CN110211156A - A kind of on-line study method of Space Time information consolidation - Google Patents
A kind of on-line study method of Space Time information consolidation Download PDFInfo
- Publication number
- CN110211156A CN110211156A CN201910480901.8A CN201910480901A CN110211156A CN 110211156 A CN110211156 A CN 110211156A CN 201910480901 A CN201910480901 A CN 201910480901A CN 110211156 A CN110211156 A CN 110211156A
- Authority
- CN
- China
- Prior art keywords
- network
- pedestrian
- target tracking
- target
- search
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000007596 consolidation process Methods 0.000 title claims abstract description 13
- 230000006399 behavior Effects 0.000 claims description 6
- LTXREWYXXSTFRX-QGZVFWFLSA-N Linagliptin Chemical compound N=1C=2N(C)C(=O)N(CC=3N=C4C=CC=CC4=C(C)N=3)C(=O)C=2N(CC#CC)C=1N1CCC[C@@H](N)C1 LTXREWYXXSTFRX-QGZVFWFLSA-N 0.000 claims description 3
- 230000008901 benefit Effects 0.000 abstract description 3
- 238000012544 monitoring process Methods 0.000 description 13
- 238000005516 engineering process Methods 0.000 description 7
- 238000001514 detection method Methods 0.000 description 5
- 238000005286 illumination Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- RLLPVAHGXHCWKJ-IEBWSBKVSA-N (3-phenoxyphenyl)methyl (1s,3s)-3-(2,2-dichloroethenyl)-2,2-dimethylcyclopropane-1-carboxylate Chemical compound CC1(C)[C@H](C=C(Cl)Cl)[C@@H]1C(=O)OCC1=CC=CC(OC=2C=CC=CC=2)=C1 RLLPVAHGXHCWKJ-IEBWSBKVSA-N 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- 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
- G06T2207/30196—Human being; Person
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a kind of on-line study methods of Space Time information consolidation, are mutually improved efficiency with target tracking algorithm and pedestrian's searching algorithm, interactively search for network with pedestrian to tracking network and are trained.The specific steps of the present invention are as follows: (1) input video flow data；(2) operational network carries out sample expansion；(3) pedestrian searches for network and target tracking network takes movement according to network state simultaneously.Pedestrian is searched for network to the present invention and target tracking network combines, and has strong robustness, the fast advantage of arithmetic speed.
Description
Technical field
It is that one kind is related to pedestrian's search and target chases after the present invention relates to a kind of on-line study method of Space Time information consolidation
The image processing techniques of track is mutually improved efficiency using pedestrian's searching algorithm and target tracking algorithm, is interactively searched to pedestrian
Rope network and target tracking network are trained.
Background technique
At present in the public place of densely populated place, government department, enterprises and institutions, residential quarters, even many residents
It is equipped with monitoring camera in family, provides reliable video monitoring to maintain public order, assuring the safety for life and property of the people
Resource.In video monitoring, since the Parameters variations such as the resolution ratio of camera, shooting angle are larger, it is difficult to realize high quality people
Stablizing for face picture obtains, so that the target tracking stability based on face recognition technology is poor.In contrast, pedestrian searches for
(Person Search) technology can provide robustness stronger target tracking solution for video monitoring.
Traditional pedestrian's search technique is divided into target detection and pedestrian and identifies two parts again, the purpose of target detection be from
Interested target is searched in picture, and it is accurately positioned, since target is under different angle and distance shootings,
Its shape, posture and relative size all change, and illumination when along with imaging, the interference for the factors such as blocking, target detection is always
It is most challenging one of the problem of computer vision field.It is to judge particular row in image or video library that pedestrian identifies again
A kind of computer vision technique that people whether there is is to confirm on the basis of target detection to the identity of pedestrian.Currently,
Feature learning, metric learning and production confrontation network model are widely applied to pedestrian and identify field again.Pedestrian's search technique
The spatial structural form of image is mainly utilized, it is not high to the inter-frame information utilization rate of video sequence.And video object tracks skill
Art mainly utilizes the inter-frame information of video, is efficiently positioned to interested target, however, due to the deformation of target, dashing forward
So reasons such as movement and environmental change, very big influence is caused to the performance of target tracking.
