CN109299687A - A kind of fuzzy anomalous video recognition methods based on CNN - Google Patents
A kind of fuzzy anomalous video recognition methods based on CNN Download PDFInfo
- Publication number
- CN109299687A CN109299687A CN201811088546.1A CN201811088546A CN109299687A CN 109299687 A CN109299687 A CN 109299687A CN 201811088546 A CN201811088546 A CN 201811088546A CN 109299687 A CN109299687 A CN 109299687A
- Authority
- CN
- China
- Prior art keywords
- training
- fuzzy
- layer
- cnn
- video
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The fuzzy anomalous video recognition methods based on CNN that the invention discloses a kind of carries out key frame interception to video first.Truncated picture is pre-processed again.Then " label " manually is beaten to sample, further, the sample of tape label is divided into training set and sample set according to a certain percentage.Training parameter is set with the artificial intelligence frame of independent research again, builds CNN network, carries out fuzzy anomalous video recognition training.Network file after training identifies unknown sample collection, then is aided with manual sort's correction, expands training dataset, is iterated training.It solves screening and monitors the problems such as extremely labor intensive, coverage be not high, accuracy is not high, operation efficiency is low when obscuring anomalous video.
Description
Technical field
The present invention relates to field of video processing, in particular to a kind of fuzzy anomalous video recognition methods based on CNN.
Background technique
Video monitoring is increasingly taken seriously as a kind of important means of security protection, is also permitted in social every field
It applies more.On the one hand, the headend equipment in video monitoring system often breaks down and quality problems, generates dropout, partially
The problems such as color, colour bar.On the other hand, object is monitored in order to hide monitoring and can go deliberately to block monitoring device by object.For
Ten hundreds of monitor videos, it is clearly extremely labor intensive, and coverage that mode manually, which screens above-mentioned anomalous video,
Not high, accuracy is not high.Therefore in a kind of method of artificial intelligence, computer is allowed to go to identify that fuzzy anomalous video is necessary.
In recent years, it is based on the feature learning method of CNN (Convolution Neural Network, convolutional neural networks)
Immense success is achieved in terms of image classification, causes the very big concern of computer vision field.
The limitation of traditional Open Framework.Current all well-known artificial intelligence learning frameworks by the U.S. company and
Mechanism publication, function, feature, the tendentiousness of these artificial intelligence frames are held by these publication tissues, this is also entire
AI industry is difficult to realize commercial one of the major reasons in different field, because small-to-medium business can only almost use this
A little disclosed frames carry out the research and development of upper layer packaging type, are limited technical strength, and it is fixed according to real demand and different application depth to lack
The ability of artificial intelligence frame processed.
Summary of the invention
It is an object of the invention to: a kind of fuzzy anomalous video recognition methods based on CNN is provided, using artificial intelligence
Mode, release manual labor significantly, improve fuzzy anomalous identification accuracy and speed, and this method is realized in low side
Hardware, in the general situation of video definition efficiently, accurately identify fuzzy anomalous video.
The technical solution adopted by the invention is as follows:
A kind of fuzzy anomalous video recognition methods based on CNN, comprising the following steps:
S1, video sample is obtained;
S2, the S1 video sample obtained is handled;
S3, building CNN network structure, the CNN network structure are training network, this training network includes three convolution
Layer, two pond layers and an output layer;
S4, it step S2 treated sample is sent into the training network that step S3 is established carries out recognition training, trained
Obtained network file is carried out Classification and Identification to unknown sample collection with test program, then is aided with artificial correction by Cheng Hou, is changed
Generation training, obtains final CNN network.
The present invention is achieved by the following technical solutions: carrying out key frame interception to video first.Again by the figure of interception
As being pre-processed.Then " label " manually is beaten to sample, further, the sample of tape label is divided into training according to a certain percentage
Collection and sample set.Training parameter is set with the artificial intelligence frame of independent research again, builds CNN network, carries out fuzzy abnormal view
Frequency recognition training.Network file after training identifies unknown sample collection, then is aided with manual sort's correction, expands training
Data set is iterated training.
