusion mould for merging the First look similarity and the space-time probability
Type；
Pedestrian's identification module, for calculating the retrieval image and the pedestrian image data set based on the Fusion Model
The second vision similarity, and be ranked up to obtain pedestrian's recognition result based on second vision similarity.
Specifically, the system also includes optimization module, the optimization module includes study sequencing unit, the first optimization list
Member, the second optimization unit and fusion optimization unit, in which:
The study sequencing unit is used to pedestrian's recognition result carrying out study sequence；
The first optimization unit is used to be trained to obtain described image classifier according to the study ranking results
Image Classifier after to optimization；
It is described second optimization unit be used for according to the study ranking results to the space-time restriction model be trained with
Space-time restriction model after being optimized；
The fusion optimization unit is used for based on the Image Classifier after the optimization and the space-time restriction mould after optimization
Type optimizes the Fusion Model, the Fusion Model after being optimized.
Specifically, the study sequencing unit includes sequence computing unit, and the first sequence computing unit is used for basis
List-wise method is ranked up pedestrian's recognition result.
Specifically, the space-time calculation module includes space-time model creating unit, and the space-time model creating unit is used for
The time difference probability distribution that the same pedestrian image that statistics training data is concentrated is taken by different cameras, and based on described
Probability distribution establishes the space-time restriction model.
Specifically, the Fusion Module includes integrated unit and computing unit, and the integrated unit is used for basis
The algorithm of Boosting merges the First look similarity and the space-time probability；
The Fusion Model computing unit is used to that the Fusion Model to be calculated according to Adaboost algorithm.
It specifically, further include training module, the training module is used for according to the data set with smooth label to two points
Class Image Classifier is trained, and obtains the trained Image Classifier.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (10)
1. pedestrian's recognition methods again of a kind of view-based access control model feature and space-time restriction, which comprises the following steps:
The First look similarity of retrieval image and pedestrian image data set is calculated using trained Image Classifier；
Establish the space-time probability that space-time restriction model calculates the retrieval image and the pedestrian image data set；
The First look similarity and the space-time probability are merged, Fusion Model is obtained；
The second vision similarity of the retrieval image and the pedestrian image data set, and base are calculated based on the Fusion Model
It is ranked up in second vision similarity to obtain pedestrian's recognition result.
2. pedestrian's recognition methods again of view-based access control model feature according to claim 1 and space-time restriction, which is characterized in that institute
It states and is ranked up to obtain pedestrian's recognition result, later based on second vision similarity further include:
Study sequence is carried out based on pedestrian's recognition result；
Will study ranking results be returned to described image classifier, based on the study ranking results to described image classifier into
Row training is with the Image Classifier after being optimized；
The study ranking results are returned to the space-time restriction model, based on the study ranking results to the space-time about
Beam model is trained with the space-time restriction model after being optimized；
The Fusion Model is optimized based on the Image Classifier after the optimization and the space-time restriction model after optimization,
Fusion Model after being optimized.
3. pedestrian's recognition methods again of view-based access control model feature according to claim 2 and space-time restriction, which is characterized in that institute
It states and study sequence is carried out based on pedestrian's recognition result, comprising:
Pedestrian's recognition result is ranked up using List-wise method.
4. pedestrian's recognition methods again of view-based access control model feature according to claim 1 and space-time restriction, which is characterized in that institute
It states and establishes space-time restriction model, specifically include:
The time difference probability distribution that the same pedestrian image that statistics training data is concentrated is taken by different cameras, is based on institute
It states probability distribution and establishes the space-time restriction model.
5. pedestrian's recognition methods again of view-based access control model feature according to claim 1 and space-time restriction, which is characterized in that institute
It states and merges the First look similarity and the space-time probability, obtain Fusion Model, specifically include:
The First look similarity and the space-time probability are merged based on the algorithm of Boosting, the fusion mould
Type are as follows:
Based on Adaboost algorithm, the value of α and β is calculated, the Fusion Model is obtained；
Wherein,For First look similarity,For space-time probability.
6. pedestrian's recognition methods again of view-based access control model feature according to claim 1 and space-time restriction, which is characterized in that institute
The First look similarity for calculating retrieval image and pedestrian image data set using trained Image Classifier is stated, is also wrapped before
It includes:
Two classification Image Classifiers are trained using the data set with smooth label, obtain the trained image point
Class device.
7. a kind of pedestrian of view-based access control model feature and space-time restriction identifying system again characterized by comprising
Vision computing module, for calculating the first of retrieval image and pedestrian image data set according to trained Image Classifier
Vision similarity；
Space-time calculation module, for establish space-time restriction model calculate it is described retrieval image and the pedestrian image data set when
Empty probability；
Fusion Module obtains Fusion Model for merging the First look similarity and the space-time probability；
Pedestrian's identification module, for calculating the of the retrieval image and the pedestrian image data set based on the Fusion Model
Two vision similarities, and be ranked up based on second vision similarity to obtain pedestrian's recognition result.
