CN107025443A - Stockyard smoke monitoring and on-time model update method based on depth convolutional neural networks - Google Patents
Stockyard smoke monitoring and on-time model update method based on depth convolutional neural networks Download PDFInfo
- Publication number
- CN107025443A CN107025443A CN201710221097.2A CN201710221097A CN107025443A CN 107025443 A CN107025443 A CN 107025443A CN 201710221097 A CN201710221097 A CN 201710221097A CN 107025443 A CN107025443 A CN 107025443A
- Authority
- CN
- China
- Prior art keywords
- model
- convolutional neural
- neural networks
- smoke
- depth convolutional
- 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; 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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
Abstract
Disclosure herein refer to a kind of stockyard smoke monitoring based on depth convolutional neural networks and on-time model update method, comprise the following steps:A, using instrument smog video switched into image sequence, and carry out smog category mark;B, the more abstract high-level characteristic of smog is obtained using depth convolutional neural networks, and be iterated training Optimized model parameter, iterative model is evaluated according to loss function, optimal models is selected;The online updating of situation implementation model is failed to report, reported by mistake to c, basis.The present invention carries out effective feature extraction by depth convolutional neural networks to smog image, and it can realize cigarette fog real-time monitoring and model modification under video monitoring, further lift accuracy of detection, simple to operate, quick effective, and has the advantages that higher robustness.
Description
Technical field
The present invention relates to a kind of smoke monitoring method, especially a kind of stockyard smog prison based on depth convolutional neural networks
Survey and on-time model update method.
Background technology
Fire is to endanger a class of public security and social development the most universal and mostly important disaster.Fire is not only ruined
Bad material property, also seriously threatens the life and health and safety of people, once occur that irremediable loss can be caused to us.
Fire explosion occurs on August 12nd, 2015, PORT OF TIANJIN, causes 165 people wrecked, and 798 people are injured, the direct economy appraised and decided
Loss is up to 68.66 hundred million yuan.The safety problem in stockyard, especially fire alarm are always an important class in security against fire field
Topic.If can early be found in fire early period of origination, and put out and can just reduce the danger that fire is caused to greatest extent
Evil." cigarette begins for fire ", the early stage that fire occurs would generally produce smog, so if smog can be timely detected, then
Fire alarm earlier can be provided, casualties and property loss is reduced.
The detection of smog is essentially a pattern recognition problem, and the abundant acquisition of smoke characteristics is accuracy of identification
With the guarantee of model generalization ability.It is to utilize single features (such as color characteristic) mostly, still in existing smog recognition methods
Because smog has the features such as shape, color are not fixed in itself, the problem of feature fully extracts difficulty is there is, model is caused
Nicety of grading is not good, and influence smog is detected in real time, thus is necessary to find new feature extraction means.
Color, texture, motion detection are three kinds of principal characters that current smog detection method is used.Color characteristic is the most
Extensive visual signature, is typically to find suitable color space to carry out smoke region segmentation, such as RGB color, HIS at present
Color space etc..Textural characteristics are a kind of visual signatures for reflecting homogeneity phenomenon in image, mainly there is statistical textural characteristics, mould
Four kinds of type type textural characteristics, signal transacting type textural characteristics, structural type textural characteristics etc..Motion feature first has to extract motion arrow
Amount, such as correlation technique based on Block- matching, optical flow method.On the one hand, it is difficult to reject the dry of class cigarette object when these methods are used
Disturb, such as when using color characteristic the interference of cloud and mist.On the other hand, these aspects can not preferably describe the essence of smog,
Adaptability to environmental change is not strong, such as influence of the illumination to motion detection, it is impossible to ensure the feature selected for difference
Applicability under the environment of stockyard.Smoke characteristics how are fully being extracted, and are meeting smoke monitoring under the conditions of video monitoring simultaneously
Real-time and high-class precision still suffer from improved space.
The content of the invention
The purpose of the present invention overcomes the deficiencies in the prior art there is provided a kind of based on depth convolutional neural networks
Stockyard smoke monitoring and on-time model update method, it can be realized carries out Real-time Smoke under the conditions of the circumstance video monitoring of stockyard
Detection, while according to the online updating for failing to report, reporting by mistake situation implementation model, further lift accuracy of detection, realize fire alarm,
Property loss is reduced, stability is higher, use more facilitates.