Further, the CNN network structure in the step S3 is by three convolutional layers, two pond layers and an output layer
Composition, in which:
First layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution;
The second layer is pond layer, according to maximum value pond;
Third layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution;
4th layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution;
Layer 5 is pond layer, using Relu activation primitive;
Layer 6 is output layer, using Softmax activation primitive.
Further, in the step S2 to the S1 method that is handled of video sample obtained the following steps are included:
The key frame of video sample in S201, interception step S1;
S202, sample classification is carried out to the key frame of video image of step S201 interception, while extracts test set and training
Collection;
S203, the key frame of video image of step S201 interception is pre-processed, reduces the calculation amount of subsequent processing.
Further, in the step S201 intercept step S1 in video sample key frame using OpenCV shape library into
Row image interception.Video by many frames image construction, therefore analyze video fog-level to a certain extent can be by analyzing
Certain frame image fog-level replaces.
Further, sample classification is carried out to the key frame of video image of step S201 interception in the step S202, it will
Blank screen, Hua Ping, smudgy screen, is classified as obscuring class with the image that various articles block camera blue screen, by remaining clear figure
As being classified as not obscuring class.
Further, carrying out pretreated method to image in the step S203 includes first carrying out gray proces, then
Compression of images is carried out, pixel value normalized is finally carried out.In order to reduce calculation amount, computational efficiency is improved, for fuzzy view
Frequency identifies this relatively simple feature, not high to image definition requirements, and color-map representation is handled at grayscale image.Then into
Row compression of images.Since fuzzy abnormal image is a kind of whole feature, to further increase efficiency, by image by high-resolution
It is compressed to low resolution.Finally carry out pixel value normalized.The grayscale image pixel value being converted between 0 to 255, for into
One step improves efficiency, and the picture element matrix of low resolution is uniformly processed on the section 0-1.
Further, trained training frame is iterated in the step S4 using the algorithm of dynamical learning rate and oneself
It is dynamic to sentence convergence algorithm.CNN training is begun to after the completion of image transformation.In order to realize intelligent training, alternatively referred to as on-hook training.This
CNN training frame uses the algorithm of dynamical learning rate and sentences convergence algorithm automatically.With the expansion of training round, learning rate meeting
According to the change of gradient in reversed gradient algorithm, dynamic is adjusted, and is gradually reduced to preset value.Change of gradient is in certain time
Interior variation is less than threshold value, then system will be completed voluntarily in deconditioning, mark training.After the completion of training, obtained with test program handle
The network file arrived carries out Classification and Identification to unknown sample collection, then is slightly aided with artificial correction, just can very easily expand training
Data set is iterated training.The network class accuracy rate finally realized can reach 99.8%.
It further, further include the method converted to training image, the method packet of image transformation in the step S4
Transverse translation, longitudinal translation are included, rotation changes picture contrast, changes brightness, setting fuzzy region range and fog-level
With adjustment noise size, and to every class transformation can all control transformation quantity in detail.Whether image mapping function is settable opens,
In the training of every wheel, if opening the transformation of training set image, training set sample will be done a round transformation again.Directly by sample
Quantity expands.EDS extended data set realizes industrial application.
In conclusion by adopting the above-described technical solution, the beneficial effects of the present invention are:
1. a kind of fuzzy anomalous video recognition methods based on CNN of the present invention solves screening and monitors fuzzy anomalous video
The problems such as Shi Jiqi labor intensive, coverage be not high, accuracy is not high, operation efficiency is low;
2. a kind of fuzzy anomalous video recognition methods based on CNN of the present invention, lower to hardware performance requirements, convenient for big rule
Mould is universal.
Detailed description of the invention
Examples of the present invention will be described by way of reference to the accompanying drawings, in which:
Fig. 1 is repetitive exercise procedure chart of the invention;
Fig. 2 is the CNN network structure that fuzzy diagnosis of the invention uses.
Specific embodiment
All features disclosed in this specification or disclosed all methods or in the process the step of, in addition to mutually exclusive
Feature and/or step other than, can combine in any way.
It elaborates below with reference to Fig. 1, Fig. 2 to the present invention.
