CN108596109A - A kind of object detection method and device based on neural network and motion vector - Google Patents
A kind of object detection method and device based on neural network and motion vector Download PDFInfo
- Publication number
- CN108596109A CN108596109A CN201810385675.0A CN201810385675A CN108596109A CN 108596109 A CN108596109 A CN 108596109A CN 201810385675 A CN201810385675 A CN 201810385675A CN 108596109 A CN108596109 A CN 108596109A
- Authority
- CN
- China
- Prior art keywords
- image
- described image
- target
- frame
- motion vector
- 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
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
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention provides a kind of object detection method and device, method includes：Parsing stream of video data records the serial number per frame image to extract multiple image；For every frame image, determine whether image is intraframe predictive coding frame according to the serial number of image, if so, executing A1；Otherwise A2 to A4 is executed；A1：Image is identified by neural network model to detect at least one target, stores the location information of each target in the picture, each target is marked to form target image in the picture；A2：Reference picture is determined according to the serial number of image, obtains corresponding motion vector information；A3：According to the location information and motion vector information of each target in a reference image, the current location information of each target in the picture is determined；A4：According to the current location information of each target in the picture, each target is marked to form target image in the picture.Technical solution provided by the invention, detection efficiency are higher.
Description
Technical field
The present invention relates to field of computer technology, more particularly to a kind of target detection based on neural network and motion vector
Method and device.
Background technology
In vehicle-mounted automated driving system, it is often necessary to be carried out to the stream of video data that image capture device is sent corresponding
Parsing and identification, each target on each frame image to determine and mark stream of video data carrying (for example, vehicle,
Pedestrian and traffic sign etc.) position, realize to entrained by stream of video data image carry out target detection.
Currently, when realization carries out target detection to the image entrained by stream of video data, need through neural network mould
Type carries out image recognition respectively to each frame image entrained by stream of video data, to determine and mark on each frame image
At least one target, and each frame image after label is exported.
In above-mentioned technical proposal, by neural network model to each frame image carry out respectively image recognition elapsed time compared with
Long, detection efficiency is relatively low.
Invention content
An embodiment of the present invention provides a kind of object detection method and device based on neural network and motion vector, school survey
Efficiency is higher.
In a first aspect, the present invention provides a kind of object detection method based on neural network and motion vector, including：
Receive the stream of video data that image capture device is sent；
The stream of video data is parsed to extract at least two field pictures, and each frame described image for recording extraction is divided
Not corresponding serial number；
It is directed to each frame described image, the serial number corresponding to described image determines whether described image is in frame
Encoded predicted frame, if so, executing A1；Otherwise, A2, A3 and A4 are executed；
A1：Described image is identified by neural network model to detect at least one mesh of described image carrying
Mark determines and stores location information of each target in described image, to each target in described image
Processing is marked to form target image in position；
A2：Serial number corresponding to described image determines the reference picture corresponding to described image, and described in acquisition
Image corresponds to the motion vector information of the reference picture；
A3：According to each target location information in the reference picture and the motion vector information, really
Fixed current location information of each target in described image；
A4：According to current location information of each target in described image, to each described in described image
Processing is marked to form target image in the position of target.
Preferably,
The serial number corresponding to described image determines whether described image is intraframe predictive coding frame, including：
The evaluation of estimate corresponding to described image is calculated by following formula：
Wherein, it is default normal more than 1 that β, which characterizes serial number, n corresponding to the evaluation coefficient, α characterization described images,
Number；
When β is integer, determine that described image is intraframe predictive coding frame, otherwise, it determines described image is inter-prediction
Coded frame.
Preferably,
The motion vector information, including：Whole pixel motion vector and intraframe predictive coding image block.
