CN108595469A - A kind of semantic-based agricultural machinery monitor video image section band Transmission system - Google Patents
A kind of semantic-based agricultural machinery monitor video image section band Transmission system Download PDFInfo
- Publication number
- CN108595469A CN108595469A CN201810181230.0A CN201810181230A CN108595469A CN 108595469 A CN108595469 A CN 108595469A CN 201810181230 A CN201810181230 A CN 201810181230A CN 108595469 A CN108595469 A CN 108595469A
- Authority
- CN
- China
- Prior art keywords
- video
- semantic
- frame
- label
- nouns
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
- G06V20/47—Detecting features for summarising video content
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/48—Matching video sequences
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention relates to intelligent network connection agricultural machinery systems technology fields, a kind of disclosed semantic-based agricultural machinery monitor video image section band Transmission system, information collection and video monitoring are carried out by vehicle-mounted camera and roadside field fixing camera, the key frame and movement locus of video are extracted again, and the original motion trajectory to extracting carries out post-processing operation, obtains its direction of motion and track G-bar information；Then Object identifying operation is carried out to key frame, recognition result is added in video semantic nouns tag set；Then name-matches are carried out with the trail file in semantic nouns label and training set, find out corresponding trail file and calculates track semantic similarity, select the highest trail file of k similarity；The present invention can improve the computing capability and storage capacity of the network equipment, and the bandwidth consumption for exchanging or saving network transmission for storage capacity is calculated using network element appropriate, that is, improves communication capacity, and be conducive to the inquiry and retrieval of video data.
Description
Technical field
The present invention relates to intelligent network connection agricultural machinery systems technology field more particularly to a kind of semantic-based agricultural machinery monitor videos
Image section band Transmission system.
Background technology
In intelligent network joins agricultural machinery system, video data is to be transferred to data center and core net by mobile network.
However, due to farmland geographical location and floor space etc., intelligent network joins the mobile communications network of agricultural machinery system：Core net with
The medium that access net is transmitted as mobile video business is but faced with bandwidth, wired and wirelessly limited congenital disadvantage.
It is for the video information equally measured, how in its useful information of guarantee to save with the specific goal in research of system is changed
It does not lose and under the premise of video quality does not reduce, reduces as far as possible in its transmission process to transfer resource, is i.e. bandwidth disappears
Consumption.
The video content that intelligent network connection agricultural machinery generates in operation process much all need to only transmit its content information,
User is simultaneously not concerned with detailed video pictures, and is intended to obtain information from video pictures, for example, encountered in traveling process barrier,
Semantic information can only be transmitted to feed back video content by driving to the regular events such as field boundary.Since semantic information is compared
In video information, its scale of construction is few too many, therefore in intelligent network joins agricultural machinery system, semantic-based section bandization technology is extremely important.
Meanwhile the semantic information by extracting video, video data can be converted from the representation based on frame based on semantic right to
The representation of elephant so that computer can more accurately understand video data, convenient for being better achieved to video content
Expression, is also more conducive to the inquiry and retrieval of video data.Intelligent network joins in agricultural machinery system, and video data is to pass through mobile network
Network is transferred to data center and core net.However, due to farmland geographical location and floor space etc., intelligent network joins agricultural machinery
The medium that the core net of the mobile communications network of system is transmitted with access net as mobile video business, is but faced with bandwidth, has
Line and the congenital disadvantage being wirelessly limited.Therefore, such as how existing hardware technology is incited somebody to action to store and transmit a large amount of video data
As one of problem in the urgent need to address.
Invention content
The present invention proposes a kind of semantic-based agricultural machinery monitor video image section band Transmission system.
For achieving the above object, the present invention adopts the following technical scheme that：
A kind of semantic-based agricultural machinery monitor video image section band Transmission system, it is special for the process that agricultural machinery is advanced
Sign is：Information collection and video monitoring are carried out by vehicle-mounted camera and roadside field fixing camera, then extracts video
Key frame and movement locus, and the original motion trajectory to extracting carry out post-processing operation, obtain its direction of motion and
Track G-bar information；Then Object identifying operation is carried out to key frame, recognition result is added to video semantic nouns mark
In label set；Then name-matches are carried out with the trail file in semantic nouns label and training set, finds out corresponding track text
Part simultaneously calculates track semantic similarity, selects the highest trail file of k similarity；Finally, it reads of the same name with k trail file
Training set mark file, result is added in video semantic verbs tag set；Fed back by transmitting semantic information
Video content；
Video data is wherein fed back, its language is expressed in such a way that semantic nouns label and semantic verbs label are combined
It is Moving Objects in video that adopted information, wherein semantic nouns label is corresponding, and corresponding semantic verbs label is Moving Objects
Specific action behavior, the two combines the semantic label as video；
1, semantic nouns tag extraction：
In terms of semantic nouns label, the operations such as shot segmentation and key-frame extraction are carried out to test video first, depending on
Then the key frame of frequency carries out Object identifying, you can obtain semantic nouns label to key frame.