Domestic and foreign scholars identify again in target detection, pedestrian and the aspect of target tracking three has carried out that system is deep to grind
Study carefully, and proposes many methods.However, effectively monitoring range in practical application scene in order to increase, camera can be pacified
It is placed on higher position, causes pedestrian target size in entire monitored picture smaller, while vulnerable to foreign matters such as trees, buildings
It blocks on ground.Big in some flows of the people, pedestrian densely region is also easy to produce overlapping between multiple pedestrian targets and blocks.By
Monitored picture clarity is not high, illumination, shooting angle is different and pedestrian influences with wearing the factors such as similar clothes, the row of different identity
People may also have similar feature.Target tracking mainly handles single camera data, and pedestrian's search can handle multi-cam
Video data, be more adaptive to practical application scene.Target tracking method can for pedestrian search for provide technology auxiliary and
The reference of method.
Summary of the invention
Goal of the invention: in order to overcome the deficiencies in the prior art, the present invention provide a kind of Space Time information consolidation
Line learning method, the evident characteristics again searched for using the space-time characterisation of target tracking and pedestrian, searches pedestrian using intensified learning
Rope network and target tracking network are updated, and arithmetic speed is very fast, and can improve pedestrian search and target tracking precision and
Timeliness.
Technical solution: to achieve the above object, the technical solution adopted by the present invention are as follows:
A kind of on-line study method of Space Time information consolidation, includes the following steps:
(1) input video flow data；
(2) pedestrian is run simultaneously and search for network and target tracking network, include the following steps:
(21) the tracking target in video flowing is tracked using target tracking network, while by target tracking network
Tracking target concurrently sets the search target that network is searched for for pedestrian；
(22) search target is sampled every n frame, sampled result is saved in chronological order to exptended sample collection C1
={ c1_t, c1_t-1, c1_t-2...；Tracking target is sampled every n frame, sampled result is saved in chronological order to expansion
Fill sample set C2={ c2_t, c2_t-1, c2_t-2...；
(3) intensified learning strategy:
(31) it is directed to the intensified learning of target tracking network, divides the following two kinds situation:
Situation is 1.: if the target tracking accuracy of target tracking network is lower than 90%, according to intensified learning strategy, will expand
Sample set C1 is extended in target tracking network sample set, optimizes current target tracking network；
Situation is 2.: if the target tracking accuracy of target tracking network is greater than or equal to 90%, current target being maintained to chase after
Track network；
(32) intensified learning that network is searched for for pedestrian, divides the following two kinds situation:
Situation is 1.: if the search accuracy that pedestrian searches for network is lower than 90%, according to intensified learning strategy, by exptended sample
Collection C2 extends to pedestrian and searches in network sample set, optimizes current pedestrian and searches for network；
Situation is 2.: if the search accuracy that pedestrian searches for network is greater than or equal to 90%, maintaining current pedestrian's dragnet
Network.
In this case, the search accuracy of network is searched for according to pedestrian, judges whether that needing to adjust pedestrian searches for network sample
Collection reuses intensified learning strategy and optimizes to pedestrian's search network；According to the target tracking accuracy of target tracking network,
Judge whether to need to adjust target tracking network sample set, reuses intensified learning strategy and target tracking network is optimized.
It is one kind in the case where given pedestrian's identity that pedestrian, which searches for network, the video shot from single or multiple cameras
The computer vision technique of the pedestrian movement track is found in stream.But in actual video monitoring environment, due to being taken the photograph
As the limitation of head position and the visual field, single monitoring camera is difficult to realize all standing to target monitoring region, even with more
A camera, it is also difficult to realize to target monitoring region it is seamless, be overlapped, cover all around so that existing pedestrian's search technique
Still it is difficult to meet the needs of real-time target matching in large-scale intelligent monitoring system.Meanwhile pedestrian's search is equally vulnerable to resolution ratio
Different, the similar pedestrian of clothing, illumination variation, visual angle change and foreign matter such as block at the influence of factors.Although currently, by a large amount of
Monitoring camera accumulates a large amount of video data, and still, acquisition data corresponding to each pedestrian are very rare.Cause
This, pedestrian's search mission is a typical big data small sample problem, how to be excavated from a large amount of video data hiding
Important information therein, while being characterized in solving pedestrian's search mission using the differentiation of a small amount of pedestrian's data study to pedestrian
It is crucial.
Target tracking network is predicted subsequent in the case where giving target sizes and the position of certain video sequence initial frame
The size of the target and position in frame.Usual target tracking faces several big difficult points: appearance deformation, illumination variation, quickly movement and
Similar interference of motion blur, background etc., these difficult points will lead to tracking target and lose.This problem, this hair are lost for target
It is bright to search for network by introducing pedestrian, the property of network is searched for according to pedestrian, i.e., in the case where given pedestrian's identity, from single
Or the computer vision technique of the pedestrian movement track is found in the video flowing of multiple camera shootings.