Second aspect, an embodiment of the present invention provides a kind of object detecting device based on neural network and motion vector,
Including
Code stream receiving module, the stream of video data for receiving image capture device transmission；
Code stream analyzing module parses the stream of video data to extract at least two field pictures, and records each of extraction
The corresponding serial number of frame described image institute；
Image determining module, for being directed to each frame described image, the serial number corresponding to described image determines
Whether described image is intraframe predictive coding frame, if so, triggering neural network calling module；Otherwise, acquisition of information mould is triggered
Block；
The neural network calling module, is identified to detect described image for passing through neural network model
At least one target of image carrying is stated, determine and stores location information of each target in described image, described
Processing is marked to form target image to the position of each target in image；
Described information acquisition module determines the ginseng corresponding to described image for the serial number corresponding to described image
Image is examined, and obtains the motion vector information that described image corresponds to the reference picture；
Position determination module, for according to each target location information in the reference picture and the fortune
Dynamic vector information determines current location information of each target in described image；
Processing module is marked, for the current location information according to each target in described image, in the figure
Processing is marked to form target image to the position of each target as in.
Preferably,
Described image determining module, for executing following steps：
The evaluation of estimate corresponding to described image is calculated by following formula：
Wherein, it is default normal more than 1 that β, which characterizes serial number, n corresponding to the evaluation coefficient, α characterization described images,
Number；
When β is integer, determines that described image is intraframe predictive coding frame, when β is non-integer, determine that described image is
Inter prediction encoding frame.
Preferably,
The motion vector information, including：Whole pixel motion vector and intraframe predictive coding image block.
An embodiment of the present invention provides a kind of object detection method and device based on neural network and motion vector, the party
It in method, is parsed by the stream of video data to reception to extract at least two field pictures, and records each frame of extraction
The corresponding serial number of image institute subsequently can then be directed to each frame image, and the serial number corresponding to the image determines should
Whether image is intraframe predictive coding frame；When image is intraframe predictive coding frame, just by neural network model to the image
It is identified to detect at least one target of image carrying, place is marked to the position of each target in the images
Reason stores the location information of each target in the images to form target image；When image is non-intraframe predictive coding frame
When, the reference picture corresponding to the image can be determined according to the serial number corresponding to the image, and obtain the image and correspond to ginseng
The motion vector information of image is examined, to true according to the location information of motion vector information and each target in a reference image
The fixed current location information of each target in the images, and then the current location information according to each target in the images,
Processing is marked to form target image to the position of each target in the images.In conclusion the embodiment of the present invention carries
The technical solution of confession, without being known respectively to each frame image entrained by stream of video data by neural network model
Not, you can detect and mark each target on each frame image, detection efficiency is higher.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is the present invention
Some embodiments for those of ordinary skill in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 is a kind of stream for object detection method based on neural network and motion vector that one embodiment of the invention provides
Cheng Tu；
Fig. 2 is a kind of knot for object detecting device based on neural network and motion vector that one embodiment of the invention provides
Structure schematic diagram.
Specific implementation mode
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments, based on the embodiments of the present invention, those of ordinary skill in the art
The every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
As shown in Figure 1, an embodiment of the present invention provides a kind of object detection method based on neural network and motion vector,
Including：
Step 101, the stream of video data that image capture device is sent is received；
Step 102, the stream of video data is parsed to extract at least two field pictures, and is recorded described in each frame of extraction
The corresponding serial number of image institute；
Step 103, the image corresponding to a non-selected serial number is selected successively；
Step 104, the serial number corresponding to the described image selected determines whether described image is compiled for intra prediction
Code frame, if so, thening follow the steps 105；Otherwise, step 106 is executed；
Step 105, described image is identified by neural network model to detect that described image carries at least
One target determines and stores location information of each target in described image, to each described in described image
Processing is marked to form target image in the position of target；
Step 106, the serial number corresponding to described image determines the reference picture corresponding to described image, and obtains
Described image corresponds to the motion vector information of the reference picture；
Step 107, believed according to each target location information in the reference picture and the motion vector
Breath, determines current location information of each target in described image；
Step 108, the current location information according to each target in described image, to each in described image
Processing is marked to form target image in the position of the target；
Step 109, if there are non-selected images, if so, thening follow the steps 103；Otherwise, terminate current process.