1) shot segmentation
Shot segmentation algorithm based on histogram, be the gray scale of each pixel between consecutive frame, brightness are divided into it is N number of
Grade, then make histogram for each grade statistical pixel number and compare, the histogram of two images is provided, then difference of histograms
Calculation formula is as follows：
Wherein, N is the sum of image frame pixel；What hm (i)-hn (i) was indicated is two video frame in this histogram list of i
Distance above position；Do not consider the location information of pixel based on histogram method, and uses the statistical value of its brightness and color；
2) key-frame extraction
Key-frame Extraction Algorithm based on cluster, the key-frame extraction based on cluster, implementation step：
The first step：If some camera lens Mi includes n picture frame, it is expressed as Mi={ N1... ..., Nn, wherein N1Headed by frame,
NnFor tail frame；If the similarity between adjacent two frame define be this adjacent two frames color histogram similarity, that is to say straight
Square figure characteristic difference predefines the density of a threshold value δ control cluster；
Second step:Calculate present frame NiWith it is existing some cluster barycenter between similarity, if the value be less than δ, the frame
Distance is larger between the cluster, therefore NiIt cannot be added in the cluster；If Ni and all existing cluster barycenter similarities are small
In δ, then it is its barycenter that Ni, which forms new a cluster and Ni,；Otherwise the frame is added to and is similarly spent in maximum cluster,
Keep the distance between the frame and this barycenter clustered minimum；
Third walks：After the n picture frame that camera lens Mi is included is referred to different clusters respectively by first two steps, pass is selected
Key frame：The representative frame that the frame nearest from cluster barycenter is clustered as this, the representative of all clusters are extracted from each cluster
Frame just constitutes the key frame of camera lens Mi；
2, semantic verbs tag extraction：
In terms of semantic verbs label, the fortune of test video is obtained by the related algorithm of Moving Objects detect and track first
Dynamic track data, then investigates the semantic similarity between the track data and the track data of training set video of test video,
The semantic nouns label of test video and the marked content of training set video are contacted simultaneously, can be obtained verb after comprehensive analysis
Semantic label；
1) movement locus extraction and analysis when extracting movement locus, usually first carry out Moving Objects detection operation, then right
The Moving Objects that detected are into line trace, by the result of tracking with coordinate representation out to get movement locus；Extraction movement rail
When mark, Moving Objects detection operation is usually first carried out, then to the Moving Objects that detected into line trace, by the result of tracking
With coordinate representation out to get movement locus；
2) video trains set construction method, using the method for machine learning come when extracting video semanteme, construction video is trained
Collection；Way is first selecting video, extracts the related data of key frame and movement locus, and longest one is chosen in movement locus
Item, and its information is extracted by track post-processing operation, include the G-bar of the direction of motion, path curves；Then right
These videos carry out manual mark, and marked content is the motor behavior of object；Wherein, key frame is used for video semantic nouns label
Extraction, motion track information and the content marked by hand are prepared for video semantic verbs tag extraction；To in training set
Each video extract its related data and carry out manual mark, after the completion of all operations, video training set i.e. construct finish；
3) semantic verbs tag extraction algorithm, semantic verbs tag extraction algorithm are to be based on machine learning thought, with training
The data of collection are matched with the data of video to be analyzed, and process is as follows：The key frame and movement locus of video are extracted first,
And the original motion trajectory to extracting carries out post-processing operation, obtains its direction of motion and track G-bar information；It connects
It and Object identifying operation is carried out to key frame, recognition result is added in video semantic nouns tag set；Then noun is used
Trail file in semantic label and training set carries out name-matches, finds out corresponding trail file and to calculate track semanteme similar
Degree, selects the highest trail file of k similarity；Finally, it reads and marks file with k trail file training set of the same name, it will
As a result it is added in video semantic verbs tag set.