Specifically, in the step (3), intensified learning strategy specifically:
(41) initialize: target tracking network sample set is U1, behavior aggregate A1={ a1_0, a1_1, a1_2, a1_3...；Pedestrian
Search network sample set is U2, behavior aggregate A2={ a2_0, a2_1, a2_2, a2_3...；Act aj_iExpression will be in exptended sample collection Cj
Newest im sample frame extends in Uj, i.e. Uj={ Uj, cj_t, cj_t-1, cj_t-2..., cj_t-im+1, act aj_iAward
Value is rj_ai, { rj_ai}=0, m is positive integer, j=1,2；Enter step (42)；
(42) the target tracking accuracy g of target tracking network is tracked respectively1The search accuracy of network is searched for pedestrian
g2If: gjLower than 90%, then (43) are entered step；Otherwise, (42) are entered step, i.e., persistently track gj, until discovery is lower than
gj90%；
(43) max { r is executedj_aiCorresponding movement aj_i, that is, the highest movement of reward value is executed, continues to track gjIf: gj
Lower than 90%, then (44) are entered step；Otherwise, (45) are entered step；
(44) i=i+1, execution act aj_i, continue to track gj: if it is lower than 90%, enters step (44)；Otherwise, enter
Step (45)；
(45) enter the step, then illustrate that execution acts aj_iAfterwards, g is improvedj, and reached 90% given threshold,
Give movement aj_iOne award, rj_ai=rj_ai+ 1, return step (42).
Be for the intensified learning strategy that target tracking network and pedestrian search for network it is identical, but the two be again it is mutual not
Interference, it is respectively independent.Can be seen that intensified learning strategy according to the whole process of intensified learning strategy is awarded by selection
It is worth highest movement to optimize the method that pedestrian searches for network or target tracking network, pedestrian is made to search for network and target tracking net
Network is able to maintain that a higher accuracy.
Specifically, in the step (43), two or more max { r if it existsj_aiCorresponding movement aj_i, it is minimum to execute i
Movement, and i ≠ 0；Restriction to step (43) indicates the side that we preferentially use when the reward value of multiple movements is identical
Method is to increase less sample frame, can improve system operational speed as much as possible in this way.
Preferably, in the step (22), n >=30；Since consecutive frame feature is close, if acquisition data sample interval is too
It is small, it is impossible to guarantee the diversity of sample characteristics；If it is too big to acquire data sample interval, target may be lost.
Preferably, in the step (41), m >=30；Decline problem when accuracy occurs, illustrates present frame target signature not
Obviously, therefore by present frame expand at least 30 frame samples forward and enter sample set, characteristic similarity is higher within 30 frames, can be compared with
The feature of accuracy decline frame is supplemented, well so as to improve search accuracy.
This method is mapped to the characteristic of movement using intensified learning from ambient condition, then by target tracking algorithm to pedestrian
It searches for target and carries out data collection.During pedestrian's search, it is to model respectively that pedestrian, which searches for network there are two kinds of movements,
The movement that the movement of fine tuning and model remain unchanged.Pedestrian searches for network, and there are two states, are the normal state of model respectively
With the insufficient state of model performance.The state that network is searched for two movement adjustment pedestrians, by the insufficient state tune of model performance
Whole is the normal state of model.The feedback information that network is searched for according to pedestrian, judges whether to need to target tracking network collection
To sample be acquired, reuse intensified learning and collected difficult sample set screened, model is finely tuned to reach
Purpose.Similarly, target tracking (search) network is updated with same method.
The utility model has the advantages that the on-line study method of Space Time information consolidation provided by the invention, have compared with prior art with
Lower advantage:
(1) present invention is directed to large-scale video data, and target tracking technology is used to automatically generate for pedestrian's searching method
There is the data set of label, the screening of data is carried out using intensified learning technology, realizes the promotion of pedestrian's searching method performance；Using
Pedestrian's search technique is that the data set that target tracking method automatically generates label can be quick due to the characteristic of pedestrian's search
It finds lost target and records the difficult sample of such target, the screening of data is carried out using intensified learning technology, realize target
The promotion of method for tracing performance.