Embodiment 1
A kind of fuzzy anomalous video recognition methods based on CNN, comprising the following steps:
S1, video sample is obtained;
S2, the S1 video sample obtained is handled;
S3, building CNN network structure, the CNN network structure are training network, this training network includes three convolution
Layer, two pond layers and an output layer;
S4, it step S2 treated sample is sent into the training network that step S3 is established carries out recognition training, trained
Obtained network file is carried out Classification and Identification to unknown sample collection with test program, then is aided with artificial correction by Cheng Hou, is changed
Generation training, obtains final CNN network.
The present invention is achieved by the following technical solutions: carrying out key frame interception to video first.Again by the figure of interception
As being pre-processed.Then " label " manually is beaten to sample, further, the sample of tape label is divided into training according to a certain percentage
Collection and sample set.Training parameter is set with the artificial intelligence frame of independent research again, builds CNN network, carries out fuzzy abnormal view
Frequency recognition training.Network file after training identifies unknown sample collection, then is aided with manual sort's correction, expands training
Data set is iterated training.
Embodiment 2
The present embodiment the difference from embodiment 1 is that, further, the CNN network structure in the step S3 is by three
Convolutional layer, two pond layers and an output layer composition, in which:
First layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution;
The second layer is pond layer, according to maximum value pond;
Third layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution;
4th layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution;
Layer 5 is pond layer, using Relu activation primitive;
Layer 6 is output layer, using Softmax activation primitive.
Further, in the step S2 to the S1 method that is handled of video sample obtained the following steps are included:
The key frame of video sample in S201, interception step S1;
S202, sample classification is carried out to the key frame of video image of step S201 interception, while extracts test set and training
Collection;
S203, the key frame of video image of step S201 interception is pre-processed, reduces the calculation amount of subsequent processing.
Further, in the step S201 intercept step S1 in video sample key frame using OpenCV shape library into
Row image interception.Video by many frames image construction, therefore analyze video fog-level to a certain extent can be by analyzing
Certain frame image fog-level replaces.
Further, sample classification is carried out to the key frame of video image of step S201 interception in the step S202, it will
Blank screen, Hua Ping, smudgy screen, is classified as obscuring class with the image that various articles block camera blue screen, by remaining clear figure
As being classified as not obscuring class.
Further, carrying out pretreated method to image in the step S203 includes first carrying out gray proces, then
Compression of images is carried out, pixel value normalized is finally carried out.In order to reduce calculation amount, computational efficiency is improved, for fuzzy view
Frequency identifies this relatively simple feature, not high to image definition requirements, and color-map representation is handled at grayscale image.Then into
Row compression of images.Since fuzzy abnormal image is a kind of whole feature, to further increase efficiency, by image by high-resolution
It is compressed to low resolution.Finally carry out pixel value normalized.The grayscale image pixel value being converted between 0 to 255, for into
One step improves efficiency, and the picture element matrix of low resolution is uniformly processed on the section 0-1.
Further, trained training frame is iterated in the step S4 using the algorithm of dynamical learning rate and oneself
It is dynamic to sentence convergence algorithm.CNN training is begun to after the completion of image transformation.In order to realize intelligent training, alternatively referred to as on-hook training.This
CNN training frame uses the algorithm of dynamical learning rate and sentences convergence algorithm automatically.With the expansion of training round, learning rate meeting
According to the change of gradient in reversed gradient algorithm, dynamic is adjusted, and is gradually reduced to preset value.Change of gradient is in certain time
Interior variation is less than threshold value, then system will be completed voluntarily in deconditioning, mark training.After the completion of training, obtained with test program handle
The network file arrived carries out Classification and Identification to unknown sample collection, then is slightly aided with artificial correction, just can very easily expand training
Data set is iterated training.The network class accuracy rate finally realized can reach 99.8%.
It further, further include the method converted to training image, the method packet of image transformation in the step S4
Transverse translation, longitudinal translation are included, rotation changes picture contrast, changes brightness, setting fuzzy region range and fog-level
With adjustment noise size, and to every class transformation can all control transformation quantity in detail.Whether image mapping function is settable opens,
In the training of every wheel, if opening the transformation of training set image, training set sample will be done a round transformation again.Directly by sample
Quantity expands.EDS extended data set realizes industrial application.