Due to using technical solution as described above, the present invention that there is following superiority：
It is to be directed to the video information equally measured that agricultural machinery net connection, which drives section with the specific goal in research of system is changed, how to be protected
Hinder its useful information not lose, under the premise of video quality does not reduce, reduce as far as possible in its transmission process to transfer resource,
That is the consumption of bandwidth.Meanwhile under the premise of communication bandwidth is limited, improving the computing capability and storage capacity of the network equipment
It is relatively easy, feasible method.Therefore, section is exactly to utilize network element appropriate with the core concept of image transmitting processing method is changed
Calculating and storage capacity, exchange or save the bandwidth consumption of network transmission, i.e. communication capacity for.
The video content that intelligent network connection agricultural machinery generates in operation process much all need to only transmit its content information,
User is simultaneously not concerned with detailed video pictures, and is intended to obtain information from video pictures, for example, encountered in traveling process barrier,
Semantic information can only be transmitted to feed back video content by driving to the regular events such as field boundary.Since semantic information is compared
In video information, its scale of construction is few too many, therefore in intelligent network joins agricultural machinery system, semantic-based section bandization technology is extremely important.
Meanwhile the semantic information by extracting video, video data can be converted from the representation based on frame based on semantic right to
The representation of elephant so that computer can more accurately understand video data, convenient for being better achieved to video content
Expression, is also more conducive to the inquiry and retrieval of video data.It can be seen that video semanteme extractive technique has important research
Meaning and application value.
Description of the drawings
Fig. 1 is video training set construction flow chart；
Fig. 2 is semantic verbs tag extraction algorithm flow chart.
Specific implementation mode
As shown in Figure 1, 2, in intelligent network joins agricultural machinery system, since the scale of construction of video information is compared to other kinds of letter
Breath, such as control information, load information, text message etc. want huge more, therefore the transmission pressure of video information is this system
Need solve critical issue.The present invention solves under the congenital disadvantage of Bandwidth-Constrained, is come with existing hardware technology
Store and transmit a large amount of video data.And propose section bandization image transmitting processing method.Wherein label is that expression is semantic
A kind of effective means, the characteristics of according to video data, may be used semantic nouns label and semantic verbs label be combined
Mode expresses its semantic information, and it is Moving Objects in video, semantic verbs label pair that wherein semantic nouns label is corresponding
What is answered is the specific action behavior of Moving Objects, and the two combines the semantic label as video.
Semantic nouns tag extraction：
In terms of semantic nouns label, the operations such as shot segmentation and key-frame extraction are carried out to test video first, depending on
Then the key frame of frequency carries out Object identifying, you can obtain semantic nouns label to key frame.
Shot segmentation
Shot segmentation algorithm based on histogram
It realizes simple and convenient, and can obtain preferable effect to agricultural machinery monitor video, is the most universal segmentation side
Method.Algorithm based on histogram is typically that the gray scale of each pixel between consecutive frame, brightness are divided into N number of grade, then needle
Histogram is made to each grade statistical pixel number to compare, and provides the histogram of two images, then difference of histograms calculation formula
As follows：
Wherein, N is the sum of image frame pixel.What hm (i)-hn (i) was indicated is two video frame in this histogram list of i
Distance above position.Do not consider the location information of pixel based on histogram method, and uses the statistical value of its brightness and color, disadvantage
Different to structure and very similar two frame of histogram cause missing inspection and also for light variation relatively acutely in the case of, frame
Difference can be by prodigious interference.
Key-frame extraction
Key frame is the picture frame that can describe Video Key content.It reflects most intuitive and most worthy letter in camera lens
Breath.In order to enable users to understand at a glance the content of video as far as possible, can generally be taken when extracting key frame " it is peaceful it is wrong not
Conservatism less ".But this conservatism is taken, the number of key frames that may result in extraction is excessive, causes key frame
Dizzy multiple and redundancy.Therefore, suitable Key-frame Extraction Algorithm can extract most representative picture frame, and not generate too many superfluous
Remaining, this is the direction of key-frame extraction technique primary study.The Key-frame Extraction Algorithm based on cluster will be introduced below, is
Next research work lays the foundation.
Key-frame extraction based on cluster
Realize step：
The first step：If some camera lens Mi includes n picture frame, Mi={ N can be expressed as1... ..., Nn, wherein N1For
First frame, NnFor tail frame.If the similarity between adjacent two frame define be this adjacent two frames color histogram similarity (namely
It is histogram feature difference), predefine the density of a threshold value δ control cluster.