(2) pedestrian is searched for network by the present invention and target tracking network combines, since pedestrian searches for the property of network,
Lost target can be searched again for when target tracking, improves the efficiency of target tracking, mesh can be made in this way
Mark tracking network searches for network with pedestrian and mutually promotes, and forms a benign cycle.
Detailed description of the invention
Fig. 1 is implementing procedure block diagram of the invention；
The logic diagram of Fig. 2 to realize the present invention.
Specific embodiment
The present invention will be further explained with reference to the accompanying drawing.
It is as shown in Figure 1 the implementing procedure block diagram of a kind of on-line study method of Space Time information consolidation, with reference to the accompanying drawing
It illustrates.
Step 1: input video flow data
At present in the public place of densely populated place, government department, enterprises and institutions, residential quarters, even many residents
It is equipped with monitoring camera in family, provides reliable video monitoring to maintain public order, assuring the safety for life and property of the people
Resource.Therefore we have a large amount of original video data, input these data as test sample collection.
Step 2: running pedestrian simultaneously searches for network and target tracking network
Since pedestrian's search technique is not high to video interframe information utilization, and target tracking method can be captured effectively
The information of video interframe, but it is weaker to the comprehension of information ability of image space.Therefore by using intensified learning technical tie-up
Pedestrian's search and target tracking technology can effectively carry out the Space Time information excavating of video data big data.
(21) the tracking target in video flowing is tracked using target tracking network, while by target tracking network
Tracking target concurrently sets the search target that network is searched for for pedestrian；
(22) search target is sampled every n frame, sampled result is saved in chronological order to exptended sample collection C1
={ c1_t, c1_t-1, c1_t-2...；Tracking target is sampled every n frame, sampled result is saved in chronological order to expansion
Fill sample set C2={ c2_t, c2_t-1, c2_t-2...；
Step 3: the application of intensified learning strategy
(31) it is directed to the intensified learning of target tracking network, divides the following two kinds situation:
Situation is 1.: if the target tracking accuracy of target tracking network is lower than 90%, according to intensified learning strategy, will expand
Sample set C1 is extended in target tracking network sample set, optimizes current target tracking network；
Situation is 2.: if the target tracking accuracy of target tracking network is greater than or equal to 90%, current target being maintained to chase after
Track network；
(32) intensified learning that network is searched for for pedestrian, divides the following two kinds situation:
Situation is 1.: if the search accuracy that pedestrian searches for network is lower than 90%, according to intensified learning strategy, by exptended sample
Collection C2 extends to pedestrian and searches in network sample set, optimizes current pedestrian and searches for network；
Situation is 2.: if the search accuracy that pedestrian searches for network is greater than or equal to 90%, maintaining current pedestrian's dragnet
Network.
Step 4: the detailed process of intensified learning strategy
(41) initialize: target tracking network sample set is U1, behavior aggregate A1={ a1_0, a1_1, a1_2, a1_3...；Pedestrian
Search network sample set is U2, behavior aggregate A2={ a2_0, a2_1, a2_2, a2_3...；Act aj_iExpression will be in exptended sample collection Cj
Newest im sample frame extends in Uj, i.e. Uj={ Uj, cj_t, cj_t-1, cj_t-2..., cj_t-im+1, act aj_iAward
Value is rj_ai, { rj_ai}=0, m is positive integer, j=1,2；Enter step (42)；
(42) the target tracking accuracy g of target tracking network is tracked respectively1The search accuracy of network is searched for pedestrian
g2If: gjLower than 90%, then (43) are entered step；Otherwise, (42) are entered step, i.e., persistently track gj, until discovery is lower than
gj90%；
(43) max { r is executedj_aiCorresponding movement aj_i, that is, the highest movement of reward value is executed, continues to track gjIf: gj
Lower than 90%, then (44) are entered step；Otherwise, (45) are entered step；
(44) i=i+1, execution act aj_i, continue to track gj: if it is lower than 90%, enters step (44)；Otherwise, enter
Step (45)；
(45) enter the step, then illustrate that execution acts aj_iAfterwards, g is improvedj, and reached 90% given threshold,
Give movement aj_iOne award, rj_ai=rj_ai+ 1, return step (42).
The above is only a preferred embodiment of the present invention, it should be pointed out that: for the ordinary skill people of the art
For member, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also answered
It is considered as protection scope of the present invention.