Embodiment 3
As shown in Figure 1, 2, specific step is as follows for the detailed training process of fuzzy video identification:
S1, vehicle-mounted monitoring video sample is obtained.
Sample data of the invention is generated based on third party's monitor supervision platform passenger-cargo carriage Vehicular video, the monitor video of acquisition
Sample resolution generally 352*288.It can certainly obtain by other means, resolution ratio is also not necessarily limited to 352*288.
S201, key frame of video interception.
After obtaining a large amount of vehicle-mounted monitoring videos, key frame interception is carried out to these videos.Video by many frames image
It constitutes, therefore the fog-level for analyzing video can be replaced to a certain extent by analyzing certain frame image fog-level.This method is adopted
Image interception is carried out with OpenCV shape library.The video length of selection opens figure in 60s or so, every section of video intercepting 5, and resolution ratio is
352*288。
S202, image pattern classification.
The image that the first step is intercepted out manually labels " ", and image two is divided.By blank screen, blue screen, Hua Ping, smudgy
Screen is classified as fuzzy class with the image that various articles block camera.It is classified as remaining clear image not obscure class.Obtain sample
Collection blurred picture 2000 is opened, and clear picture 6000 is opened.Then sample set is randomly selected 10%, as test set.It is converted
Before, training set (2000+6000) * (1-10%)=7200.
S203 image preprocessing.
By step 201 and 202, the sample image of 8000 classification is obtained, it is necessary to pre-process to image.It is advanced
Row gray proces.In order to reduce calculation amount, raising computational efficiency identifies this relatively simple feature for fuzzy video,
It is not high to image definition requirements, it is handled using by RGB255 color-map representation at grayscale image.It is criticized using OpenCV shape library
Amount processing.Then compression of images is carried out.It, will to further increase efficiency since fuzzy abnormal image is a kind of whole feature
Image is compressed to 88*72 by 352*288.Finally carry out pixel value normalized.The grayscale image pixel value being converted to is arrived 0
Between 255, to further increase efficiency, the picture element matrix of 88*72 is uniformly processed on the section 0-1.
S3, CNN network structure is constructed in the neural network framework of independent research.
As shown in Fig. 2, the present invention is trained using CNN network structure.This training network has six layers, three convolutional layers two
A pond layer and an output layer composition.
First layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution, and it is 3x3 that 16 sizes, which are arranged,
Convolution kernel, moving step length 1.Input dimension of picture is 72*88, and output dimension of picture is 72*88.
The second layer is pond layer, and according to maximum value pond, the core that 16 sizes are 2x2, moving step length 2 does not cover pond
Change.Input dimension of picture is 72*88, and output dimension of picture is 36*44.
Third layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution, and it is 3x3 that 16 sizes, which are arranged,
Convolution kernel, moving step length 1.Input dimension of picture is 36*44, and output dimension of picture is 36*44.
4th layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution, and it is 3x3 that 32 sizes, which are arranged,
Convolution kernel, moving step length 1.Input dimension of picture is 36*44, and output dimension of picture is 36*44.
Layer 5 is pond layer, and using Relu activation primitive, the core that 32 sizes are 2x2, moving step length 2 is not covered.
Input dimension of picture is 36*44, and output dimension of picture is 18*22.
Layer 6 is output layer, and using Softmax activation primitive, number of classifying is 2.
S401, the transformation of training set image.
The method of image transformation has: lateral, longitudinal translation, rotation change picture contrast, change brightness, and mould is arranged
Regional scope and fog-level are pasted, noise size is adjusted.And transformation quantity can all be controlled in detail to the transformation of every class.Image transformation
Whether function is settable opens, and in the training of every wheel, if opening the transformation of training set image, will do again training set sample
One round transformation.Sample size is directly expanded to 80000 from 7200.EDS extended data set realizes industrial application.
S4, repetitive exercise.