8. the pedestrian of view-based access control model feature according to claim 7 and space-time restriction identifying system again, which is characterized in that institute
The system of stating further includes optimization module, the optimization module include study sequencing unit, first optimization unit, second optimization unit with
And fusion optimization unit, in which:
The study sequencing unit is used to pedestrian's recognition result carrying out study sequence；
The first optimization unit is excellent to obtain for being trained according to the study ranking results to described image classifier
Image Classifier after change；
The second optimization unit is for being trained to obtain the space-time restriction model according to the study ranking results
Space-time restriction model after optimization；
The fusion optimization unit is used for based on the Image Classifier after the optimization and the space-time restriction model pair after optimization
The Fusion Model optimizes, the Fusion Model after being optimized.
9. the pedestrian of view-based access control model feature according to claim 8 and space-time restriction identifying system again, which is characterized in that institute
Stating study sequencing unit includes sequence computing unit, and the first sequence computing unit is used for according to List-wise method to institute
Pedestrian's recognition result is stated to be ranked up.
10. the pedestrian of view-based access control model feature according to claim 7 and space-time restriction identifying system again, which is characterized in that
The space-time calculation module includes space-time model creating unit, and the space-time model creating unit is for counting training data concentration
The time difference probability distribution that is taken by different cameras of the same pedestrian image, and described in being established based on the probability distribution
Space-time restriction model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910263596.7A CN110110598A (en) | 2019-04-01 | 2019-04-01 | The pedestrian of a kind of view-based access control model feature and space-time restriction recognition methods and system again |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910263596.7A CN110110598A (en) | 2019-04-01 | 2019-04-01 | The pedestrian of a kind of view-based access control model feature and space-time restriction recognition methods and system again |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110110598A true CN110110598A (en) | 2019-08-09 |
Family
ID=67485021
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910263596.7A Pending CN110110598A (en) | 2019-04-01 | 2019-04-01 | The pedestrian of a kind of view-based access control model feature and space-time restriction recognition methods and system again |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110110598A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110765863A (en) * | 2019-09-17 | 2020-02-07 | 清华大学 | Target clustering method and system based on space-time constraint |
| CN112232424A (en) * | 2020-10-21 | 2021-01-15 | 成都商汤科技有限公司 | Identity recognition method and device, electronic equipment and storage medium |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101169808A (en) * | 2007-11-26 | 2008-04-30 | 赵明 | Super time and super space combating game design for infusing network and realization |
| US20100318359A1 (en) * | 2009-06-10 | 2010-12-16 | Microsoft Corporation | Application-dependent information for recognition processing |
| US20110191098A1 (en) * | 2010-02-01 | 2011-08-04 | Stratify, Inc. | Phrase-based document clustering with automatic phrase extraction |
| CN103049751A (en) * | 2013-01-24 | 2013-04-17 | 苏州大学 | Improved weighting region matching high-altitude video pedestrian recognizing method |
| CN103310233A (en) * | 2013-06-28 | 2013-09-18 | 青岛科技大学 | Similarity mining method of similar behaviors between multiple views and behavior recognition method |
| CN103593661A (en) * | 2013-11-27 | 2014-02-19 | 天津大学 | Human body action recognition method based on ordering method |
| CN104102910A (en) * | 2014-08-07 | 2014-10-15 | 吉林农业大学 | Sports video tactical behavior recognition method based on space-time local mode |
| CN104318327A (en) * | 2014-10-17 | 2015-01-28 | 北京邮电大学 | Predictive parsing method for track of vehicle |
| CN105046211A (en) * | 2015-07-02 | 2015-11-11 | 中国人民解放军国防科学技术大学 | Spatio-temporal interest point feature encoding method in human motion recognition |
| CN105930768A (en) * | 2016-04-11 | 2016-09-07 | 武汉大学 | Spatial-temporal constraint-based target re-identification method |
| CN107563272A (en) * | 2017-06-14 | 2018-01-09 | 南京理工大学 | Target matching method in a kind of non-overlapping visual field monitoring system |
| CN108764269A (en) * | 2018-04-03 | 2018-11-06 | 华南理工大学 | A kind of cross datasets pedestrian recognition methods again based on space-time restriction incremental learning |
| CN108764167A (en) * | 2018-05-30 | 2018-11-06 | 上海交通大学 | A kind of target of space time correlation recognition methods and system again |
| CN109325471A (en) * | 2018-10-31 | 2019-02-12 | 中山大学 | A kind of binary-flow network pedestrian recognition methods again of combination appearance features and spatial and temporal distributions |
-
2019
- 2019-04-01 CN CN201910263596.7A patent/CN110110598A/en active Pending
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101169808A (en) * | 2007-11-26 | 2008-04-30 | 赵明 | Super time and super space combating game design for infusing network and realization |