Claims (5)
1. a kind of on-line study method of Space Time information consolidation, characterized by the following steps:
(1) input video flow data；
(2) pedestrian is run simultaneously and search for network and target tracking network, include the following steps:
(21) the tracking target in video flowing is tracked using target tracking network, while by the tracking of target tracking network
Target concurrently sets the search target that network is searched for for pedestrian；
(22) search target is sampled every n frame, sampled result is saved in chronological order to exptended sample collection C1=
{c1_t,c1_t-1,c1_t-2,…}；Tracking target is sampled every n frame, sampled result is saved in chronological order to expansion
Sample set C2={ c2_t,c2_t-1,c2_t-2,…}；
(3) intensified learning strategy:
(31) it is directed to the intensified learning of target tracking network, divides the following two kinds situation:
Situation is 1.: if the target tracking accuracy of target tracking network is lower than 90%, according to intensified learning strategy, by exptended sample
Collection C1 is extended in target tracking network sample set, optimizes current target tracking network；
Situation is 2.: if the target tracking accuracy of target tracking network is greater than or equal to 90%, maintaining current target tracking net
Network；
(32) intensified learning that network is searched for for pedestrian, divides the following two kinds situation:
Situation is 1.: if the search accuracy that pedestrian searches for network is lower than 90%, according to intensified learning strategy, by exptended sample collection C2
It extends to pedestrian to search in network sample set, optimizes current pedestrian and search for network；
Situation is 2.: if the search accuracy that pedestrian searches for network is greater than or equal to 90%, current pedestrian being maintained to search for network.
2. the on-line study method of Space Time information consolidation according to claim 1, it is characterised in that: the step (3)
In, intensified learning strategy specifically:
(41) initialize: target tracking network sample set is U1, behavior aggregate A1={ a1_0,a1_1,a1_2,a1_3,…}；Pedestrian's search
Network sample set is U2, behavior aggregate A2={ a2_0,a2_1,a2_2,a2_3,…}；Act aj_iExpression will be newest in exptended sample collection Cj
Im sample frame extend in Uj, i.e. Uj={ Uj, cj_t,cj_t-1,cj_t-2,…,cj_t-im+1, act aj_iReward value be
rj_ai, { rj_ai}=0, m is positive integer, j=1,2；Enter step (42)；
(42) the target tracking accuracy g of target tracking network is tracked respectively1The search accuracy g of network is searched for pedestrian2If:
gjLower than 90%, then (43) are entered step；Otherwise, (42) are entered step, i.e., persistently track gj, until discovery is lower than gj90%；
(43) max { r is executedj_aiCorresponding movement aj_i, that is, the highest movement of reward value is executed, continues to track gjIf: gjIt is lower than
90%, then enter step (44)；Otherwise, (45) are entered step；
(44) i=i+1, execution act aj_i, continue to track gj: if it is lower than 90%, enters step (44)；Otherwise, it enters step
(45)；
(45) enter the step, then illustrate that execution acts aj_iAfterwards, g is improvedj, and reached 90% given threshold, give
Act aj_iOne award, rj_ai=rj_ai+ 1, return step (42).
3. the on-line study method of Space Time information consolidation according to claim 1, it is characterised in that: the step (43)
In, two or more max { r if it existsj_aiCorresponding movement aj_i, execute the smallest movement of i, and i ≠ 0.
4. the on-line study method of Space Time information consolidation according to claim 1, it is characterised in that: described to state step
(22) in, n >=30.