CNN training is begun to after the completion of image transformation.In order to realize intelligent training, alternatively referred to as on-hook training.This CNN instruction
Practice frame to use the algorithm of dynamical learning rate and sentence convergence algorithm automatically.With the expansion of training round, learning rate can basis
Change of gradient in reversed gradient algorithm and dynamic adjusts, gradually reduce to preset value.Change of gradient becomes within a certain period of time
Change and be less than threshold value, then system will be completed voluntarily in deconditioning, mark training.After the completion of training, with test program obtaining
Network file carries out Classification and Identification to unknown sample collection, then is slightly aided with artificial correction, just can very easily expand training data
Collection, is iterated training.The network class accuracy rate finally realized can reach 99.8%.
Embodiment 4
The program designed using this method, using the calculated performance of the x86 framework CPU of 4 core 2GHz as under standard, for CIF
Resolution ratio Vehicular video intercepts five figures by standard of 60s duration, and recognition speed is less than 100ms.
The above, only the preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto, it is any
Those skilled in the art within the technical scope disclosed by the invention, can without the variation that creative work is expected or
Replacement, should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be limited with claims
Subject to fixed protection scope.
Claims (8)
1. a kind of fuzzy anomalous video recognition methods based on CNN, it is characterised in that: the following steps are included:
S1, video sample is obtained;
S2, the S1 video sample obtained is handled;
S3, building CNN network structure, the CNN network structure are training network, this training network includes three convolutional layers, two
A pond layer and an output layer;
S4, it step S2 treated sample is sent into the training network that step S3 is established carries out recognition training, after the completion of training,
Obtained network file is carried out Classification and Identification to unknown sample collection with test program, then is aided with artificial correction, is iterated instruction
Practice, obtains final CNN network.
2. a kind of fuzzy anomalous video recognition methods based on CNN according to claim 1, it is characterised in that: the step
CNN network structure in rapid S3 is made of three convolutional layers, two pond layers and an output layer, in which:
First layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution;
The second layer is pond layer, according to maximum value pond;
Third layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution;
4th layer is convolutional layer, and using Relu activation primitive, convolution mode is band edge convolution;
Layer 5 is pond layer, using Relu activation primitive;
Layer 6 is output layer, using Softmax activation primitive.
3. a kind of fuzzy anomalous video recognition methods based on CNN according to claim 1, it is characterised in that: the step
The S1 video sample obtained is handled in rapid S2 method the following steps are included:
The key frame of video sample in S201, interception step S1;
S202, sample classification is carried out to the key frame of video image of step S201 interception, while extracts test set and training set;
S203, the key frame of video image of step S201 interception is pre-processed, reduces the calculation amount of subsequent processing.
4. a kind of fuzzy anomalous video recognition methods based on CNN according to claim 3, it is characterised in that: the step
The key frame that video sample in step S1 is intercepted in rapid S201 carries out image interception using OpenCV shape library.
5. a kind of fuzzy anomalous video recognition methods based on CNN according to claim 3, it is characterised in that: the step
Sample classification is carried out to the key frame of video image of step S201 interception in rapid S202, by blank screen, blue screen, Hua Ping, smudgy
Screen is classified as fuzzy class with the image that various articles block camera, is classified as remaining clear image not obscure class.
6. a kind of fuzzy anomalous video recognition methods based on CNN according to claim 3, it is characterised in that: the step
Carrying out pretreated method to image in rapid S203 includes first carrying out gray proces, then carries out compression of images, finally carries out picture
Element value normalized.
7. a kind of fuzzy anomalous video recognition methods based on CNN according to claim 1, it is characterised in that: the step
The training frame that recognition training is carried out in rapid S4 uses the algorithm of dynamical learning rate and sentences convergence algorithm automatically.