According to the technical scheme that provides of the present invention, the stockyard smoke monitoring based on depth convolutional neural networks and online
Model update method, it is characterized in that, comprise the following steps:
(1) video data of collection switched into sequence of pictures, and carries out smoke region position, category to mark, by position,
Category information saves as category file, and picture constitutes initial positive and negative sample set Y with position, two files of category information;
(2) the initial positive and negative sample set Y for obtaining step (1) is divided into training set Y in the ratio of setting11, test set Y2With
Checking collection Y12;By training set Y11Collect Y with checking12It is put into general depth convolutional neural networks model to be trained, to Universal Die
Type carries out small parameter perturbations, is constantly iterated training counting loss function simultaneously, multiple initial models are selected according to loss function,
By to test set Y2Tested, optimal initial model X is selected according to ROC curve;
(3) the initial model X obtained using step (2), Smoke Detection is carried out to video monitoring under the environment of stockyard, is protected simultaneously
Testing result and the value of the confidence are deposited, positive and negative sample set is updated according to testing result at regular intervals and carries out model modification and obtain more
New model X', is evaluated by test model precision.
Further, the step (1) comprises the steps of:
Step 1.1, the video data of collection switched into sequence of pictures, will there is the region of smog in all pictures with rectangle
The form of frame, which is marked out, to be come, and stamps Smoke category;
The position of rectangle frame and category information are stored in position, class with standard VOC2007 forms in step 1.2, picture
Mark in file, name and corresponding picture are of the same name, picture constitutes initial positive negative sample with position, two files of category information
Collect Y.
Further, the step (2) comprises the steps of:
Step 2.1, the initial positive and negative sample set Y of general press 9:1 ratio is divided into training set Y in a random way1And test set
Y2, by training set Y1It is same to press 9:1 ratio random fashion is divided into training set Y11With checking collection Y12;
Step 2.2, by training set Y1It is put into general depth convolutional neural networks model and trains, training set Y11For micro-
Mode transfer shape parameter, checking collection Y12For determining model structure, while being missed according to the classification of the testing result of model and actual result
Poor LclsAnd position deviation LlocWeighted calculation loss function evaluates training pattern, and loss function L calculation formula is:
The low model of step 2.3, the several loss functions of selection uses test set Y as initial model2Carry out test model essence
Degree, evaluation criteria is defined by the ROC curve of rate of failing to report and rate of false alarm, selects best model to be used as initial model X.
Further, the step (3) comprises the steps of:
Step 3.1, using initial model X, Real-time Smoke detection is carried out to video monitoring under the environment of stockyard, detected in real time
Interface display site environment, if model thinks there is smog appearance, marks smoke region and shows the value of the confidence, preservation model detection
The smoke region arrived and the value of the confidence;
Step 3.2, it regard the value of the confidence as negative sample collection Z for 0~0.2 testing result at regular intervals1, by the value of the confidence
Positive sample collection Z is used as more than 0.2 testing result2, and by negative sample collection Z1, positive sample collection Z2With training set Y1Merge into instruction
Practice collection Z;
Step 3.3, training set Z is put into general depth convolutional neural networks model and trained, in repeat step (2)
Step 2.2, step 2.3 obtain more new model X', more will be compared with initial model X new model X', test set Y2It is used as test specimens
This, selects preferable model as initial model X and proceeds Real-time Smoke detection.
Stockyard smoke monitoring and on-time model update method of the present invention based on depth convolutional neural networks, Neng Goushi
Carry out Real-time Smoke detection under the conditions of present stockyard circumstance video monitoring, at the same according to failing to report, report by mistake situation implementation model
Line updates, and further lifts accuracy of detection, realizes fire alarm, reduces property loss, and stability is higher, and use more facilitates.
Brief description of the drawings
Fig. 1 is stockyard smoke monitoring of the present invention and the flow chart of on-time model update method.
Fig. 2 is the mark schematic diagram of smog picture.
Fig. 3 schemes for the picture feature visualization of depth convolutional neural networks model extraction.
Fig. 4 is model rate of failing to report and the ROC curve figure of rate of false alarm.
Fig. 5 schemes for the Real-time Smoke detection of stockyard circumstance video monitoring.
Embodiment
With reference to specific accompanying drawing, the invention will be further described.