Embodiment as shown in Figure 1 is parsed by the stream of video data to reception to extract at least two frame figures
Picture, and the corresponding serial number of each frame image institute of extraction is recorded, it subsequently can be then directed to each frame image, according to the image
Corresponding serial number determines whether the image is intraframe predictive coding frame；When image is intraframe predictive coding frame, just pass through
Neural network model is identified the image to detect at least one target of image carrying, in the images to each
Processing is marked to form target image in the position of target, and stores the location information of each target in the images；Work as figure
When as being non-intraframe predictive coding frame, the reference picture corresponding to the image can be determined according to the serial number corresponding to the image,
And the motion vector information that the image corresponds to reference picture is obtained, to joined according to motion vector information and each target
It examines the location information in image and determines the current location information of each target in the images, and then according to each target in the figure
Processing is marked to form target image to the position of each target in the images in current location information as in.To sum up
It is described, technical solution provided in an embodiment of the present invention, without passing through neural network model to every entrained by stream of video data
One frame image is identified respectively, you can detects and marks each target on each frame image, detection efficiency is higher.
Correspondingly, without being known respectively to each frame image entrained by stream of video data by neural network model
Not, the workload of neural network model can be also saved, each frame image progress mesh realized and carried to stream of video data is reduced
The power consumption for marking the terminal of detection executes other tasks for terminal and provides enough computer storage threshold doseags, while end also can be improved
The processing capacity that video data caused by end is blocked and deforms upon for target distorts, improves its robustness.
In one embodiment of the invention, the serial number corresponding to described image determines whether described image is frame
Intraprediction encoding frame, including：
The evaluation of estimate corresponding to described image is calculated by following formula：
Wherein, it is default normal more than 1 that β, which characterizes serial number, n corresponding to the evaluation coefficient, α characterization described images,
Number；
When β is integer, determine that described image is intraframe predictive coding frame, otherwise, it determines described image is inter-prediction
Coded frame.
In above-described embodiment, n is preset constant more than 1, which (for example can hold in conjunction with practical business scene
The computing capability or storage capacity of the equipment of row this method) rationally it is arranged, the size of preset constant would generally influence image
The accuracy of detection of middle target, in order to ensure with higher accuracy of detection, preset constant should be relatively small, can usually be arranged
For more than 1 and less than 6 integer, for example, could be provided as any one in 2,3,4,5.
In above-described embodiment, as α=1, i.e., when the serial number corresponding to image is 1, it is from reception to characterize the image
First image extracted in stream of video data needs to pass through nerve net at this point, the image should be intraframe predictive coding frame
Network model is identified the image to detect one or more targets of image carrying, in the images to each target
Position be marked processing to form target image, meanwhile, should also store each target in the images position letter
Breath.
It is understandable, it is every to meet item defined by above-described embodiment from each frame image of the video extraction of reception
Each frame image of part should all be confirmed as town intraprediction encoding frame, for example, be taken when in n=3 and stream of video data
When 9 frame image of band, each frame image that corresponding serial number is 1,4,7 can be confirmed as town intraprediction encoding frame；Correspondingly, corresponding
Each frame image that serial number is 2,3,5,6,8,9 is inter prediction encoding frame.
In one embodiment of the invention, when being parsed due to the video data to reception, typically receiving while parses,
It therefore, that is, can be according to image institute when determining that image is inter prediction encoding frame when it is not intraframe predictive coding frame to determine
Corresponding serial number carries out backtracking traversal according to descending sequence to each image of extraction
No is intraframe predictive coding frame, is determined as the present image of traversal when the present image of traversal is intraframe predictive coding frame
Reference picture.It for example, then can be first when the present image for determining that corresponding serial number is 6 is not intraframe predictive coding frame
First traversal order number 5, when determining image corresponding to serial number 5 nor when intraframe predictive coding frame, traversal order number 4,
When it is intraframe predictive coding frame to determine the image corresponding to serial number 4, then the image corresponding to serial number 4 can be determined as
Reference picture corresponding with the present image.
