CN109598240B - Video object quickly recognition methods and system again - Google Patents
Video object quickly recognition methods and system again Download PDFInfo
- Publication number
- CN109598240B CN109598240B CN201811478944.4A CN201811478944A CN109598240B CN 109598240 B CN109598240 B CN 109598240B CN 201811478944 A CN201811478944 A CN 201811478944A CN 109598240 B CN109598240 B CN 109598240B
- Authority
- CN
- China
- Prior art keywords
- image
- camera
- target
- probability
- weight
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/49—Segmenting video sequences, i.e. computational techniques such as parsing or cutting the sequence, low-level clustering or determining units such as shots or scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of video object adjusted based on various dimensions probability statistics quickly recognition methods and systems again, it is therefore an objective to the processing time that target identifies again is reduced in extensive camera network.This method comprises: obtaining the probability weight of each camera in video network；According to the sequence of probability weight from big to small, target image is carried out to the image of each camera pre-acquired and is identified again, target route track is generated；The probability weight of each camera is adjusted according to the target route track of generation.Since the larger camera of the subsequent probability of occurrence of target is given priority in arranging for retrieval, greatly reduces the degree of dispersion of the same target image in search procedure, the time of subsequent artefacts' link is greatly decreased；Simultaneously because weight is constantly adjusted, so whole system operation is the process that iteration optimization, a performance are constantly promoted；Therefore, technical solution of the present invention can be used for reducing the processing time that target identifies again in extensive camera network.
Description
Technical field
The present invention relates to technical field of image processing, and in particular to a kind of video mesh adjusted based on various dimensions probability statistics
Mark quick recognition methods and system again.
Background technique
Rapidly increase in view of the scale of present real time monitoring video network, traditional artificial access video finds targets of interest
Manpower costs it is increasing, cost rises violently.So thering is the image object weight identification technology of strength miscellaneous function to come into being, i.e.,
The image of the same target is searched in huge camera network by image recognition and comparison technology and in camera map
Middle generation route track.
Mainstream weighs identification technology substantially with the following method: to the image and mesh of camera pre-acquired all in video network
Logo image carries out characteristics of image comparison one by one, is ranked up according to comparing result (that is: target image distance), the figure after sequence
In image set conjunction since apart from nearest image, manually the picture for belonging to same target is selected, until choosing same mesh
Until piece of marking on a map reaches the threshold value of user demand, and camera belonging to these target images and capture time are automatically recorded,
Then generate route track.
Practice discovery, existing heavy identification technology has following defects that similar according to the image of distance definition due to machine
Degree has certain difference with the similarity that human eye defines, and with the growth of camera network size, same target image hardly may be used
Can be continuously distributed in result set, huge dispersion degree causes last hand picking process time-consuming extremely long.
Summary of the invention
The embodiment of the present invention provide it is a kind of based on various dimensions probability statistics adjust video object quickly again recognition methods and
System, it is therefore an objective to the processing time that target identifies again is reduced in extensive camera network.
The technical solution adopted by the invention is as follows:
First aspect present invention provides a kind of video object adjusted based on various dimensions probability statistics quickly identification side again
Method, this method comprises: obtaining the probability weight of each camera in video network；According to the sequence of probability weight from big to small,
Target image is carried out to the image of each camera pre-acquired to identify again, generates target route track；According to the target road of generation
Line tracking adjusts the probability weight of each camera.
Second aspect of the present invention, provide it is a kind of based on various dimensions probability statistics adjust video object quickly again identification system
System, which includes: acquisition module, for obtaining the probability weight of each camera in video network；Identification module, for pressing
According to the sequence of probability weight from big to small, target image is carried out to the image of each camera pre-acquired and is identified again, target is generated
Route track；Module is adjusted, the probability weight of each camera is adjusted for the target route track according to generation.
Third aspect present invention provides a kind of computer equipment, including processor and memory, stores in the memory
There is computer executable program, the processor can be performed following steps: obtaining view by executing the program stored in memory
The probability weight of each camera in frequency network；According to the sequence of probability weight from big to small, to each camera pre-acquired
Image carries out target image and identifies again, generates target route track；Each camera is adjusted according to the target route track of generation
Probability weight.