As shown in figure 1, stockyard smoke monitoring and on-time model of the present invention based on depth convolutional neural networks update
Method, comprises the following steps:
(1) the smog video of collection is switched into sequence of pictures, there is no particular/special requirement to crossover tool and picture format, this
In embodiment, crossover tool is using self-control python instruments, and picture format is jpg forms;
(2) mark out to come in the form of rectangle frame to there is the region of smog in picture, and stamp Smoke category (such as
Shown in Fig. 2), for class cigarette object, class is designated as likesmoke, does not have to annotation tool in particular/special requirement, the present embodiment, mark
Instrument is using self-control python instruments;
(3) position of rectangle frame in the picture marked and category information are preserved in place with standard VOC2007 forms
Put, in category file, name and corresponding picture are of the same name;
(4) initial positive and negative sample set Y is made up of picture with position, two files of category information, and JPEGImages is figure
Piece collection, Annotations is position, category message file folder;
(5) initial positive and negative sample set Y is pressed 9:1 ratio is divided into training set Y in a random way1With test set Y2, will instruct
Practice collection Y1It is same to press 9:1 ratio random fashion is divided into training set Y11With checking collection Y12;Test.txt is test set,
Train.txt is training set, and val.txt is checking collection, and trainval.txt is training and checking collection;
(6) by training set Y1It is put into general depth convolutional neural networks model and trains, obtained a variety of highers is special
Levy visualization figure as shown in Figure 3;
(7) according to the error in classification L of the testing result of model and actual resultclsAnd position deviation LlocWeighted calculation is damaged
Lose function and evaluate training pattern, loss function L calculation formula is:
Wherein, NclsFor the number of picture in pictures;NlocFor the number of the rectangle frame detected;P represent testing result with
The judgement of actual result, is correctly 1, mistake is 0;T represents the position of rectangle frame;
The low model of several loss functions is chosen as initial model, test set Y is used2Carry out test model precision, assess
Standard is defined (as shown in Figure 4) by the ROC curve of rate of failing to report and rate of false alarm, curve closer to origin illustrate model rate of false alarm and
Rate of failing to report is lower, selects best model to be used as initial model X;
(8) initial model X is utilized, Smoke Detection is carried out to video monitoring under the environment of stockyard, in detection interface display in real time
Site environment, if model thinks there is smog appearance, marks smoke region and shows the value of the confidence (as shown in Figure 5), preservation model
The smoke region detected and the value of the confidence;
It is considered herein that the value of the confidence is, than the object of relatively similar smog, these objects to be regarded for 0~0.2 testing result
Negative sample, the value of the confidence is smoke target for more than 0.2 testing result, and these objects are regarded into positive sample, updates positive and negative samples
Re -training more new model afterwards, if more new model X' compared with initial model X to initial positive and negative sample set Y Detection results more
It is good, then current model modification is effective;Comprise the following steps that:
A, it regard the value of the confidence as negative sample collection Z for 0~0.2 testing result at regular intervals1, it is 0.2 by the value of the confidence
Testing result above is used as positive sample collection Z2, and by negative sample collection Z1, positive sample collection Z2With training set Y1Merge into training set Z;
B, by training set Z be put into general depth convolutional neural networks model train, duplication model training step update
Model X', more will be compared new model X', test set Y with initial model X2As test sample, preferable model is selected as first
Beginning model X proceeds Real-time Smoke detection.
Claims (4)
1. a kind of stockyard smoke monitoring and on-time model update method based on depth convolutional neural networks, it is characterized in that, including
Following steps:
(1) video data of collection is switched into sequence of pictures, and position, category mark is carried out to smoke region, by position, category
Information saves as category file, and picture constitutes initial positive and negative sample set Y with position, two files of category information;
(2) the initial positive and negative sample set Y for obtaining step (1) is divided into training set Y in the ratio of setting11, test set Y2And checking
Collect Y12;By training set Y11Collect Y with checking12It is put into general depth convolutional neural networks model to be trained, universal model is entered
Row small parameter perturbations, are constantly iterated training counting loss function simultaneously, select multiple initial models according to loss function, pass through
To test set Y2Tested, optimal initial model X is selected according to ROC curve;
(3) the initial model X obtained using step (2), Smoke Detection is carried out to video monitoring under the environment of stockyard, while preserving inspection
Result and the value of the confidence are surveyed, positive and negative sample set is updated according to testing result at regular intervals and model modification is carried out obtains updating mould
Type X', is evaluated by test model precision.