In one embodiment of the invention, the motion vector information, in including but not limited to whole pixel motion vector and frame
Predictive-coded picture block, for example, can also include dividing pixel motion vector.
For example, when the present image A of corresponding serial number 6 is inter prediction encoding frame, the reference corresponding to the image
Image is the image B (intraframe predictive coding frame) that corresponding serial number is 4, when motion vector information includes intraframe predictive coding figure
When as block, characterizes in image A and image B and all have the intraframe predictive coding image block, the intraframe predictive coding image can be passed through
Whole pixel motion vector of the block in image B with respect to it in image A, determines the same target in image A and image B
Relative motion vectors then can be according to relative motion vectors and each due to position of each target in image A it is known that therefore
Location determination of the target in image A goes out position of each target in image B, to mark each target in image B.
As shown in Fig. 2, an embodiment of the present invention provides a kind of object detecting device based on neural network and motion vector,
Including：
Code stream receiving module 201, the stream of video data for receiving image capture device transmission；
Code stream analyzing module 202 parses the stream of video data to extract at least two field pictures, and records the every of extraction
The corresponding serial number of one frame described image institute；
Image determining module 203, for being directed to each frame described image, the serial number corresponding to described image is true
Determine whether described image is intraframe predictive coding frame, if so, triggering neural network calling module 204；Otherwise, triggering information obtains
Modulus block 205；
The neural network calling module 204 is identified to detect described image for passing through neural network model
Go out at least one target of described image carrying, determine and stores location information of each target in described image,
Processing is marked to form target image to the position of each target in described image；
Described information acquisition module 205 is determined for the serial number corresponding to described image corresponding to described image
Reference picture, and obtain described image correspond to the reference picture motion vector information；
Position determination module 206, for according to each target location information in the reference picture and institute
Motion vector information is stated, determines current location information of each target in described image；
Processing module 207 is marked, for the current location information according to each target in described image, described
Processing is marked to form target image to the position of each target in image.
In one embodiment of the invention, described image determining module 203, for executing following steps：
The evaluation of estimate corresponding to described image is calculated by following formula：
Wherein, it is default normal more than 1 that β, which characterizes serial number, n corresponding to the evaluation coefficient, α characterization described images,
Number；
When β is integer, determines that described image is intraframe predictive coding frame, when β is non-integer, determine that described image is
Inter prediction encoding frame.
In one embodiment of the invention, the motion vector information, including：Whole pixel motion vector and intraframe predictive coding
Image block.
The contents such as the information exchange between each unit, implementation procedure in above-mentioned apparatus, due to implementing with the method for the present invention
Example is based on same design, and particular content can be found in the narration in the method for the present invention embodiment, and details are not described herein again.
In conclusion each embodiment of the present invention at least has the advantages that：
1, it in one embodiment of the invention, is parsed by the stream of video data to reception to extract at least two frame figures
Picture, and the corresponding serial number of each frame image institute of extraction is recorded, it subsequently can be then directed to each frame image, according to the image
Corresponding serial number determines whether the image is intraframe predictive coding frame；When image is intraframe predictive coding frame, just pass through
Neural network model is identified the image to detect at least one target of image carrying, in the images to each
Processing is marked to form target image in the position of target, and stores the location information of each target in the images；Work as figure
When as being non-intraframe predictive coding frame, the reference picture corresponding to the image can be determined according to the serial number corresponding to the image,
And the motion vector information that the image corresponds to reference picture is obtained, to joined according to motion vector information and each target
It examines the location information in image and determines the current location information of each target in the images, and then according to each target in the figure
Processing is marked to form target image to the position of each target in the images in current location information as in.To sum up
It is described, technical solution provided in an embodiment of the present invention, without passing through neural network model to every entrained by stream of video data
One frame image is identified respectively, you can detects and marks each target on each frame image, detection efficiency is higher.
2, in one embodiment of the invention, without passing through neural network model to each frame figure entrained by stream of video data
As being identified respectively, the workload of neural network model can be also saved, reduction realization carries stream of video data each
Frame image carries out the power consumption of the terminal of target detection, and executing other tasks for terminal provides enough computer storage threshold doseags, together
When the processing capacity that video data caused by terminal is blocked and deforms upon for target distorts also can be improved, improve its robust
Property.