8. a kind of fuzzy anomalous video recognition methods based on CNN according to claim 1, it is characterised in that: the step
It further include the method converted to training image in rapid S4, the method for image transformation includes transverse translation, longitudinal translation, rotation
Turn, change picture contrast, change brightness, setting fuzzy region range and fog-level and adjustment noise size, and is right
Every class transformation can all control transformation quantity in detail.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811088546.1A CN109299687A (en) | 2018-09-18 | 2018-09-18 | A kind of fuzzy anomalous video recognition methods based on CNN |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811088546.1A CN109299687A (en) | 2018-09-18 | 2018-09-18 | A kind of fuzzy anomalous video recognition methods based on CNN |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109299687A true CN109299687A (en) | 2019-02-01 |
Family
ID=65163579
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811088546.1A Withdrawn CN109299687A (en) | 2018-09-18 | 2018-09-18 | A kind of fuzzy anomalous video recognition methods based on CNN |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109299687A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109886963A (en) * | 2019-03-28 | 2019-06-14 | 北京奇艺世纪科技有限公司 | A kind of image processing method and system |
CN110348299A (en) * | 2019-06-04 | 2019-10-18 | 上海交通大学 | The recognition methods of three-dimension object |
CN110866512A (en) * | 2019-11-21 | 2020-03-06 | 南京大学 | Monitoring camera shielding detection method based on video classification |
CN112395903A (en) * | 2019-08-12 | 2021-02-23 | 顺丰科技有限公司 | Method and device for determining spatial characteristics, network equipment and storage medium |
CN112818735A (en) * | 2020-12-03 | 2021-05-18 | 中国舰船研究设计中心 | Article identification-based vessel spare part identification method |
CN113111766A (en) * | 2021-04-09 | 2021-07-13 | 湖南科技大学 | Artificial intelligent feedback control method for coagulation dosing signal |
CN114494979A (en) * | 2022-03-29 | 2022-05-13 | 杭州定川信息技术有限公司 | Method for video recognition of ecological flow discharge |
CN116112645A (en) * | 2023-04-11 | 2023-05-12 | 重庆华悦生态环境工程研究院有限公司深圳分公司 | Multi-image transmission method and device for reservoir environment |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101900991A (en) * | 2010-07-02 | 2010-12-01 | 长沙理工大学 | Composite PID (Proportion Integration Differentiation) neural network control method based on nonlinear dynamic factor |
CN103763515A (en) * | 2013-12-24 | 2014-04-30 | 浙江工业大学 | Video anomaly detection method based on machine learning |
CN105955708A (en) * | 2016-05-09 | 2016-09-21 | 西安北升信息科技有限公司 | Sports video lens classification method based on deep convolutional neural networks |
CN106203501A (en) * | 2016-07-08 | 2016-12-07 | 青岛大学 | Video image based on machine learning obscures method for detecting abnormality |
CN106548208A (en) * | 2016-10-28 | 2017-03-29 | 杭州慕锐科技有限公司 | A kind of quick, intelligent stylizing method of photograph image |
CN107341805A (en) * | 2016-08-19 | 2017-11-10 | 北京市商汤科技开发有限公司 | Background segment and network model training, image processing method and device before image |
CN108337505A (en) * | 2018-01-30 | 2018-07-27 | 百度在线网络技术(北京)有限公司 | Information acquisition method and device |
-
2018
- 2018-09-18 CN CN201811088546.1A patent/CN109299687A/en not_active Withdrawn
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101900991A (en) * | 2010-07-02 | 2010-12-01 | 长沙理工大学 | Composite PID (Proportion Integration Differentiation) neural network control method based on nonlinear dynamic factor |
CN103763515A (en) * | 2013-12-24 | 2014-04-30 | 浙江工业大学 | Video anomaly detection method based on machine learning |
CN105955708A (en) * | 2016-05-09 | 2016-09-21 | 西安北升信息科技有限公司 | Sports video lens classification method based on deep convolutional neural networks |
CN106203501A (en) * | 2016-07-08 | 2016-12-07 | 青岛大学 | Video image based on machine learning obscures method for detecting abnormality |
CN107341805A (en) * | 2016-08-19 | 2017-11-10 | 北京市商汤科技开发有限公司 | Background segment and network model training, image processing method and device before image |
CN106548208A (en) * | 2016-10-28 | 2017-03-29 | 杭州慕锐科技有限公司 | A kind of quick, intelligent stylizing method of photograph image |