Fourth aspect present invention provides a kind of computer-readable medium, wherein be stored with computer executable program, when depositing
When the program of storage is included that the computer equipment of processor executes, so that the computer equipment is executed following steps: obtaining video network
The probability weight of each camera in network；According to the sequence of probability weight from big to small, to the image of each camera pre-acquired
It carries out target image to identify again, generates target route track；The general of each camera is adjusted according to the target route track of generation
Rate weight.
As can be seen from the above technical solutions, the embodiment of the present invention has the advantage that
Technical solution of the present invention abandons mainstream weight identification technology " carrying out global random comparison one by one to camera network "
Thinking introduces weight distribution concept, constantly adjusts the subsidiary probability of each camera according to the process that target each time is searched and weighs
Value, so according to after weighed value adjusting by image set that each camera acquires by weight size order import one by one it is to be found collect into
Row identification and hand picking again.Since the larger camera of the subsequent probability of occurrence of target is given priority in arranging for retrieval, make same target figure
The degree of dispersion of the picture in search procedure greatly reduces, and the time of subsequent artefacts' link is greatly decreased；Simultaneously because weight is
It is constantly adjusted, so whole system operation is the process that iteration optimization, a performance are constantly promoted.Technical solution of the present invention
It can be used for reducing the processing time that target identifies again in extensive camera network.
Detailed description of the invention
Technical solution in order to illustrate the embodiments of the present invention more clearly, below will be to needed in embodiment description
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for ability
For the those of ordinary skill of domain, without creative efforts, it can also be obtained according to these attached drawings other attached
Figure.
Fig. 1 is that the video object provided by one embodiment of the present invention adjusted based on various dimensions probability statistics is quickly identified again
The flow diagram of method；
Fig. 2 is the schematic diagram of the various dimensions image information for the camera that one embodiment of the invention provides；
Fig. 3 is that the video object provided by one embodiment of the present invention adjusted based on various dimensions probability statistics is quickly identified again
The structural schematic diagram of system.
Specific embodiment
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only
The embodiment of a part of the invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people
The model that the present invention protects all should belong in member's every other embodiment obtained without making creative work
It encloses.
Description and claims of this specification and term " first " in above-mentioned attached drawing, " second ", " third " etc. are
For distinguishing different objects, it is not use to describe a particular order.In addition, term " includes " and " having " and they are any
Deformation, it is intended that cover and non-exclusive include.Such as contain the process, method, system, product of a series of steps or units
Or equipment is not limited to listed step or unit, but optionally further comprising the step of not listing or unit, or can
Selection of land further includes the other step or units intrinsic for these process, methods, product or equipment.
Below by specific embodiment, it is described in detail respectively.
Referring to FIG. 1, one embodiment of the present of invention, provides a kind of video object adjusted based on various dimensions probability statistics
Quick recognition methods again, this method comprises:
11, the probability weight of each camera in video network is obtained；
12, the sequence according to probability weight from big to small carries out target image weight to the image of each camera pre-acquired
Identification generates target route track；
13, the probability weight of each camera is adjusted according to the target route track of generation.
Optionally, the sequence in step 12 according to probability weight from big to small, to the image of each camera pre-acquired into
Row target image identifies again, generates target route track, it may include:
121, the sequence according to probability weight from big to small imports the image of each camera pre-acquired one by one or in batches
Image set to be found；
122, after each import operation, target image is carried out to the image imported in image set to be found and is identified again, is obtained
Result figure image set is imported with the matched matching image of target image；
123, when the quantity for the matching image concentrated when result images reaches given threshold, according to of result images concentration
Target route track is generated with image.
Optionally, in step 122, target image is carried out to the image that imports in image set to be found and is identified again, obtain with
The matched matching image of target image imports result figure image set, it may include:
1221, by image set to be found image and target image carry out characteristic distance comparison one by one, according to feature away from
From from closely to remote sequence, being ranked up to the image in image set to be found；
1222, it carries out image one by one according to sequence to select, choose from image set to be found and target image matched
With image, the matching image and the target image include same target；
1223, the matching image chosen is imported into result figure image set.
Wherein, camera pre-acquired refers to: the embedded program of camera itself is defined the frame image of whole picture
The capture of target, for example defining target is pedestrian, just captures pedestrian in the picture；For vehicle, vehicle is just captured.