2. stockyard smoke monitoring and on-time model update method as claimed in claim 1 based on depth convolutional neural networks,
It is characterized in that:The step (1) comprises the steps of:
Step 1.1, the video data of collection switched into sequence of pictures, will there is the region of smog in all pictures with rectangle frame
Form, which is marked out, to be come, and stamps Smoke category;
The position of rectangle frame and category information are stored in position, category text with standard VOC2007 forms in step 1.2, picture
In part folder, name and corresponding picture are of the same name, and picture constitutes initial positive and negative sample set Y with position, two files of category information.
3. stockyard smoke monitoring and on-time model update method as claimed in claim 1 based on depth convolutional neural networks,
It is characterized in that:The step (2) comprises the steps of:
Step 2.1, the initial positive and negative sample set Y of general press 9:1 ratio is divided into training set Y in a random way1With test set Y2, will
Training set Y1It is same to press 9:1 ratio random fashion is divided into training set Y11With checking collection Y12;
Step 2.2, by training set Y1It is put into general depth convolutional neural networks model and trains, training set Y11For finely tuning mould
Shape parameter, checking collection Y12For determining model structure, while according to the error in classification L of the testing result of model and actual resultcls
And position deviation LlocWeighted calculation loss function evaluates training pattern, and loss function L calculation formula is:
The low model of step 2.3, the several loss functions of selection uses test set Y as initial model2Carry out test model precision, comment
Estimate standard to be defined by the ROC curve of rate of failing to report and rate of false alarm, select best model to be used as initial model X.
4. stockyard smoke monitoring and on-time model update method as claimed in claim 3 based on depth convolutional neural networks,
It is characterized in that:The step (3) comprises the steps of:
Step 3.1, using initial model X, Real-time Smoke detection is carried out to video monitoring under the environment of stockyard, at detection interface in real time
Displaying scene environment, if model thinks there is smog appearance, marks smoke region and shows the value of the confidence, what preservation model was detected
Smoke region and the value of the confidence;
Step 3.2, it regard the value of the confidence as negative sample collection Z for 0~0.2 testing result at regular intervals1, it is 0.2 by the value of the confidence
Testing result above is used as positive sample collection Z2, and by negative sample collection Z1, positive sample collection Z2With training set Y1Merge into training set Z;
Step 3.3, training set Z is put into general depth convolutional neural networks model and trained, the step in repeat step (2)
2.2nd, step 2.3 obtains more new model X', more will be compared with initial model X new model X', test set Y2It is used as test sample, choosing
Preferable model is selected as initial model X and proceeds Real-time Smoke detection.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710221097.2A CN107025443A (en) | 2017-04-06 | 2017-04-06 | Stockyard smoke monitoring and on-time model update method based on depth convolutional neural networks |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710221097.2A CN107025443A (en) | 2017-04-06 | 2017-04-06 | Stockyard smoke monitoring and on-time model update method based on depth convolutional neural networks |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107025443A true CN107025443A (en) | 2017-08-08 |
Family
ID=59527849
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710221097.2A Pending CN107025443A (en) | 2017-04-06 | 2017-04-06 | Stockyard smoke monitoring and on-time model update method based on depth convolutional neural networks |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107025443A (en) |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107517251A (en) * | 2017-08-16 | 2017-12-26 | 北京小度信息科技有限公司 | Information-pushing method and device |
CN107730488A (en) * | 2017-09-21 | 2018-02-23 | 滨州学院 | A kind of method planted using unmanned plane low-altitude remote sensing image automatic detection opium poppy |