Second step:Calculate present frame NiWith it is existing some cluster barycenter between similarity, if the value be less than δ, the frame
Distance is larger between the cluster, therefore NiIt cannot be added in the cluster.If Ni and all existing cluster barycenter similarities are small
In δ, then it is its barycenter that Ni, which forms new a cluster and Ni,；Otherwise the frame is added to and is similarly spent in maximum cluster,
Keep the distance between the frame and this barycenter clustered minimum.
Third walks：After the n picture frame that camera lens Mi is included is referred to different clusters respectively by first two steps, so that it may with
Select key frame：The representative frame that the frame nearest from cluster barycenter is clustered as this, all clusters are extracted from each cluster
Representative frame just constitute the key frame of camera lens Mi.
Semantic verbs tag extraction：
In terms of semantic verbs label, the fortune of test video is obtained by the related algorithm of Moving Objects detect and track first
Dynamic track data, then investigates the semantic similarity between the track data and the track data of training set video of test video,
The semantic nouns label of test video and the marked content of training set video are contacted simultaneously, can be obtained verb after comprehensive analysis
Semantic label.
◆ movement locus extracts and analysis
When extracting movement locus, usually first carry out Moving Objects detection operation, then to the Moving Objects that detected into
Line trace, by the result of tracking with coordinate representation out to get movement locus.When extracting movement locus, usually first moved
Object detection operates, then to the Moving Objects that detected into line trace, by the result of tracking with coordinate representation out to get
Movement locus.
◆ video trains set construction method
Using the method for machine learning come when extracting video semanteme, construction video training set is very important a step
Suddenly.Way herein is first to choose a certain number of videos, extracts the related datas such as key frame and movement locus, is being moved
Longest one is chosen in track and its information, including the direction of motion, path curves are extracted by track post-processing operation
G-bar etc.；Then manual mark is carried out to these videos, marked content is mainly the motor behavior of object.Wherein, it closes
Key frame is used for video semantic nouns tag extraction, and motion track information and the content marked by hand are then for video semantic verbs mark
Label extraction is prepared.Its related data is extracted to each video in training set and carries out manual mark, all operations are completed
Afterwards, video training set is constructed and is finished.
◆ semantic verbs tag extraction algorithm
Semantic verbs tag extraction algorithm is based primarily upon machine learning thought, with data and the video to be analyzed of training set
Data are matched, and main process is as follows：The key frame and movement locus of video are extracted first, and original to what is extracted
Movement locus carries out post-processing operation, obtains the information such as its direction of motion and track G-bar；Then key frame is carried out pair
As identification operation, recognition result is added in video semantic nouns tag set；Then semantic nouns label and training set are used
In trail file carry out name-matches, find out corresponding trail file and calculate track semantic similarity, select k similarity
Highest trail file；Finally, it reads and marks file with k trail file training set of the same name, result is added to video and is moved
In word justice tag set.
In the process that agricultural machinery is advanced, information collection and video are carried out by vehicle-mounted camera and roadside field fixing camera
Monitoring, then the key frame and movement locus of video are extracted, and the original motion trajectory to extracting carries out post-processing behaviour
Make, obtains the information such as its direction of motion and track G-bar；Then Object identifying operation is carried out to key frame, by recognition result
It is added in video semantic nouns tag set；Then title is carried out with the trail file in semantic nouns label and training set
Match, find out corresponding trail file and calculate track semantic similarity, selects the highest trail file of k similarity；Finally, it reads
It takes and marks file with k trail file training set of the same name, result is added in video semantic verbs tag set.Thus
Video content can be fed back by transmitting semantic information.
Claims (1)
1. a kind of semantic-based agricultural machinery monitor video image section band Transmission system, for the process that agricultural machinery is advanced, feature
It is：Information collection and video monitoring are carried out by vehicle-mounted camera and roadside field fixing camera, then extracts video
Key frame and movement locus, and the original motion trajectory to extracting carries out post-processing operation, obtains its direction of motion and rail
Mark G-bar information；Then Object identifying operation is carried out to key frame, recognition result is added to video semantic nouns label
In set；Then name-matches are carried out with the trail file in semantic nouns label and training set, finds out corresponding trail file
And track semantic similarity is calculated, select the highest trail file of k similarity；Finally, it reads of the same name with k trail file
Training set marks file, and result is added in video semantic verbs tag set；Regarded by transmitting semantic information to feed back
Frequency content；
Video data is wherein fed back, its semanteme letter is expressed in such a way that semantic nouns label and semantic verbs label are combined
It is Moving Objects in video that breath, wherein semantic nouns label be corresponding, and it is the tool of Moving Objects that semantic verbs label is corresponding
Body action behavior, the two combine the semantic label as video；
(1), semantic nouns tag extraction：
In terms of semantic nouns label, the operations such as shot segmentation and key-frame extraction are carried out to test video first, obtain video
Then key frame carries out Object identifying, you can obtain semantic nouns label to key frame.