CN108337505A (en) * | 2018-01-30 | 2018-07-27 | 百度在线网络技术(北京)有限公司 | Information acquisition method and device |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109886963A (en) * | 2019-03-28 | 2019-06-14 | 北京奇艺世纪科技有限公司 | A kind of image processing method and system |
CN109886963B (en) * | 2019-03-28 | 2021-08-24 | 北京奇艺世纪科技有限公司 | Image processing method and system |
CN110348299A (en) * | 2019-06-04 | 2019-10-18 | 上海交通大学 | The recognition methods of three-dimension object |
CN110348299B (en) * | 2019-06-04 | 2022-06-10 | 上海交通大学 | Method for recognizing three-dimensional object |
CN112395903A (en) * | 2019-08-12 | 2021-02-23 | 顺丰科技有限公司 | Method and device for determining spatial characteristics, network equipment and storage medium |
CN110866512A (en) * | 2019-11-21 | 2020-03-06 | 南京大学 | Monitoring camera shielding detection method based on video classification |
CN112818735A (en) * | 2020-12-03 | 2021-05-18 | 中国舰船研究设计中心 | Article identification-based vessel spare part identification method |
CN113111766A (en) * | 2021-04-09 | 2021-07-13 | 湖南科技大学 | Artificial intelligent feedback control method for coagulation dosing signal |
CN114494979A (en) * | 2022-03-29 | 2022-05-13 | 杭州定川信息技术有限公司 | Method for video recognition of ecological flow discharge |
CN116112645A (en) * | 2023-04-11 | 2023-05-12 | 重庆华悦生态环境工程研究院有限公司深圳分公司 | Multi-image transmission method and device for reservoir environment |
CN116112645B (en) * | 2023-04-11 | 2023-11-21 | 重庆华悦生态环境工程研究院有限公司深圳分公司 | Multi-image transmission method and device for reservoir environment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109299687A (en) | A kind of fuzzy anomalous video recognition methods based on CNN | |
Fan et al. | No reference image quality assessment based on multi-expert convolutional neural networks | |
CN114359727B (en) | Tea disease identification method and system based on lightweight optimization Yolo v4 | |
CN108780508A (en) | System and method for normalized image | |
CN107909005A (en) | Personage's gesture recognition method under monitoring scene based on deep learning | |
CN112215795B (en) | Intelligent detection method for server component based on deep learning | |
WO2019114145A1 (en) | Head count detection method and device in surveillance video | |
CN108052950B (en) | MIA-based fused magnesia furnace dynamic flame segmentation and feature extraction method | |
CN112990392A (en) | New material floor defect target detection system based on improved YOLOv5 algorithm | |
CN105809643B (en) | A kind of image enchancing method based on adaptive block channel extrusion | |
CN109472193A (en) | Method for detecting human face and device | |
US20180096481A1 (en) | Process to isolate object of interest in image | |
US10891478B2 (en) | Method for correction of the eyes image using machine learning and method for machine learning | |
CN112686172A (en) | Method and device for detecting foreign matters on airport runway and storage medium | |
CN107527045A (en) | A kind of human body behavior event real-time analysis method towards multi-channel video | |
CN109389569A (en) | Based on the real-time defogging method of monitor video for improving DehazeNet | |
CN109325533A (en) | A kind of artificial intelligence frame progress CNN repetitive exercise method | |
CN110458027A (en) | One kind being based on marblized fresh meat stage division, system and device | |
CN114170227A (en) | Product surface defect detection method, device, equipment and storage medium | |
CN111582056A (en) | Automatic detection method for fire-fighting equipment in oil unloading operation site of gas station | |
CN112561813A (en) | Face image enhancement method and device, electronic equipment and storage medium | |
CN111104976B (en) | Blue algae coverage calculating method based on time sequence images | |
CN114708532A (en) | Monitoring video quality evaluation method, system and storage medium | |
CN115115950A (en) | Unmanned aerial vehicle image duplicate checking method based on image histogram features | |
CN115019340A (en) | Night pedestrian detection algorithm based on deep learning |
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 | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20190201 |
|
WW01 | Invention patent application withdrawn after publication |