CN108256496A (en) * | 2018-02-01 | 2018-07-06 | 江南大学 | A kind of stockyard smog detection method based on video |
CN108334902A (en) * | 2018-02-02 | 2018-07-27 | 北京华纵科技有限公司 | A kind of track train equipment room smog fireproof monitoring method based on deep learning |
CN108428324A (en) * | 2018-04-28 | 2018-08-21 | 温州大学激光与光电智能制造研究院 | The detection device of smog in a kind of fire scenario based on convolutional network |
CN108537215A (en) * | 2018-03-23 | 2018-09-14 | 清华大学 | A kind of flame detecting method based on image object detection |
CN108764264A (en) * | 2018-03-16 | 2018-11-06 | 深圳中兴网信科技有限公司 | Smog detection method, smoke detection system and computer installation |
CN108985192A (en) * | 2018-06-29 | 2018-12-11 | 东南大学 | A kind of video smoke recognition methods based on multitask depth convolutional neural networks |
CN109034049A (en) * | 2018-07-23 | 2018-12-18 | 北京密境和风科技有限公司 | The recognition methods of dancing video and device |
CN109147254A (en) * | 2018-07-18 | 2019-01-04 | 武汉大学 | A kind of video outdoor fire disaster smog real-time detection method based on convolutional neural networks |
CN109271906A (en) * | 2018-09-03 | 2019-01-25 | 五邑大学 | A kind of smog detection method and its device based on depth convolutional neural networks |
CN109766780A (en) * | 2018-12-20 | 2019-05-17 | 武汉理工大学 | A kind of ship smog emission on-line checking and method for tracing based on deep learning |
CN109766872A (en) * | 2019-01-31 | 2019-05-17 | 广州视源电子科技股份有限公司 | Image-recognizing method and device |
CN109961042A (en) * | 2019-03-22 | 2019-07-02 | 中国人民解放军国防科技大学 | Smoke detection method combining deep convolutional neural network and visual change diagram |
CN110096942A (en) * | 2018-12-20 | 2019-08-06 | 北京以萨技术股份有限公司 | A kind of Smoke Detection algorithm based on video analysis |
CN111783968A (en) * | 2020-06-30 | 2020-10-16 | 山东信通电子股份有限公司 | Power transmission line monitoring method and system based on cloud edge cooperation |
CN111986199A (en) * | 2020-09-11 | 2020-11-24 | 征图新视(江苏)科技股份有限公司 | Unsupervised deep learning-based wood floor surface flaw detection method |
CN112396042A (en) * | 2021-01-20 | 2021-02-23 | 鹏城实验室 | Real-time updated target detection method and system, and computer-readable storage medium |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106097346A (en) * | 2016-06-13 | 2016-11-09 | 中国科学技术大学 | A kind of video fire hazard detection method of self study |
-
2017
- 2017-04-06 CN CN201710221097.2A patent/CN107025443A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106097346A (en) * | 2016-06-13 | 2016-11-09 | 中国科学技术大学 | A kind of video fire hazard detection method of self study |
Non-Patent Citations (6)
Title |
---|
SHAOQING REN等: "《Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks》", 《HTTPS://ARXIV.ORG/PDF/1506.01497.PDF》 * |
万维: "《基于深度学习的目标检测算法研究及应用》", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
冯子勇: "《基于深度学习的图像特征学习和分类方法的研究及应用》", 《中国博士学位论文全文数据库 信息科技辑》 * |
吴养会等: "《农村金融计量研究方法与应用》", 30 November 2015 * |
施彦等: "《神经网络设计方法与实例分析》", 31 December 2009, 北京邮电大学出版社 * |
汪子杰: "《基于深度神经网络的视频烟雾检测研究》", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107517251B (en) * | 2017-08-16 | 2020-12-15 | 北京星选科技有限公司 | Information pushing method and device |
CN107517251A (en) * | 2017-08-16 | 2017-12-26 | 北京小度信息科技有限公司 | Information-pushing method and device |
CN107730488A (en) * | 2017-09-21 | 2018-02-23 | 滨州学院 | A kind of method planted using unmanned plane low-altitude remote sensing image automatic detection opium poppy |
CN108256496A (en) * | 2018-02-01 | 2018-07-06 | 江南大学 | A kind of stockyard smog detection method based on video |
CN108334902A (en) * | 2018-02-02 | 2018-07-27 | 北京华纵科技有限公司 | A kind of track train equipment room smog fireproof monitoring method based on deep learning |
CN108764264A (en) * | 2018-03-16 | 2018-11-06 | 深圳中兴网信科技有限公司 | Smog detection method, smoke detection system and computer installation |
CN108537215A (en) * | 2018-03-23 | 2018-09-14 | 清华大学 | A kind of flame detecting method based on image object detection |
CN108428324A (en) * | 2018-04-28 | 2018-08-21 | 温州大学激光与光电智能制造研究院 | The detection device of smog in a kind of fire scenario based on convolutional network |
CN108985192A (en) * | 2018-06-29 | 2018-12-11 | 东南大学 | A kind of video smoke recognition methods based on multitask depth convolutional neural networks |