1) shot segmentation
Shot segmentation algorithm based on histogram, be the gray scale of each pixel between consecutive frame, brightness are divided into it is N number of etc.
Grade, then make histogram for each grade statistical pixel number and compare, the histogram of two images is provided, then difference of histograms meter
It is as follows to calculate formula：
Wherein, N is the sum of image frame pixel；What hm (i)-hn (i) was indicated is two video frame in this histogram map unit of i
The distance in face；Do not consider the location information of pixel based on histogram method, and uses the statistical value of its brightness and color；
2) key-frame extraction
Key-frame Extraction Algorithm based on cluster, the key-frame extraction based on cluster, implementation step：
The first step：If some camera lens Mi includes n picture frame, it is expressed as Mi={ N1... ..., Nn, wherein N1Headed by frame, NnFor
Tail frame；If the similarity between adjacent two frame define be this adjacent two frames color histogram similarity, that is to say histogram
Characteristic difference predefines the density of a threshold value δ control cluster；
Second step:Calculate present frame NiWith the similarity between some existing cluster barycenter, if the value is less than δ, the frame is poly- with this
Distance is larger between class, therefore NiIt cannot be added in the cluster；If NiIt is respectively less than δ with all existing cluster barycenter similarities, then Ni
Form new a cluster and NiFor its barycenter；Otherwise the frame is added to and is similarly spent in maximum cluster, make the frame with
The distance between the barycenter of this cluster minimum；
Third walks：After the n picture frame that camera lens Mi is included is referred to different clusters respectively by first two steps, key is selected
Frame：The representative frame that the frame nearest from cluster barycenter is clustered as this, the representative frame of all clusters are extracted from each cluster
Just constitute the key frame of camera lens Mi；
(2), semantic verbs tag extraction：
In terms of semantic verbs label, the movement rail of test video is obtained by the related algorithm of Moving Objects detect and track first
Then mark data investigate the semantic similarity between the track data and the track data of training set video of test video, simultaneously
The semantic nouns label of test video and the marked content of training set video are contacted, can be obtained semantic verbs after comprehensive analysis
Label；
1) movement locus extraction and analysis when extracting movement locus, usually first carry out Moving Objects detection operation, then to detection
Moving Objects out are into line trace, by the result of tracking with coordinate representation out to get movement locus；Extract movement locus
When, Moving Objects detection operation is usually first carried out, then the Moving Objects that detected are used the result of tracking into line trace
Coordinate representation is out to get movement locus；
2) video trains set construction method, and when extracting video semanteme, video training set is constructed using the method for machine learning；It does
Method is first selecting video, extracts the related data of key frame and movement locus, chooses longest one in movement locus, and
Its information is extracted by track post-processing operation, includes the G-bar of the direction of motion, path curves；Then these are regarded
Frequency carries out manual mark, and marked content is the motor behavior of object；Wherein, key frame is used for video semantic nouns tag extraction,
Motion track information and the content marked by hand are prepared for video semantic verbs tag extraction；To each of training set
Video all extracts its related data and carries out manual mark, and after the completion of all operations, video training set is constructed and finished；
3) semantic verbs tag extraction algorithm, semantic verbs tag extraction algorithm are to be based on machine learning thought, with training set
Data are matched with the data of video to be analyzed, and process is as follows：The key frame and movement locus of video are extracted first, and right
The original motion trajectory extracted carries out post-processing operation, obtains its direction of motion and track G-bar information；Then right
Key frame carries out Object identifying operation, and recognition result is added in video semantic nouns tag set；Then semantic nouns are used
Trail file in label and training set carries out name-matches, finds out corresponding trail file and calculates track semantic similarity,
Select the highest trail file of k similarity；Finally, it reads and marks file with k trail file training set of the same name, by result