Wherein, characteristic distance, that is, target image distance refers to that image passes through convolutional neural networks or other image analysis moulds
The distance between the image feature vector that type is obtained by filtration, such as: geometric distance, manhatton distance.We can under normal conditions
Think the closer image of distance, content is more similar.
Optionally, step 11 obtains in video network before the probability weight of each camera, further includes:
Establish various dimensions image information, in number, camera view including camera it is each go out-number of boundary point and
It is each enter-number of boundary point；Weight, which is established, for each camera safeguards that matrix, the probability weight in matrix are denoted as An [x]
[y], the target for indicating that the out-boundary point x for the camera for being n from number disappears appear in the camera view that number is y
Probability size, wherein n, x, y are nonnegative integer.
Schematic diagram shown in Fig. 2 is specifically referred to, which is the image of certain camera shooting, and plurality of oval circle indicates
It is multiple go out-boundary point and/or enter-boundary point.
Following two Data Structures can be set up in the present invention:
One, singleton various dimensions image information class is established, includes:
" out-boundary point ": being beginning integer serial number with 0, note: out-boundary point, that is, target may disappear in video
Position；
" entering-boundary point " is beginning integer serial number with 0, note: enter-boundary point i.e. target may initially go out in video
Existing position；
Camera number: being beginning integer serial number with 0.
The same boundary point may be both out-boundary point and enter-boundary point.It will go out and enter separated mark in actual implementation
Note is the possibility because with the presence of some boundary points " only export but no import " or " only import but no export ".
Two, weight maintenance matrix is established
Weight, which is established, for each camera safeguards matrix.It is showed in machine with two-dimensional array, is set as An [x] [y]=weight
It is worth (i.e. probability weight), wherein n is camera number, and x value is that out-boundary point number, y take the photograph in the camera for remaining
As the number of head.
Weighted value be from the camera it is specific go out-boundary point disappear target occur again in specific camera head it is general
The class proportion of rate size.Weighted value is initialized as lamda, and the value of lamda is according to camera parameter and the environment of camera network
Parameter is determined by user, such as can be initialized as 1.
Weight safeguards matrix exemplary diagram:
Example enter one in the network containing 6 cameras number is 3 and there are 4 to go out-boundary point cameras, it initial
Weight two-dimensional matrix are as follows:
Optionally, in step 13, the probability weight of each camera is adjusted according to the target route track of generation, can be wrapped
Include: according to the target route track of generation, if target from the number of camera that number is m be that p goes out-go out after boundary point disappears
In the image for appearing in the camera shooting that number is q, then increase Am [p] [q] according to predetermined amplitude, wherein m, p, q are non-
Negative integer.Such as 0.1 can be increased every time.
More than, present invention method is illustrated.For ease of understanding, one is given further below specifically to answer
With embodiment, it is as follows that algorithm is run substantially:
For the identification again of each every single target
S1. Target Photo is chosen in the video frame.
S2. search queue is generated according to the sequence of corresponding probability weight, (batch size can one by one or by batch wherein
It is customized) pre-acquisition image set corresponding to subsequent camera is imported, 1. characteristic distance comparisons are carried out after each import；2.
Choose the picture of same target and imports result set.
S3. until in result set amount of images reach user's given threshold and draw target trajectory.
S4. the sequencing that the camera of target process is determined according to target trajectory, is then removed most in sequence in sequence
Each camera except the latter proceeds as follows:
Out-boundary point of the target is found in the camera, then according to the number of the camera of the next appearance of target
Corresponding weight is modified in weight maintenance matrix, such as: target, which disappears, in No. 3 cameras goes out-boundary points and immediately with No. 5
Occur in No. 8 cameras, then modify A3 [5] [8]=A3 [5] [8]+delta
Delta occurrence can have user customized according to environment, such as initialization delta=0.1.
Referring to FIG. 3, one embodiment of the present of invention, also provides a kind of video mesh adjusted based on various dimensions probability statistics
The quick weight identifying system of mark, which is characterized in that the system includes:
Module 31 is obtained, for obtaining the probability weight of each camera in video network；
Identification module 32, for the sequence according to probability weight from big to small, to the image of each camera pre-acquired into
Row target image identifies again, generates target route track；
Module 33 is adjusted, the probability weight of each camera is adjusted for the target route track according to generation.