It should be noted that herein, such as first and second etc relational terms are used merely to an entity
Or operation is distinguished with another entity or operation, is existed without necessarily requiring or implying between these entities or operation
Any actual relationship or order.Moreover, the terms "include", "comprise" or its any other variant be intended to it is non-
It is exclusive to include, so that the process, method, article or equipment including a series of elements includes not only those elements,
But also include other elements that are not explicitly listed, or further include solid by this process, method, article or equipment
Some elements.In the absence of more restrictions, the element limited by sentence " including one ", is not arranged
Except there is also other identical factors in the process, method, article or apparatus that includes the element.
Finally, it should be noted that：The foregoing is merely presently preferred embodiments of the present invention, is merely to illustrate the skill of the present invention
Art scheme, is not intended to limit the scope of the present invention.Any modification for being made all within the spirits and principles of the present invention,
Equivalent replacement, improvement etc., are included within the scope of protection of the present invention.
Claims (6)
1. a kind of object detection method based on neural network and motion vector, which is characterized in that including：
Receive the stream of video data that image capture device is sent；
The stream of video data is parsed to extract at least two field pictures, and each frame described image institute for recording extraction is right respectively
The serial number answered；
It is directed to each frame described image, the serial number corresponding to described image determines whether described image is intra prediction
Coded frame, if so, executing A1；Otherwise, A2, A3 and A4 are executed；
A1：Described image is identified by neural network model to detect at least one target of described image carrying,
It determines and stores location information of each target in described image, to the position of each target in described image
Processing is marked to form target image；
A2：Serial number corresponding to described image determines the reference picture corresponding to described image, and obtains described image
Corresponding to the motion vector information of the reference picture；
A3：According to each target location information in the reference picture and the motion vector information, determine each
Current location information of a target in described image；
A4：According to current location information of each target in described image, to each target in described image
Position be marked processing to form target image.
2. according to the method described in claim 1, it is characterized in that,
The serial number corresponding to described image determines whether described image is intraframe predictive coding frame, including：
The evaluation of estimate corresponding to described image is calculated by following formula：
Wherein, it is the preset constant more than 1 that β, which characterizes serial number, n corresponding to the evaluation coefficient, α characterization described images,；
When β is integer, determine that described image is intraframe predictive coding frame, otherwise, it determines described image is inter prediction encoding
Frame.
3. method according to claim 1 or 2, which is characterized in that
The motion vector information, including：Whole pixel motion vector and intraframe predictive coding image block.
4. a kind of object detecting device based on neural network and motion vector, which is characterized in that including：
Code stream receiving module, the stream of video data for receiving image capture device transmission；
Code stream analyzing module parses the stream of video data to extract at least two field pictures, and records each frame institute of extraction
State the corresponding serial number of image institute；
Image determining module, for being directed to each frame described image, described in the serial number determination corresponding to described image
Whether image is intraframe predictive coding frame, if so, triggering neural network calling module；Otherwise, data obtaining module is triggered；
The neural network calling module is identified described image for passing through neural network model to detect the figure
As at least one target carried, determines and store location information of each target in described image, in described image
In to the position of each target be marked processing to form target image；
Described information acquisition module determines the reference chart corresponding to described image for the serial number corresponding to described image
Picture, and obtain the motion vector information that described image corresponds to the reference picture；
Position determination module, for according to each target location information in the reference picture and movement arrow
Information is measured, determines current location information of each target in described image；
Processing module is marked, for the current location information according to each target in described image, in described image
Processing is marked to form target image to the position of each target.
5. device according to claim 4, which is characterized in that
Described image determining module, for executing following steps：
The evaluation of estimate corresponding to described image is calculated by following formula：
Wherein, it is the preset constant more than 1 that β, which characterizes serial number, n corresponding to the evaluation coefficient, α characterization described images,；
It when β is integer, determines that described image is intraframe predictive coding frame, when β is non-integer, determines that described image is interframe
Encoded predicted frame.