It is added in video semantic verbs tag set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810181230.0A CN108595469A (en) | 2018-03-06 | 2018-03-06 | A kind of semantic-based agricultural machinery monitor video image section band Transmission system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810181230.0A CN108595469A (en) | 2018-03-06 | 2018-03-06 | A kind of semantic-based agricultural machinery monitor video image section band Transmission system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108595469A true CN108595469A (en) | 2018-09-28 |
Family
ID=63625725
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810181230.0A Pending CN108595469A (en) | 2018-03-06 | 2018-03-06 | A kind of semantic-based agricultural machinery monitor video image section band Transmission system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108595469A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110097094A (en) * | 2019-04-15 | 2019-08-06 | 天津大学 | It is a kind of towards personage interaction multiple semantic fusion lack sample classification method |
| CN111246176A (en) * | 2020-01-20 | 2020-06-05 | 北京中科晶上科技股份有限公司 | Video transmission method for realizing banding |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104239501A (en) * | 2014-09-10 | 2014-12-24 | 中国电子科技集团公司第二十八研究所 | Mass video semantic annotation method based on Spark |
-
2018
- 2018-03-06 CN CN201810181230.0A patent/CN108595469A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104239501A (en) * | 2014-09-10 | 2014-12-24 | 中国电子科技集团公司第二十八研究所 | Mass video semantic annotation method based on Spark |
Non-Patent Citations (3)
| Title |
|---|
| 任希: "视频语义提取方法研究", 《万方数据》 * |
| 李全栋: "基于内容视频检索中镜头分割与关键帧提取研究", 《万方数据》 * |
| 李全栋等: "一种改进的无监督聚类的关键帧提取算法", 《应用光学》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110097094A (en) * | 2019-04-15 | 2019-08-06 | 天津大学 | It is a kind of towards personage interaction multiple semantic fusion lack sample classification method |
| CN110097094B (en) * | 2019-04-15 | 2023-06-13 | 天津大学 | A Few-shot Classification Method Based on Multiple Semantic Fusion Oriented to Character Interaction |
| CN111246176A (en) * | 2020-01-20 | 2020-06-05 | 北京中科晶上科技股份有限公司 | Video transmission method for realizing banding |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103593464B (en) | Video fingerprint detecting and video sequence matching method and system based on visual features | |
| CN107230267B (en) | Intelligence In Baogang Kindergarten based on face recognition algorithms is registered method | |
| CN101719144A (en) | Method for segmenting and indexing scenes by combining captions and video image information | |
| CN104717468B (en) | Cluster scene intelligent monitoring method and system based on the classification of cluster track | |
| CN111738218B (en) | Human body abnormal behavior recognition system and method | |
| Mahmoodi et al. | Violence detection in videos using interest frame extraction and 3D convolutional neural network | |
| CN113378675A (en) | Face recognition method for simultaneous detection and feature extraction | |
| CN107688830A (en) | It is a kind of for case string and show survey visual information association figure layer generation method | |
| CN118968044B (en) | Multi-target detection tracking statistical method and device based on edge calculation | |
| Kuang et al. | A dual-branch neural network for DeepFake video detection by detecting spatial and temporal inconsistencies | |
| Yang et al. | A method of pedestrians counting based on deep learning | |
| CN104504733A (en) | Video abstraction method and system based on moving target detection | |
| CN108595469A (en) | A kind of semantic-based agricultural machinery monitor video image section band Transmission system | |
| Qin et al. | Application of video scene semantic recognition technology in smart video | |
| Wang et al. | Research on pedestrian detection based on multi-level fine-grained YOLOX algorithm | |
| KR102526263B1 (en) | Method and System for Auto Multiple Image Captioning | |
| CN118172713B (en) | Video tag identification method, device, computer equipment and storage medium | |
| He et al. | Chronological video synopsis via events rearrangement optimization | |
| Feng et al. | Similarity-and quality-guided relation learning for joint detection and tracking | |
| Wang et al. | Aerobics Action Recognition Algorithm Based on Three‐Dimensional Convolutional Neural Network and Multilabel Classification | |
| Zhu et al. | [Retracted] Basketball Object Extraction Method Based on Image Segmentation Algorithm | |
| Ma et al. | A lightweight neural learning algorithm for real-time facial feature tracking system via split-attention and heterogeneous convolution | |
| Keyvanpour et al. | Detection of individual activities in video sequences based on fast interference discovery and semi-supervised method | |
| CN116758364A (en) | Video classification method, device, electronic equipment and computer readable storage medium | |
| CN113709559B (en) | Video dividing method, device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180928 |
|
| RJ01 | Rejection of invention patent application after publication |