Optionally, the identification module 32 can include:
Import unit, for the sequence according to probability weight from big to small, one by one by the image of each camera pre-acquired
Or image set to be found is imported in batches；
Recognition unit, for carrying out target image weight to the image imported in image set to be found after each import operation
Identification obtains importing result figure image set with the matched matching image of target image；
When the quantity of trajectory unit, the matching image for concentrating when result images reaches given threshold, according to result figure
Matching image in image set generates target route track.
Optionally, the recognition unit is specifically used for:
By in image set to be found image and target image carry out characteristic distance comparison one by one, according to characteristic distance from close
To remote sequence, the image in image set to be found is ranked up；
It carries out image one by one according to sequence to select, choose from image set to be found and the matched matching figure of target image
Picture, the matching image and the target image include same target；
The matching image chosen is imported into result figure image set.
It is optional, the system also includes:
Preprocessing module 30, it is each in number, camera view including camera for establishing various dimensions image information
Out-boundary point number and it is each enter-number of boundary point；And establish weight for each camera and safeguard matrix, in matrix
Probability weight be denoted as An [x] [y], indicate from number be n camera go out-boundary point x disappear target appear in number
For the probability size in the camera view of y.
Optionally, the adjustment module 33 is specifically used for:
According to the target route track of generation, if target from the number of camera that number is m be that the going out of p-goes out boundary point
In the image for appearing in the camera shooting that number is q after disappearance, then increase Am [p] [q] according to predetermined amplitude.
One embodiment of the present of invention also provides a kind of computer equipment, including processor and memory, the memory
In be stored with computer executable program, following steps can be performed by executing the program stored in memory in the processor:
Obtain the probability weight of each camera in video network；It is pre- to each camera according to the sequence of probability weight from big to small
The image of acquisition carries out target image and identifies again, generates target route track；It is adjusted according to the target route track of generation each
The probability weight of camera.
One embodiment of the present of invention also provides a kind of computer-readable medium, wherein being stored with computer can be performed journey
Sequence makes the computer equipment execute following steps: obtaining when the computer equipment that the program of storage is included processor executes
The probability weight of each camera in video network；According to the sequence of probability weight from big to small, to each camera pre-acquired
Image carry out target image identify again, generate target route track；Each camera shooting is adjusted according to the target route track of generation
The probability weight of head.
To sum up, in order to reduce the processing time that target identifies in extensive camera network again, the embodiment of the present invention is mentioned
A kind of video object adjusted based on various dimensions probability statistics quickly recognition methods and system again are supplied.Technical solution of the present invention
Key point is: operation algorithm substantially, various dimensions image information concept, probability weight concept, iteration optimization concept.
By using above-mentioned technical proposal, the present invention achieves following technical effect:
Technical solution of the present invention abandons mainstream weight identification technology " carrying out global random comparison one by one to camera network "
Thinking introduces weight distribution concept, constantly adjusts the subsidiary probability of each camera according to the process that target each time is searched and weighs
Value, so according to after weighed value adjusting by image set that each camera acquires by weight size order import one by one it is to be found collect into
Row identification and hand picking again.Since the larger camera of the subsequent probability of occurrence of target is given priority in arranging for retrieval, make same target figure
The degree of dispersion of the picture in search procedure greatly reduces, and the time of subsequent artefacts' link is greatly decreased；Simultaneously because weight is
It is constantly adjusted, so whole system operation is the process that iteration optimization, a performance are constantly promoted.Technical solution of the present invention
It can be used for reducing the processing time that target identifies again in extensive camera network.
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, is not described in some embodiment
Part, may refer to the associated description of other embodiments.
Above-described embodiment is merely illustrative of the technical solution of the present invention, rather than its limitations；The ordinary skill people of this field
Member is it is understood that it can still modify to technical solution documented by the various embodiments described above, or to part of skill
Art feature is equivalently replaced；And these are modified or replaceed, each reality of the present invention that it does not separate the essence of the corresponding technical solution
Apply the spirit and scope of a technical solution.
Claims (6)
1. a kind of video object quickly recognition methods again adjusted based on various dimensions probability statistics, which is characterized in that this method packet
It includes:
Obtain the probability weight of each camera in video network；
According to the sequence of probability weight from big to small, target image is carried out to the image of each camera pre-acquired and is identified again, it is raw
At target route track；
The probability weight of each camera is adjusted according to the target route track of generation；
Wherein, it obtains in video network before the probability weight of each camera further include:
Establish various dimensions image information, in number, camera view including camera it is each go out-number of boundary point and each
Enter-the number of boundary point；
Weight is established for each camera and safeguards that matrix, the probability weight in matrix are denoted as An [x] [y], indicates that from number be n's
The target that out-boundary point x of camera disappears appears in the probability size in the camera view that number is y, wherein n, x, y
It is nonnegative integer；
Wherein, include: according to the probability weight that the target route track of generation adjusts each camera
According to the target route track of generation, if target disappears from the boundary point that goes out-go out that the number for numbering the camera for being m is p
In the image for appearing in the camera shooting that number is q afterwards, then increase Am [p] [q] according to predetermined amplitude, wherein m, p, q are equal
For nonnegative integer.
2. the method according to claim 1, wherein the sequence according to probability weight from big to small, takes the photograph to each
It is identified again as the image of head pre-acquired carries out target image, generating target route track includes:
According to the sequence of probability weight from big to small, the image of each camera pre-acquired is imported into figure to be found one by one or in batches
Image set；
After each import operation, target image is carried out to the image imported in image set to be found and is identified again, is obtained and target figure
As matched matching image imports result figure image set；
When the quantity for the matching image that result images are concentrated reaches given threshold, the matching image concentrated according to result images is raw
At target route track.
3. according to the method described in claim 2, it is characterized in that, carrying out target figure to the image imported in image set to be found
As identifying again, obtain importing result figure image set with the matched matching image of target image include:
Image in image set to be found is subjected to characteristic distance comparison with target image one by one, according to characteristic distance from closely to remote
Sequence, the image in image set to be found is ranked up；
It carries out image one by one according to sequence to select, choose from image set to be found and the matched matching image of target image, institute
It states matching image and the target image includes same target；
The matching image chosen is imported into result figure image set.
4. a kind of video object adjusted based on various dimensions probability statistics quickly weighs identifying system, which is characterized in that the system packet
It includes:
Module is obtained, for obtaining the probability weight of each camera in video network；
Identification module carries out target to the image of each camera pre-acquired for the sequence according to probability weight from big to small
Image identifies again, generates target route track；
Module is adjusted, the probability weight of each camera is adjusted for the target route track according to generation；
Further include: preprocessing module, it is each in number, camera view including camera for establishing various dimensions image information
It is a go out-number of boundary point and it is each enter-number of boundary point；And weight is established for each camera and safeguards matrix, matrix
In probability weight be denoted as An [x] [y], indicate from number be n camera go out-boundary point x disappear target appear in volume
Number for y camera view in probability size, wherein n, x, y are nonnegative integer；
The adjustment module is specifically used for:
According to the target route track of generation, if target disappears from the boundary point that goes out-go out that the number for numbering the camera for being m is p
In the image for appearing in the camera shooting that number is q afterwards, then increase Am [p] [q] according to predetermined amplitude, wherein m, p, q are equal
For nonnegative integer.
5. system according to claim 4, which is characterized in that the identification module includes:
Import unit, for the sequence according to probability weight from big to small, by the image of each camera pre-acquired one by one or point
It criticizes and imports image set to be found；
Recognition unit, for carrying out target image to the image imported in image set to be found and identifying after each import operation again,
It obtains importing result figure image set with the matched matching image of target image；
When the quantity of trajectory unit, the matching image for concentrating when result images reaches given threshold, according to result figure image set
In matching image generate target route track.
6. system according to claim 5, which is characterized in that the recognition unit is specifically used for:
Image in image set to be found is subjected to characteristic distance comparison with target image one by one, according to characteristic distance from closely to remote
Sequence, the image in image set to be found is ranked up；
It carries out image one by one according to sequence to select, choose from image set to be found and the matched matching image of target image, institute
It states matching image and the target image includes same target；
The matching image chosen is imported into result figure image set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811478944.4A CN109598240B (en) | 2018-12-05 | 2018-12-05 | Video object quickly recognition methods and system again |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811478944.4A CN109598240B (en) | 2018-12-05 | 2018-12-05 | Video object quickly recognition methods and system again |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109598240A CN109598240A (en) | 2019-04-09 |
| CN109598240B true CN109598240B (en) | 2019-11-05 |
Family
ID=65961078
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811478944.4A Active CN109598240B (en) | 2018-12-05 | 2018-12-05 | Video object quickly recognition methods and system again |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109598240B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111107319B (en) * | 2019-12-25 | 2021-05-28 | 眸芯科技（上海）有限公司 | Target tracking method, device and system based on regional camera |
| CN112613496A (en) * | 2021-03-08 | 2021-04-06 | 深圳市安软科技股份有限公司 | Pedestrian re-identification method and device, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016106294A (en) * | 2015-12-28 | 2016-06-16 | 墫野 和夫 | Fully automatic robot household electric system appliance |
| 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 |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104469324B (en) * | 2014-12-25 | 2018-03-06 | 浙江宇视科技有限公司 | A kind of mobile target tracking method and device based on video |
| CN105930768A (en) * | 2016-04-11 | 2016-09-07 | 武汉大学 | Spatial-temporal constraint-based target re-identification method |
| CN106096577B (en) * | 2016-06-24 | 2019-05-31 | 安徽工业大学 | A kind of target tracking method in camera distribution map |
| CN106846378B (en) * | 2017-01-23 | 2019-12-03 | 中山大学 | A kind of across the video camera object matching and tracking of the estimation of combination topology of spacetime |
-
2018
- 2018-12-05 CN CN201811478944.4A patent/CN109598240B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016106294A (en) * | 2015-12-28 | 2016-06-16 | 墫野 和夫 | Fully automatic robot household electric system appliance |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109598240A (en) | 2019-04-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108062574B (en) | Weak supervision target detection method based on specific category space constraint | |
| CN110148120B (en) | Intelligent disease identification method and system based on CNN and transfer learning | |
| WO2021254205A1 (en) | Target detection method and apparatus | |
| CN104463117B (en) | A kind of recognition of face sample collection method and system based on video mode | |
| Fu et al. | Fast and accurate detection of banana fruits in complex background orchards | |
| CN111860375A (en) | Plant protection unmanned aerial vehicle ground monitoring system and monitoring method thereof | |
| CN109359666A (en) | A kind of model recognizing method and processing terminal based on multiple features fusion neural network | |
| CN110427922A (en) | One kind is based on machine vision and convolutional neural networks pest and disease damage identifying system and method | |
| CN101763502B (en) | High-efficiency method and system for sensitive image detection | |
| CN111709285A (en) | Epidemic situation protection monitoring method and device based on unmanned aerial vehicle and storage medium | |
| CN109117897A (en) | Image processing method, device and readable storage medium storing program for executing based on convolutional neural networks | |
| CN112990103B (en) | String mining secondary positioning method based on machine vision | |
| CN109598240B (en) | Video object quickly recognition methods and system again | |
| CN109271932A (en) | Pedestrian based on color-match recognition methods again | |
| Zhang et al. | An algorithm for automatic identification of multiple developmental stages of rice spikes based on improved Faster R-CNN | |
| CN107223242A (en) | Efficient local feature description's symbol filtering | |
| CN110827312A (en) | Learning method based on cooperative visual attention neural network | |
| CN116052222A (en) | Cattle face recognition method for naturally collecting cattle face image | |
| CN115861799A (en) | Light-weight air-to-ground target detection method based on attention gradient | |
| CN111291780B (en) | Cross-domain network training and image recognition method | |
| CN107704509A (en) | A kind of method for reordering for combining stability region and deep learning | |
| CN114445691A (en) | Model training method and device, electronic equipment and storage medium | |
| CN104008374B (en) | Miner's detection method based on condition random field in a kind of mine image | |
| CN111191527B (en) | Attribute identification method, attribute identification device, electronic equipment and readable storage medium | |
| Dubey et al. | A review of image segmentation using clustering methods |
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 |