| US20100318359A1 (en) * | 2009-06-10 | 2010-12-16 | Microsoft Corporation | Application-dependent information for recognition processing |
| US20110191098A1 (en) * | 2010-02-01 | 2011-08-04 | Stratify, Inc. | Phrase-based document clustering with automatic phrase extraction |
| CN103049751A (en) * | 2013-01-24 | 2013-04-17 | 苏州大学 | Improved weighting region matching high-altitude video pedestrian recognizing method |
| CN103310233A (en) * | 2013-06-28 | 2013-09-18 | 青岛科技大学 | Similarity mining method of similar behaviors between multiple views and behavior recognition method |
| CN103593661A (en) * | 2013-11-27 | 2014-02-19 | 天津大学 | Human body action recognition method based on ordering method |
| CN104102910A (en) * | 2014-08-07 | 2014-10-15 | 吉林农业大学 | Sports video tactical behavior recognition method based on space-time local mode |
| CN104318327A (en) * | 2014-10-17 | 2015-01-28 | 北京邮电大学 | Predictive parsing method for track of vehicle |
| CN105046211A (en) * | 2015-07-02 | 2015-11-11 | 中国人民解放军国防科学技术大学 | Spatio-temporal interest point feature encoding method in human motion recognition |
| CN105930768A (en) * | 2016-04-11 | 2016-09-07 | 武汉大学 | Spatial-temporal constraint-based target re-identification method |
| CN107563272A (en) * | 2017-06-14 | 2018-01-09 | 南京理工大学 | Target matching method in a kind of non-overlapping visual field monitoring system |
| CN108764269A (en) * | 2018-04-03 | 2018-11-06 | 华南理工大学 | A kind of cross datasets pedestrian recognition methods again based on space-time restriction incremental learning |
| CN108764167A (en) * | 2018-05-30 | 2018-11-06 | 上海交通大学 | A kind of target of space time correlation recognition methods and system again |
| CN109325471A (en) * | 2018-10-31 | 2019-02-12 | 中山大学 | A kind of binary-flow network pedestrian recognition methods again of combination appearance features and spatial and temporal distributions |
Non-Patent Citations (3)
| Title |
|---|
| 姬晓飞等: "基于分阶段视觉共生矩阵序列的双人交互识别", 《计算机工程与设计》, no. 09, 16 September 2017 (2017-09-16), pages 216 - 221 * |
| 江志浩等: "基于时空模型视频监控的行人活动预测算法", 《计算机应用与软件》, no. 01, 15 January 2017 (2017-01-15), pages 155 - 159 * |
| 胡琼等: "基于视觉的人体动作识别综述", 《计算机学报》, no. 12, 15 December 2013 (2013-12-15), pages 144 - 156 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110765863A (en) * | 2019-09-17 | 2020-02-07 | 清华大学 | Target clustering method and system based on space-time constraint |
| CN110765863B (en) * | 2019-09-17 | 2022-05-17 | 清华大学 | Target clustering method and system based on space-time constraint |
| CN112232424A (en) * | 2020-10-21 | 2021-01-15 | 成都商汤科技有限公司 | Identity recognition method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sarfraz et al. | A pose-sensitive embedding for person re-identification with expanded cross neighborhood re-ranking | |
| Van Gemert et al. | APT: Action localization proposals from dense trajectories. | |
| Wang et al. | A survey of vehicle re-identification based on deep learning | |
| Chen et al. | Survey of pedestrian action recognition techniques for autonomous driving | |
| Liu et al. | Nonparametric scene parsing: Label transfer via dense scene alignment | |
| Kuo et al. | How does person identity recognition help multi-person tracking? | |
| Su et al. | Learning a dense multi-view representation for detection, viewpoint classification and synthesis of object categories | |
| CN104881637B (en) | Multimodal information system and its fusion method based on heat transfer agent and target tracking | |
| Spinello et al. | Tracking people in 3D using a bottom-up top-down detector | |
| Kawewong et al. | Online and incremental appearance-based SLAM in highly dynamic environments | |
| Min et al. | Traffic sign recognition based on semantic scene understanding and structural traffic sign location | |
| Liu et al. | Learning deep features via congenerous cosine loss for person recognition | |
| Lee et al. | Place recognition using straight lines for vision-based SLAM | |
| Amato et al. | A comparison of face verification with facial landmarks and deep features | |
| CN104751136A (en) | Face recognition based multi-camera video event retrospective trace method | |
| WO2021082168A1 (en) | Method for matching specific target object in scene image | |
| CN109993061B (en) | Face detection and recognition method, system and terminal equipment | |
| CN104091171A (en) | Vehicle-mounted far infrared pedestrian detection system and method based on local features | |
| TWI525574B (en) | Collaborative face annotation method and collaborative face annotation system | |
| CN104463232A (en) | Density crowd counting method based on HOG characteristic and color histogram characteristic | |
| Premebida et al. | A cascade classifier applied in pedestrian detection using laser and image-based features | |
| CN107038400A (en) | Face identification device and method and utilize its target person tracks of device and method | |
| CN112001279B (en) | Cross-modal pedestrian re-identification method based on dual attribute information | |
| Symeonidis et al. | Neural attention-driven non-maximum suppression for person detection | |
| CN110110598A (en) | The pedestrian of a kind of view-based access control model feature and space-time restriction recognition methods and system again |
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: 20190809 |
|
| RJ01 | Rejection of invention patent application after publication |