6. device according to claim 4 or 5, which is characterized in that
The motion vector information, including：Whole pixel motion vector and intraframe predictive coding image block.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810385675.0A CN108596109B (en) | 2018-04-26 | 2018-04-26 | Target detection method and device based on neural network and motion vector |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810385675.0A CN108596109B (en) | 2018-04-26 | 2018-04-26 | Target detection method and device based on neural network and motion vector |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108596109A true CN108596109A (en) | 2018-09-28 |
| CN108596109B CN108596109B (en) | 2021-04-13 |
Family
ID=63610266
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810385675.0A Active CN108596109B (en) | 2018-04-26 | 2018-04-26 | Target detection method and device based on neural network and motion vector |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108596109B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111292354A (en) * | 2020-01-22 | 2020-06-16 | 维沃移动通信有限公司 | False detection suppression method and electronic device |
| CN111654700A (en) * | 2020-06-19 | 2020-09-11 | 杭州海康威视数字技术股份有限公司 | Privacy mask processing method and device, electronic equipment and monitoring system |
| CN111950517A (en) * | 2020-08-26 | 2020-11-17 | 司马大大（北京）智能系统有限公司 | Target detection method, model training method, electronic device and storage medium |
| CN112714316A (en) * | 2020-12-21 | 2021-04-27 | 太原智林信息技术股份有限公司 | Regular mark detection and classification identification method based on video code stream |
| CN112889283A (en) * | 2018-10-19 | 2021-06-01 | 三星电子株式会社 | Encoding method and apparatus thereof, and decoding method and apparatus thereof |
| WO2021120866A1 (en) * | 2019-12-18 | 2021-06-24 | 深圳云天励飞技术股份有限公司 | Target object tracking method and apparatus, and terminal device |
| CN113610900A (en) * | 2021-10-11 | 2021-11-05 | 深圳佑驾创新科技有限公司 | Method and device for predicting scale change of vehicle tail sequence and computer equipment |
| CN113794886A (en) * | 2021-08-11 | 2021-12-14 | 镕铭微电子(济南)有限公司 | Method and apparatus for assisting target detection using video coding information |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0723366A2 (en) * | 1995-01-17 | 1996-07-24 | Graphics Communications Laboratories | Motion estimation method and apparatus for calculating a motion vector |
| CN103218830A (en) * | 2013-04-07 | 2013-07-24 | 北京航空航天大学 | Method for extracting video object contour based on centroid tracking and improved GVF Snake |
| CN104683802A (en) * | 2015-03-24 | 2015-06-03 | 江南大学 | H.264/AVC compressed domain based moving target tracking method |
-
2018
- 2018-04-26 CN CN201810385675.0A patent/CN108596109B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0723366A2 (en) * | 1995-01-17 | 1996-07-24 | Graphics Communications Laboratories | Motion estimation method and apparatus for calculating a motion vector |
| CN103218830A (en) * | 2013-04-07 | 2013-07-24 | 北京航空航天大学 | Method for extracting video object contour based on centroid tracking and improved GVF Snake |
| CN104683802A (en) * | 2015-03-24 | 2015-06-03 | 江南大学 | H.264/AVC compressed domain based moving target tracking method |
Non-Patent Citations (1)
| Title |
|---|
| 詹翊强 等: "《基于MPEG-2视频流的目标跟踪快速算法》", 《上海交通大学学报》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112889283A (en) * | 2018-10-19 | 2021-06-01 | 三星电子株式会社 | Encoding method and apparatus thereof, and decoding method and apparatus thereof |
| WO2021120866A1 (en) * | 2019-12-18 | 2021-06-24 | 深圳云天励飞技术股份有限公司 | Target object tracking method and apparatus, and terminal device |
| US11620756B2 (en) | 2019-12-18 | 2023-04-04 | Shenzhen Intellifusion Technologies Co., Ltd. | Target object tracking method and apparatus, and terminal device |
| CN111292354A (en) * | 2020-01-22 | 2020-06-16 | 维沃移动通信有限公司 | False detection suppression method and electronic device |