5. the on-line study method of Space Time information consolidation according to claim 2, it is characterised in that: described to state step
(41) in, m >=30.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910480901.8A CN110211156B (en) | 2019-06-04 | 2019-06-04 | Time-space information combined online learning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910480901.8A CN110211156B (en) | 2019-06-04 | 2019-06-04 | Time-space information combined online learning method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110211156A true CN110211156A (en) | 2019-09-06 |
| CN110211156B CN110211156B (en) | 2021-02-12 |
Family
ID=67790530
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910480901.8A Active CN110211156B (en) | 2019-06-04 | 2019-06-04 | Time-space information combined online learning method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110211156B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130339353A1 (en) * | 2010-12-06 | 2013-12-19 | Fraunhofe-Gesellschaft zur Foerderung der angewand ten Forschung e.V. | Method for Operating a Geolocation Database and a Geolocation Database System |
| CN106408610A (en) * | 2015-04-16 | 2017-02-15 | 西门子公司 | Method and system for machine learning based assessment of fractional flow reserve |
| CN107729953A (en) * | 2017-09-18 | 2018-02-23 | 清华大学 | Robot plume method for tracing based on continuous state behavior domain intensified learning |
| CN108932840A (en) * | 2018-07-17 | 2018-12-04 | 北京理工大学 | Automatic driving vehicle urban intersection passing method based on intensified learning |
-
2019
- 2019-06-04 CN CN201910480901.8A patent/CN110211156B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130339353A1 (en) * | 2010-12-06 | 2013-12-19 | Fraunhofe-Gesellschaft zur Foerderung der angewand ten Forschung e.V. | Method for Operating a Geolocation Database and a Geolocation Database System |
| CN106408610A (en) * | 2015-04-16 | 2017-02-15 | 西门子公司 | Method and system for machine learning based assessment of fractional flow reserve |
| CN107729953A (en) * | 2017-09-18 | 2018-02-23 | 清华大学 | Robot plume method for tracing based on continuous state behavior domain intensified learning |
| CN108932840A (en) * | 2018-07-17 | 2018-12-04 | 北京理工大学 | Automatic driving vehicle urban intersection passing method based on intensified learning |
Non-Patent Citations (3)
| Title |
|---|
| MIN YANG ETC.: ""Multiagent-Based Simulation of Temporal-Spatial Characteristics of Activity-Travel Patterns Using Interactive Reinforcement Learning"", 《MATHEMATICAL PROBLEMS IN ENGINEERING》 * |
| 舒朗等: ""基于回归与深度强化学习的目标检测算法"", 《软件导刊》 * |
| 陈宗海等: ""基于节点生长k-均值聚类算法的强化学习方法"", 《计算机研究与发展》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110211156B (en) | 2021-02-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103546726B (en) | Method for automatically discovering illegal land use | |
| CN102831617A (en) | Method and system for detecting and tracking moving object | |
| CN107833239B (en) | Optimization matching target tracking method based on weighting model constraint | |
| CN109389086A (en) | Detect the method and system of unmanned plane silhouette target | |
| CN104598883A (en) | Method for re-recognizing target in multi-camera monitoring network | |
| CN111161309B (en) | Searching and positioning method for vehicle-mounted video dynamic target | |
| CN109448027A (en) | A kind of adaptive, lasting motion estimate method based on algorithm fusion | |
| CN106851302B (en) | A kind of Moving Objects from Surveillance Video detection method based on intraframe coding compression domain | |
| CN110765880A (en) | Light-weight video pedestrian heavy identification method | |
| CN105913452A (en) | Real-time space debris detection and tracking method | |
| CN106709938A (en) | Multi-target tracking method based on improved TLD (tracking-learning-detected) | |
| CN102663362A (en) | Moving target detection method t based on gray features | |
| CN112801019A (en) | Method and system for eliminating re-identification deviation of unsupervised vehicle based on synthetic data | |
| CN106572387A (en) | video sequence alignment method and system | |
| Gadde et al. | Transductive weakly-supervised player detection using soccer broadcast videos | |
| Casagrande et al. | Abnormal motion analysis for tracking-based approaches using region-based method with mobile grid | |
| Vora et al. | Bringing generalization to deep multi-view pedestrian detection | |
| Gemignani et al. | A novel background subtraction approach based on multi layered self-organizing maps | |
| CN103218829A (en) | Foreground extracting method suitable for dynamic background | |
| Shu et al. | Small moving object detection and tracking based on event signals | |
| CN110211156A (en) | A kind of on-line study method of Space Time information consolidation | |
| CN110210405A (en) | A kind of pedestrian's search sample extending method based on target tracking | |
| Zhao et al. | An Improved Method for Infrared Vehicle and Pedestrian Detection Based on YOLOv5s | |
| CN110276233A (en) | A kind of polyphaser collaboration tracking system based on deep learning | |
| CN110197163A (en) | A kind of target tracking sample extending method based on pedestrian's search |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20221129 Address after: 221000 Building A13, Safety Technology Industrial Park, Tongshan District, Xuzhou City, Jiangsu Province Patentee after: XUZHOU GUANGLIAN TECHNOLOGY Co.,Ltd. Address before: 221008 Tongshan University Road, Xuzhou City, Jiangsu Province, Institute of Scientific Research, China University of Mining and Technology Patentee before: CHINA University OF MINING AND TECHNOLOGY |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: An Online Learning Method Based on Temporal Spatial Information Union Effective date of registration: 20231108 Granted publication date: 20210212 Pledgee: Xuzhou Huaichang Investment Co.,Ltd. Pledgor: XUZHOU GUANGLIAN TECHNOLOGY Co.,Ltd. Registration number: Y2023980064378 |