CN109147254A (en) * | 2018-07-18 | 2019-01-04 | 武汉大学 | A kind of video outdoor fire disaster smog real-time detection method based on convolutional neural networks |
CN109034049A (en) * | 2018-07-23 | 2018-12-18 | 北京密境和风科技有限公司 | The recognition methods of dancing video and device |
CN109271906A (en) * | 2018-09-03 | 2019-01-25 | 五邑大学 | A kind of smog detection method and its device based on depth convolutional neural networks |
CN110096942A (en) * | 2018-12-20 | 2019-08-06 | 北京以萨技术股份有限公司 | A kind of Smoke Detection algorithm based on video analysis |
CN109766780A (en) * | 2018-12-20 | 2019-05-17 | 武汉理工大学 | A kind of ship smog emission on-line checking and method for tracing based on deep learning |
CN109766872A (en) * | 2019-01-31 | 2019-05-17 | 广州视源电子科技股份有限公司 | Image-recognizing method and device |
CN109961042A (en) * | 2019-03-22 | 2019-07-02 | 中国人民解放军国防科技大学 | Smoke detection method combining deep convolutional neural network and visual change diagram |
CN109961042B (en) * | 2019-03-22 | 2021-07-27 | 中国人民解放军国防科技大学 | Smoke detection method combining deep convolutional neural network and visual change diagram |
CN111783968A (en) * | 2020-06-30 | 2020-10-16 | 山东信通电子股份有限公司 | Power transmission line monitoring method and system based on cloud edge cooperation |
CN111986199A (en) * | 2020-09-11 | 2020-11-24 | 征图新视(江苏)科技股份有限公司 | Unsupervised deep learning-based wood floor surface flaw detection method |
CN111986199B (en) * | 2020-09-11 | 2024-04-16 | 征图新视(江苏)科技股份有限公司 | Method for detecting surface flaws of wood floor based on unsupervised deep learning |
CN112396042A (en) * | 2021-01-20 | 2021-02-23 | 鹏城实验室 | Real-time updated target detection method and system, and computer-readable storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107025443A (en) | Stockyard smoke monitoring and on-time model update method based on depth convolutional neural networks | |
CN106815604B (en) | Method for viewing points detecting based on fusion of multi-layer information | |
CN105608456B (en) | A kind of multi-direction Method for text detection based on full convolutional network | |
CN105608479B (en) | In conjunction with the anomaly detection method and system of depth data | |
CN108537215A (en) | A kind of flame detecting method based on image object detection | |
CN110322659A (en) | A kind of smog detection method | |
CN110060238A (en) | Pcb board based on deep learning marks print quality inspection method | |
CN109559302A (en) | Pipe video defect inspection method based on convolutional neural networks | |
CN110378232B (en) | Improved test room examinee position rapid detection method of SSD dual-network | |
CN112801230B (en) | Intelligent acceptance method for unmanned aerial vehicle of power distribution line | |
CN107818321A (en) | A kind of watermark date recognition method for vehicle annual test | |
CN108648169A (en) | The method and device of high voltage power transmission tower defects of insulator automatic identification | |
CN110598693A (en) | Ship plate identification method based on fast-RCNN | |
CN108564077A (en) | It is a kind of based on deep learning to detection and recognition methods digital in video or picture | |
CN111126293A (en) | Flame and smoke abnormal condition detection method and system | |
CN108256496A (en) | A kind of stockyard smog detection method based on video | |
CN106934386A (en) | A kind of natural scene character detecting method and system based on from heuristic strategies | |
CN107220638A (en) | A kind of car plate detection recognition methods based on deep learning convolutional neural networks | |
CN112766218B (en) | Cross-domain pedestrian re-recognition method and device based on asymmetric combined teaching network | |
CN105930798A (en) | Tongue image quick detection and segmentation method based on learning and oriented to handset application | |
CN107145879A (en) | A kind of floristics automatic identifying method and system | |
CN112395998A (en) | Verification method for airport space position in global open database | |
CN108805210A (en) | A kind of shell hole recognition methods based on deep learning | |
CN108537790A (en) | Heterologous image change detection method based on coupling translation network | |
CN112183438B (en) | Image identification method for illegal behaviors based on small sample learning neural network |
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: 20170808 |
|
RJ01 | Rejection of invention patent application after publication |