| CN111654700A (en) * | 2020-06-19 | 2020-09-11 | 杭州海康威视数字技术股份有限公司 | Privacy mask processing method and device, electronic equipment and monitoring system |
| CN111654700B (en) * | 2020-06-19 | 2022-12-06 | 杭州海康威视数字技术股份有限公司 | Privacy mask processing method and device, electronic equipment and monitoring system |
| CN111950517A (en) * | 2020-08-26 | 2020-11-17 | 司马大大（北京）智能系统有限公司 | Target detection method, model training method, electronic device and storage medium |
| CN112714316A (en) * | 2020-12-21 | 2021-04-27 | 太原智林信息技术股份有限公司 | Regular mark detection and classification identification method based on video code stream |
| CN113794886A (en) * | 2021-08-11 | 2021-12-14 | 镕铭微电子(济南)有限公司 | Method and apparatus for assisting target detection using video coding information |
| CN113610900A (en) * | 2021-10-11 | 2021-11-05 | 深圳佑驾创新科技有限公司 | Method and device for predicting scale change of vehicle tail sequence and computer equipment |
| CN113610900B (en) * | 2021-10-11 | 2022-02-15 | 深圳佑驾创新科技有限公司 | Method and device for predicting scale change of vehicle tail sequence and computer equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108596109B (en) | 2021-04-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108596109A (en) | A kind of object detection method and device based on neural network and motion vector | |
| CN109255322B (en) | A kind of human face in-vivo detection method and device | |
| US9158964B2 (en) | Object recognizing apparatus and method | |
| WO2016173277A1 (en) | Video coding and decoding methods and apparatus | |
| CN110782433A (en) | Dynamic information violent parabolic detection method and device based on time sequence and storage medium | |
| US8982209B2 (en) | Method and apparatus for operating a video system | |
| Zhang et al. | An imbalance compensation framework for background subtraction | |
| US9123133B1 (en) | Method and apparatus for moving object detection based on cerebellar model articulation controller network | |
| CN109711344B (en) | Front-end intelligent specific abnormal behavior detection method | |
| US6823011B2 (en) | Unusual event detection using motion activity descriptors | |
| CN109903172A (en) | Claims Resolution information extracting method and device, electronic equipment | |
| CN110849885A (en) | Hard spot monitoring method, device and system in bow net system | |
| CN106385585B (en) | The method, apparatus of frame encoding and decoding, system, storage medium | |
| WO2024037660A1 (en) | Method and apparatus for determining abnormal sorting areas, electronic device, and storage medium | |
| CN105959633A (en) | Video transmission method and device | |
| CN111510731A (en) | System and method for splicing traffic images | |
| CN108668129A (en) | A kind of coding method and device | |
| CN115941939A (en) | Video frame alignment method, device, equipment and storage medium | |
| JPWO2020039897A1 (en) | Station monitoring system and station monitoring method | |
| CN111027482B (en) | Behavior analysis method and device based on motion vector segmentation analysis | |
| Ciuntu et al. | Real-time traffic sign detection and classification using machine learning and optical character recognition | |
| KR102179077B1 (en) | syntax-based method of providing object classification in compressed video by use of neural network which is learned by cooperation with an external commercial classifier | |
| US20240221426A1 (en) | Behavior detection method, electronic device, and computer readable storage medium | |
| CN112560557A (en) | People number detection method, face detection device and electronic equipment | |
| CN108304817B (en) | The method and apparatus for realizing gesture operation |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20210319 Address after: No. 1036, Shandong high tech Zone wave road, Ji'nan, Shandong Applicant after: INSPUR GROUP Co.,Ltd. Address before: 250100 First Floor of R&D Building 2877 Kehang Road, Sun Village Town, Jinan High-tech Zone, Shandong Province Applicant before: JINAN INSPUR HI-TECH INVESTMENT AND DEVELOPMENT Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |