CN110309973A - A kind of converter splash prediction technique and system based on video intelligent algorithm - Google Patents
A kind of converter splash prediction technique and system based on video intelligent algorithm Download PDFInfo
- Publication number
- CN110309973A CN110309973A CN201910583656.3A CN201910583656A CN110309973A CN 110309973 A CN110309973 A CN 110309973A CN 201910583656 A CN201910583656 A CN 201910583656A CN 110309973 A CN110309973 A CN 110309973A
- Authority
- CN
- China
- Prior art keywords
- flame
- splash
- converter
- frame
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000009628 steelmaking Methods 0.000 claims abstract description 37
- 238000005457 optimization Methods 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 20
- 238000013527 convolutional neural network Methods 0.000 claims description 57
- 238000013528 artificial neural network Methods 0.000 claims description 43
- 230000008859 change Effects 0.000 claims description 27
- 230000000306 recurrent effect Effects 0.000 claims description 26
- 230000003068 static effect Effects 0.000 claims description 24
- 239000000284 extract Substances 0.000 claims description 16
- 238000007781 pre-processing Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 2
- 238000004519 manufacturing process Methods 0.000 abstract description 8
- 238000003912 environmental pollution Methods 0.000 abstract description 4
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 12
- 230000011218 segmentation Effects 0.000 description 12
- 230000003542 behavioural effect Effects 0.000 description 10
- 238000003709 image segmentation Methods 0.000 description 6
- 229910052742 iron Inorganic materials 0.000 description 6
- 238000003723 Smelting Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 239000007921 spray Substances 0.000 description 5
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 4
- 239000000428 dust Substances 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 210000005036 nerve Anatomy 0.000 description 4
- 210000004218 nerve net Anatomy 0.000 description 4
- 239000001301 oxygen Substances 0.000 description 4
- 229910052760 oxygen Inorganic materials 0.000 description 4
- 108090000623 proteins and genes Proteins 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000009434 installation Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 239000002994 raw material Substances 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 229910000831 Steel Inorganic materials 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 239000007788 liquid Substances 0.000 description 2
- 239000002184 metal Substances 0.000 description 2
- 229910052751 metal Inorganic materials 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000004321 preservation Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000010959 steel Substances 0.000 description 2
- 239000000956 alloy Substances 0.000 description 1
- 229910045601 alloy Inorganic materials 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000007664 blowing Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 206010016256 fatigue Diseases 0.000 description 1
- 239000007789 gas Substances 0.000 description 1
- 230000002045 lasting effect Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000010926 purge Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- VEMKTZHHVJILDY-UHFFFAOYSA-N resmethrin Chemical compound CC1(C)C(C=C(C)C)C1C(=O)OCC1=COC(CC=2C=CC=CC=2)=C1 VEMKTZHHVJILDY-UHFFFAOYSA-N 0.000 description 1
- 239000002893 slag Substances 0.000 description 1
- 239000000779 smoke Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/49—Segmenting video sequences, i.e. computational techniques such as parsing or cutting the sequence, low-level clustering or determining units such as shots or scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P10/00—Technologies related to metal processing
- Y02P10/25—Process efficiency
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Economics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Multimedia (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Manufacturing & Machinery (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of converter splash prediction technique and forecasting system based on video intelligent algorithm, includes fire door flame consecutive image information in image capture device acquisition converter steelmaking process;The section that splash occurs in fire door flame consecutive image information is marked frame by frame;Converter splash prediction model is constructed, and optimization is trained to converter splash prediction model;Whether splash occurs according to the converter splash prediction model prediction converter steelmaking process after training optimization.The present invention can accurately and consistently predict the generation of splash, so as to be controlled in advance in convertor steelmaking process, avoid the occurrence of splash, while can reduce production cost and environmental pollution, improve converter producing safety.
Description
Technical field
The present invention relates to iron and steel smelting technology fields, pre- more particularly to a kind of converter splash based on video intelligent algorithm
Survey method and forecasting system.
Background technique
Pneumatic steelmaking with molten iron, steel scrap, alloy etc. for primary raw material, not by external energy, by the physics of iron liquid itself
Chemical reaction generates heat and completes steelmaking process in converter between heat and iron liquid component.Converter current steel-making is most to lead in the world
The STEELMAKING PRODUCTION method wanted.Since there are the high-temperature chemical reaction processes of various raw material complexity in converter steelmaking process, in raw material
Splash is easy to happen when conditional fluctuation is larger, be mainly shown as fire door have foamed slag or molten metal it is excessive or spray.Converter splash
It will cause the metal loss of 0.5%-5%, increase cost;Also volume of smoke can be generated, environment is polluted;It also will cause ejecta heap
Product, cleaning is difficult, even causes accident when serious, jeopardizes the person and equipment safety.
Outstanding converter smelting operation worker can be changed according to the light and shade and lines of converter flame and shapes such as " soft or hard "
Whether state prediction converter can occur splash, and take the operation for adding auxiliary material, promotion oxygen rifle, reduction blowing oxygen quantity etc in advance to keep away
Exempt from splash.However not all operator can be skilled grasp this " seeing fire " technology, to capture related with splash
Static state, dynamic flame feature and make Accurate Prediction, and worker's lasting for a long time very bright converter fire of viewing at the scene
Flame is also inevitably tired out or relaxes, these can all cause splash prediction inaccurate in time, sprays to increase in converter steelmaking process
The probability splashed.
In recent years, with the development of computer vision technique, computer some fields of image recognition ability even
It has been more than the mankind.Therefore a kind of intelligent analysis method of view-based access control model is needed at present to simulate " seeing fire " mistake of converter operator
Journey, to realize the prediction to converter splash.
Summary of the invention
In view of the foregoing deficiencies of prior art, the purpose of the present invention is to provide a kind of based on video intelligent algorithm
Converter splash prediction technique and forecasting system cannot accurately predict that converter is sprayed for solving in convertor steelmaking process in the prior art
The problem of splashing.
In order to achieve the above objects and other related objects, the present invention provides a kind of converter splash based on video intelligent algorithm
Prediction technique the described method comprises the following steps:
Acquire fire door flame consecutive image information in converter steelmaking process;
The section that splash occurs in fire door flame consecutive image information is marked frame by frame;
Converter splash prediction model is constructed, and optimization is trained to converter splash prediction model;
Whether splash occurs according to the converter splash prediction model prediction converter steelmaking process after training optimization;
The building of converter splash prediction model includes:
According to the preceding N frame consecutive image information of any point-in-time H in fire door flame consecutive image information as input, rear M
Whether there is splash label in frame consecutive image information as the form of output and establishes sample data;Wherein, N and M is natural number;
Extract the flame still image feature in sample data in preceding N frame consecutive image information;
Extract the flame dynamic features that there are timing variations with the flame still image feature.Optionally, the converter
In splash model construction process, the flame in sample data in preceding N frame consecutive image information is extracted by convolutional neural networks CNN
Still image feature;
It is special that the flame that there are timing variations with flame still image feature dynamic is extracted by Recognition with Recurrent Neural Network RNN
Sign.
Optionally, the flame static nature includes flame brightness, flame texture, flame color and flame contours;
The flame dynamic features include the frequency that the amplitude that changes over time of flame brightness, flame brightness change over time
Change with time trend, flame texture of range that rate, flame brightness change over time, flame texture changes with time frequency
Change with time trend, flame contours of rate, flame color change with time amplitude and flame contours change with time frequency
Rate.
It optionally, further include having before constructing converter splash prediction model, the fire door flame marked frame by frame to completion is continuous
Image information is pre-processed, and the pretreatment includes:
According to the flame brightness in fire door flame consecutive image information in splash section, it is bright that extraction flame is divided using figure
It spends gray value of image and is greater than the pre-set image greater than the region of a pre-set image gray threshold and flame luminance picture gray value
Enclosure rectangle corresponding to the region of gray threshold;
And
The fire door flame image information storing data that the splash period occurs is zoomed in the range of 0 to 1.
Optionally, the convolutional neural networks CNN and the Recognition with Recurrent Neural Network RNN are connected by full Connection Neural Network FC
It connects;
And whether converter splash prediction model prediction converter steelmaking process occurs the result of splash by connecting nerve net entirely
Network FC output.
Optionally, video frame rate when acquiring fire door flame consecutive image information is located in a frame per second section, and the frame
The minimum value in rate section is greater than or equal to a preset frame rate threshold value.
The converter splash forecasting system based on video intelligent algorithm that the present invention also provides a kind of includes:
Acquisition module, for acquiring fire door flame consecutive image information in converter steelmaking process;
Mark module is connect with acquisition module, for in fire door flame consecutive image information occur splash section into
Row marks frame by frame;
Prediction module is connect with mark module, for constructing converter splash prediction model, and to converter splash prediction model
It is trained optimization, whether splash is occurred according to the converter splash prediction model prediction converter steelmaking process after training optimization;
Wherein, prediction module building converter splash prediction model includes:
According to the preceding N frame consecutive image information of any point-in-time H in fire door flame consecutive image information as input, rear M
Whether there is splash label in frame consecutive image information as the form of output and establishes sample data;Wherein, N and M is natural number;
Extract the flame still image feature in sample data in preceding N frame consecutive image information;
Extract the flame dynamic features that there are timing variations with the flame still image feature.
It optionally, further include having preprocessing module, the preprocessing module is connect with mark module and prediction module respectively,
For completing the fire door flame consecutive image marked frame by frame to mark module before mark module constructs converter splash prediction model
Information is pre-processed, and the pretreatment includes:
According to the flame brightness in fire door flame consecutive image information in splash section, it is bright that extraction flame is divided using figure
It spends gray value of image and is greater than the pre-set image greater than the region of a pre-set image gray threshold and flame luminance picture gray value
Enclosure rectangle corresponding to the region of gray threshold;
And
The fire door flame image information storing data that the splash period occurs is zoomed in the range of 0 to 1.
Optionally, it during the prediction module building converter splash prediction model, is extracted by convolutional neural networks CNN
Flame still image feature in sample data in preceding N frame consecutive image information;
It is special that the flame that there are timing variations with flame still image feature dynamic is extracted by Recognition with Recurrent Neural Network RNN
Sign;
Convolutional neural networks CNN in the prediction module is with the Recognition with Recurrent Neural Network RNN by connecting nerve net entirely
Network FC connection;
And whether converter splash prediction model prediction converter steelmaking process occurs the result of splash by connecting nerve net entirely
Network FC output.
It optionally, further include having alarm modules, the alarm modules are connect with prediction module, for according to prediction module
Prediction result issues splash alarm.
As described above, a kind of converter splash prediction technique and forecasting system based on video intelligent algorithm of the invention, tool
Have following the utility model has the advantages that splash prediction technique of the invention is by using convolutional neural networks CNN and Recognition with Recurrent Neural Network RNN phase
Combination constructs model, has carried out supervision to the model by the history flame video data largely Jing Guo standard splash result
Training, and reach the splash precision of prediction requirement for meeting engineer application, it can be mentioned automatically by the prediction model after training optimization
The static state and behavioral characteristics for taking fire door flame carry out the prediction of converter splash according to fire door flame static nature and behavioral characteristics.
The trained model of the above method is deployed in converter producing scene by splash early warning system of the invention, to imminent splash
Early warning promptly and accurately is made, foundation is provided for smelting operation adjustment, splash probability can be effectively reduced, to reduce production
Cost, environmental pollution simultaneously improve production security.
Detailed description of the invention
Fig. 1 is the flow diagram of the converter splash prediction technique based on video intelligent algorithm.
Fig. 2 is the structural schematic diagram of converter splash prediction model.
Fig. 3 is the connection schematic diagram of the converter splash forecasting system based on video intelligent algorithm.
Component label instructions
01 industrial camera
02 camera platform
03 communication apparatus
04 cloud server
The work station of 05 converter producing operating platform
CNN convolutional neural networks
RNN Recognition with Recurrent Neural Network
The full Connection Neural Network of FC
Any point-in-time of H fire door flame consecutive image information
N natural number
M natural number
Specific embodiment
Illustrate embodiments of the present invention below by way of specific specific example, those skilled in the art can be by this specification
Other advantages and efficacy of the present invention can be easily understood for disclosed content.The present invention can also pass through in addition different specific realities
The mode of applying is embodied or practiced, the various details in this specification can also based on different viewpoints and application, without departing from
Various modifications or alterations are carried out under spirit of the invention.It should be noted that in the absence of conflict, following embodiment and implementation
Feature in example can be combined with each other.
It should be noted that illustrating the basic structure that only the invention is illustrated in a schematic way provided in following embodiment
Think, only shown in schema then with related component in the present invention rather than component count, shape and size when according to actual implementation
Draw, when actual implementation kenel, quantity and the ratio of each component can arbitrarily change for one kind, and its assembly layout kenel
It is likely more complexity.
Fig. 1 and Fig. 2 are please referred to, the converter splash prediction technique based on video intelligent algorithm that the present embodiment provides a kind of, side
Method the following steps are included:
S1. fire door flame consecutive image information in image capture device acquisition converter steelmaking process;As an example, image is adopted
Collection equipment for example may include industrial camera, and for fire door flame consecutive image information for example including fire door flame video, the present embodiment is logical
Cross fire door flame video in industrial camera acquisition converter steelmaking process.
S2. the section that splash occurs in fire door flame consecutive image information is marked frame by frame;As an example, this implementation
Example by artificial observation flame by being regarded on the collected fire door flame transmission of video of image capture device to a display equipment
Frequency or computer graphical processing flame video mark the specific period that splash occurs frame by frame;I.e. in fire door flame video
The period that splash occurs is marked frame by frame.
S3. converter splash prediction model is constructed, and optimization is trained to converter splash prediction model;
S4. whether splash is occurred according to the converter splash prediction model prediction converter steelmaking process after training optimization.
The building of converter splash prediction model in step S3 has specifically included:
S31. it is used as input according to the preceding N frame consecutive image information of any point-in-time H in fire door flame consecutive image information,
Whether there is splash label in M frame consecutive image information afterwards as the form of output and establishes sample data;Wherein, N and M is nature
Number, and N and M can be set according to actual needs.If sample data is single sample, with any in fire door flame video
On the basis of time point H, preceding N frame continuous videos correspond to the flame data in the time as input, and M frame regards after predicting time point H
Frequently converter splash whether occurs in the corresponding period.
S32. convolutional neural networks (Convolutional corresponding with preceding N frame consecutive image information content is constructed
Neural Networks, CNN) extract flame still image feature in sample data in preceding N frame continuous videos;I.e. one volume
Product neural network respectively corresponds to a frame continuous videos.And the parameter sharing between each convolutional neural networks CNN, to each other
Parameter setting is completely the same.Parameter includes the transformation parameter, normalized of the location parameter of fire door key area, image enhancement
Range parameter etc., parameter setting is consistent, and when guaranteeing to extract flame dynamic features in step S33, the feature of each frame image is being united
Has comparativity under one standard.Wherein, the static nature of flame refers to feature contained by single-frame images, includes that flame is bright
Degree, flame texture, flame color and flame contours etc..
S33. according to flame still image feature, circulation nerve net corresponding with convolutional neural networks CNN quantity is constructed
There are timing variations with flame still image feature in network (Recurrent Neural Network, RNN) extraction step S32
Flame dynamic features, i.e. Recognition with Recurrent Neural Network RNN a cycle period correspond to a convolutional neural networks CNN.Flame is static
It is characterized in that flame dynamic features are the features changed over time for single-frame images.For example flame static nature can be furnace
The flame brightness of certain moment point in mouth flame video, and flame dynamic features can be the case where flame brightness changes over time,
It include that the frequency that changes over time of the amplitude that changes over time of flame brightness, flame brightness, flame brightness change over time
Range, flame texture change with time frequency, the flame color of trend, flame texture that change with time change with time
Frequency etc. that gesture, flame contours change with time amplitude and flame contours change with time.
After the foundation of converter splash prediction model, also by a large amount of sample data to by convolutional neural networks CNN and
The converter splash prediction model that Recognition with Recurrent Neural Network RNN is built into exercises supervision training, pre- to converter splash by sample data
The precision for surveying model is iterated optimization, obtains the Optimized model for meeting engineer application.The prediction essence of converter splash prediction model
The steelmaking feed condition of degree and Different field, operating habit etc. are closely related, and the difference of use environment and purpose is to precision of prediction
Target call is also different, as an example, the precision of prediction range of converter splash Early-warning Model is 80%- in the present embodiment
95%.
It further include having to completing to mark frame by frame in step S2 before converter splash prediction model building in step s3
Fire door flame video is pre-processed, and pretreatment includes figure segmentation and normalization;
The mode of figure segmentation is for example including having: bright according to the flame in fire door flame consecutive image information in splash section
Degree is divided using figure and extracts region and flame luminance graph that flame luminance picture gray value is greater than a pre-set image gray threshold
As gray value is greater than enclosure rectangle corresponding to the region of the pre-set image gray threshold;As an example, in the present embodiment
Image grayscale threshold value for example can be set to 150.
The purpose of image segmentation is the segmentation plug for outlet flame region from fire door flame video, because industrial camera shooting
Video field range is larger, in addition to fire door flame region is there are also other backgrounds, needs to be partitioned into fire door region of concern, avoids
Interference of other background images to converter splash precision of forecasting model.Image segmentation is image procossing profession particular study side
All compare to, method it is more, such as image segmentation just have it is based on threshold value, based on region, based on edge, based on gene coding
, based on wavelet transformation, gene neural network many class dividing methods.The same problem may be ok there are many method
Realize segmentation, effect or variant.For fire door region segmentation in embodiment, demand is fairly simple, therefore it is ratio that we, which use,
It is easier to the thresholding method realized, the higher feature of flame brightness ratio is based on, relatively high threshold value is taken to divide.As showing
, gray value of image is come out greater than 150 extracted region in the present embodiment and takes gray value of image right greater than 150 region institute
The enclosure rectangle answered is as segmentation result.Here the thresholding method that uses, there are also threshold value 150 be not it is unique, actually
To cut plug for outlet flame region as target, also there are many kinds of dividing methods or threshold value to be able to achieve.
Normalized mode is for example including having: scaling to the fire door flame image information storing data that the splash period occurs
In the range of to 0 to 1, the convenient training to converter splash prediction model optimizes.Normal image storage at present generallys use 8,
I.e. pixel coverage is 0 to 28Between -1, therefore normalizing is exactly the pixel value by video storing data divided by 255.
In an example of the present embodiment, convolutional neural networks CNN is with Recognition with Recurrent Neural Network RNN by connecting nerve entirely
Network FC connection;And whether converter splash prediction model prediction converter steelmaking process occurs the result of splash by connecting nerve entirely
Network FC output.Flame static nature and the flame dynamic that convolutional neural networks CNN and Recognition with Recurrent Neural Network RNN are extracted respectively are special
Sign, full Connection Neural Network FC carry out flame static nature and flame dynamic features and corresponding default feature weight value
It calculates, finally exports splash prediction result.As an example, using the full Connection Neural Network number of plies in the present embodiment is 2 layers, mainly
For establishing between convolutional neural networks CNN and the Recognition with Recurrent Neural Network RNN static state captured and behavioral characteristics and output result
Association.Here full Connection Neural Network is 2 layers, and non-exclusive alternative, and more numbers of plies are same available as a result, but counting
Calculation amount can be bigger, the present embodiment use 2 layers only demand and computing capability used tradeoff as a result, having in practical application more
Selection.
In an example of the present embodiment, furnace in converter steelmaking process is acquired by image capture device in step sl
Video frame rate when mouth flame video is located in a frame per second section, and the minimum value in frame per second section is greater than or equal to a preset frame rate
Threshold value.As an example, preset frame per second threshold value for example can be 30fbs.There is oxygen to spray in converter steelmaking process from top high speed
Molten iron in blown converter, the variations such as light and shade, the lines of fire door flame are very fast.Such as the texture of flame with the time be in constantly become
Change, and trend, the frequency etc. that change are not getable from single frames picture, it is necessary to be obtained from consecutive image.However
Video acquisition and memory technology are all to include when per second there are in each frame image by the Continuous Vision discretization of physical world
Frame number, i.e., when frame per second is higher, the continuity of video is better, and the multidate information as caused by discretization is lost fewer.And due to splash
Prediction needs while considering the static state and behavioral characteristics of flame, and flame video acquisition needs enough frame per second;Video frame rate is bigger,
The preservation of flame multidate information is more complete, but corresponding video intelligent analytical scale is huger, it is therefore desirable to according to video analysis
Hardware capabilities select frame per second, but cannot be below 30fps, and otherwise flame multidate information missing is excessive, will affect splash prediction model
Precision.
The present embodiment combines mode with Recognition with Recurrent Neural Network RNN by using convolutional neural networks CNN and constructs converter spray
Prediction model is splashed, is changed by the history flame video data largely by the standard splash result actually occurred to the model
Generation training optimization, reaches the splash precision of prediction requirement for meeting engineer application.It simultaneously can be certainly by the model after training optimization
Whether the dynamic static state and behavioral characteristics for extracting fire door flame, can predict converter according to flame static nature and flame dynamic features
Splash can occur.
Embodiment 2, as shown in Figures 2 and 3, the converter splash prediction based on video intelligent algorithm that the present invention also provides a kind of
System includes:
Acquisition module, for acquiring fire door flame consecutive image information in converter steelmaking process, fire door flame consecutive image
Information is for example including fire door flame video;As an example, the acquisition module in the present embodiment, which is arranged near fire door, clearly to be adopted
Collect the safety zone of fire door flame, acquisition module includes industrial camera, dust guard pollution abatement equipment and communication apparatus etc., is had between them
There are many installation form, the installation form in the present embodiment is for example can include: industrial camera 01 is mounted in dust cover, and fixed
On camera head 02, collected fire door flame video is sent to one or more by communication apparatus 03 by industrial camera 01
It shows in equipment.Wherein, industrial camera 01 acquires fire door flame video with the frame per second not less than 30fps in real time, and dust cover passes through
It introduces gas and purges the dust that can be cleared up in time near camera lens;Camera head 02 can position according to installation point relative to fire door
Shooting angle is adjusted, to collect fire door flame data clear and completely.
Mark module is connect with acquisition module, for being marked frame by frame to the section that splash occurs in fire door flame video
Note;In the present embodiment, according to the fire door flame video shown in display equipment, pass through artificial observation flame video or computer graphic
Shape handles flame video, marks the specific period that splash occurs frame by frame;
Prediction module is connect with mark module, for constructing converter splash prediction model, and to converter splash prediction model
It is trained optimization, whether splash is occurred according to the converter splash prediction model prediction converter steelmaking process after training optimization;In advance
Surveying module includes cloud server 04, which pre-processes the fire door flame video that mark module transmits
It is transferred to the converter splash prediction model built afterwards, prediction result is obtained according to converter splash prediction model.
Prediction module building converter splash prediction model has specifically included:
It is used as and inputs according to the preceding N frame continuous videos of any point-in-time H in fire door flame video, in rear M frame continuous videos
Whether there is splash label as the form of output and establishes sample data;Wherein, N and M is natural number, and N and M can be according to reality
Demand is set.If sample data is single sample, on the basis of any point-in-time H in fire door flame video, preceding N frame connects
Continuous video correspond to the flame data in the time as inputting, and predicts whether sent out after time point H in the M frame video corresponding period
Raw converter splash.
Construct convolutional neural networks (Convolutional Neural corresponding with preceding N frame continuous videos quantity
Networks, CNN) extract flame still image feature in sample data in preceding N frame continuous videos;That is a convolutional Neural net
Network respectively corresponds to a frame continuous videos.And the parameter sharing between each convolutional neural networks CNN, parameter setting to each other
It is completely the same.Parameter includes the location parameter of fire door key area, the transformation parameter of image enhancement, normalized range parameter
Deng, parameter setting is consistent, guarantee extract flame dynamic features when, the feature of each frame image has comparable under unified standard
Property.Wherein, the static nature of flame refers to feature contained by single-frame images, includes flame brightness, flame texture, flame color
With flame contours etc..
According to flame still image feature, Recognition with Recurrent Neural Network corresponding with convolutional neural networks CNN quantity is constructed
It is special that (Recurrent Neural Network, RNN) extracts the flame dynamic for having timing variations with flame still image feature
The corresponding convolutional neural networks CNN of sign, i.e. Recognition with Recurrent Neural Network RNN a cycle period.Flame static nature is to be directed to
Single-frame images, flame dynamic features are the features changed over time.For example flame static nature can be fire door flame video
In certain moment point flame brightness, and flame dynamic features can be the case where flame brightness changes over time, and include flame
Range that frequency that amplitude that brightness changes over time, flame brightness change over time, flame brightness change over time, Flame Grain
Reason change with time frequency, the flame color of trend, flame texture that change with time changes with time trend, flame contours
The amplitude that changes with time and flame contours change with time frequency etc..
After the foundation of converter splash prediction model, also by a large amount of sample data to by convolutional neural networks CNN and
The converter splash prediction model that Recognition with Recurrent Neural Network RNN is built into exercises supervision training, pre- to converter splash by sample data
The precision for surveying model is iterated optimization, obtains the Optimized model for meeting engineer application.The prediction essence of converter splash prediction model
The steelmaking feed condition of degree and Different field, operating habit etc. are closely related, and the difference of use environment and purpose is to precision of prediction
Target call is also different, as an example, the precision of prediction range of converter splash Early-warning Model is 80%- in the present embodiment
95%.
It further include having to carry out in advance the fire door flame video that completion marks frame by frame before the building of converter splash prediction model
Processing, pretreatment include figure segmentation and normalization;
The mode of figure segmentation is for example including having: bright according to the flame in fire door flame consecutive image information in splash section
Degree is divided using figure and extracts region and flame luminance graph that flame luminance picture gray value is greater than a pre-set image gray threshold
As gray value is greater than enclosure rectangle corresponding to the region of the pre-set image gray threshold;As an example, in the present embodiment
Image grayscale threshold value for example can be set to 150.
The purpose of image segmentation is the segmentation plug for outlet flame region from fire door flame video, because industrial camera shooting
Video field range is larger, in addition to fire door flame region is there are also other backgrounds, needs to be partitioned into fire door region of concern, avoids
Interference of other background images to converter splash precision of forecasting model.Image segmentation is image procossing profession particular study side
All compare to, method it is more, such as image segmentation just have it is based on threshold value, based on region, based on edge, based on gene coding
, based on wavelet transformation, gene neural network many class dividing methods.The same problem may be ok there are many method
Realize segmentation, effect or variant.For fire door region segmentation in embodiment, demand is fairly simple, therefore it is ratio that we, which use,
It is easier to the thresholding method realized, the higher feature of flame brightness ratio is based on, relatively high threshold value is taken to divide.As showing
, the extracted region in the present embodiment by gray value of image greater than 150 comes out and flame luminance picture gray value of getting fire is greater than 150
Enclosure rectangle corresponding to region is as segmentation result.Here the thresholding method that uses there are also threshold value 150 is not unique
, actually to cut plug for outlet flame region as target, also there are many kinds of dividing methods or threshold value to be able to achieve.
Normalized mode is for example including having: scaling to the fire door flame image information storing data that the splash period occurs
In the range of to 0 to 1, the convenient training to converter splash prediction model optimizes.Normal image storage at present generallys use 8,
I.e. pixel coverage is 0 to 28Between -1, therefore normalizing is exactly the pixel value by video storing data divided by 255.
In an example of the present embodiment, convolutional neural networks CNN is with Recognition with Recurrent Neural Network RNN by connecting nerve entirely
Network FC connection;And whether converter splash prediction model prediction converter steelmaking process occurs the result of splash by connecting nerve entirely
Network FC output.Flame static nature and the flame dynamic that convolutional neural networks CNN and Recognition with Recurrent Neural Network RNN are extracted respectively are special
Sign, full Connection Neural Network FC carry out flame static nature and flame dynamic features and corresponding default feature weight value
It calculates, finally exports splash prediction result.As an example, using the full Connection Neural Network number of plies in the present embodiment is 2 layers, mainly
For establishing between convolutional neural networks CNN and the Recognition with Recurrent Neural Network RNN static state captured and behavioral characteristics and output result
Association.Here full Connection Neural Network is 2 layers, and non-exclusive alternative, and more numbers of plies are same available as a result, but counting
Calculation amount can be bigger, the present embodiment use 2 layers only demand and computing capability used tradeoff as a result, having in practical application more
Selection.
In an example of the present embodiment, acquisition module acquires fire door in converter steelmaking process by image capture device
Video frame rate when flame video is located in a frame per second section, and the minimum value in frame per second section is greater than or equal to a preset frame rate threshold
Value.As an example, preset frame per second threshold value for example can be 30fbs.There is oxygen to be blown in converter steelmaking process from top high speed
Molten iron in furnace, the variations such as light and shade, the lines of fire door flame are very fast.Such as the texture of flame with the time be in constantly variation
, and trend, the frequency etc. that change are not getable from single frames picture, it is necessary to it is obtained from consecutive image.However it regards
Frequency acquisition and storage technology is all by the Continuous Vision discretization of physical world there are in each frame image, when the frame per second for including
Number, i.e., when frame per second is higher, the continuity of video is better, and the multidate information as caused by discretization is lost fewer.And since splash is pre-
The static state and behavioral characteristics for needing while considering flame are surveyed, flame video acquisition needs enough frame per second;Video frame rate is bigger, fire
The preservation of flame multidate information is more complete, but corresponding video intelligent analytical scale is huger, it is therefore desirable to according to the hard of video analysis
Part ability selects frame per second, but cannot be below 30fps, and otherwise flame multidate information missing is excessive, will affect splash prediction model essence
Degree.
It further include having alarm modules in an example of the present embodiment, alarm modules are connect with prediction module, are used for root
It is predicted that the prediction result of module issues splash alarm.In the present embodiment, alarm modules are mounted on converter producing operating platform
In work station 05, and splash alarm is issued according to the pre- geodesic structure of splash that prediction module transmits, i.e., imminent splash is sent out
Splash alarm out.
The present embodiment combines mode with Recognition with Recurrent Neural Network RNN by using convolutional neural networks CNN and constructs converter spray
Prediction model is splashed, is changed by the history flame video data largely by the standard splash result actually occurred to the model
For training optimization, and the model optimized will be trained to be deployed in converter producing scene, converter splash prediction model may be implemented certainly
The dynamic static state and behavioral characteristics for extracting fire door flame, makes promptly and accurately imminent splash according to flame dynamic features
Early warning provides foundation for smelting operation adjustment, splash probability can be effectively reduced, to reduce production cost, environmental pollution
And improve production security.
To sum up, the present invention provides a kind of converter splash prediction technique and forecasting system based on video intelligent algorithm, including
There is fire door flame video in image capture device acquisition converter steelmaking process;To in fire door flame video occur splash section into
Row marks frame by frame;Converter splash prediction model is constructed, and optimization is trained to converter splash prediction model, is optimized according to training
Whether converter splash prediction model prediction converter steelmaking process afterwards occurs splash.Splash prediction technique of the invention by using
Convolutional neural networks CNN combines mode with Recognition with Recurrent Neural Network RNN and constructs model, by largely passing through standard splash result
History flame video data carries out Training to the model, and reaches the splash precision of prediction requirement for meeting engineer application,
The static state and behavioral characteristics of fire door flame can be automatically extracted by the prediction model after training optimization, it is static according to fire door flame
Feature and behavioral characteristics carry out the prediction of converter splash.Splash early warning system of the invention is by the trained mold portion of the above method
Administration makes early warning promptly and accurately to imminent splash at converter producing scene, provides foundation for smelting operation adjustment, can
To effectively reduce splash probability, to reduce production cost, environmental pollution and improve production security.So the present invention has
Effect overcomes various shortcoming in the prior art and has high industrial utilization value.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.It is any ripe
The personage for knowing this technology all without departing from the spirit and scope of the present invention, carries out modifications and changes to above-described embodiment.Cause
This, institute is complete without departing from the spirit and technical ideas disclosed in the present invention by those of ordinary skill in the art such as
At all equivalent modifications or change, should be covered by the claims of the present invention.
Claims (10)
1. a kind of converter splash prediction technique based on video intelligent algorithm, which is characterized in that the described method comprises the following steps:
Acquire fire door flame consecutive image information in converter steelmaking process;
The section that splash occurs in fire door flame consecutive image information is marked frame by frame;
Converter splash prediction model is constructed, and optimization is trained to converter splash prediction model;
Whether splash occurs according to the converter splash prediction model prediction converter steelmaking process after training optimization;
The building of converter splash prediction model includes:
According to the preceding N frame consecutive image information of any point-in-time H in fire door flame consecutive image information as input, rear M frame connects
Whether there is splash label in continuous image information as the form of output and establishes sample data;Wherein, N and M is natural number;
Extract the flame still image feature in sample data in preceding N frame consecutive image information;
Extract the flame dynamic features that there are timing variations with the flame still image feature.
2. the converter splash prediction technique according to claim 1 based on video intelligent algorithm, it is characterised in that: described turn
In furnace splash model construction process,
The flame still image feature in sample data in preceding N frame consecutive image information is extracted by convolutional neural networks CNN;
The flame dynamic features that there are timing variations with the flame still image feature are extracted by Recognition with Recurrent Neural Network RNN.
3. the converter splash prediction technique according to claim 1 based on video intelligent algorithm, it is characterised in that: the fire
Flame static nature includes flame brightness, flame texture, flame color and flame contours;
The flame dynamic features include the amplitude that changes over time of flame brightness, flame brightness change over time frequency,
Change with time trend, flame texture of range that flame brightness changes over time, flame texture changes with time frequency fire
Change with time trend, the flame contours of frequency, flame color that flame texture change with time change with time amplitude and fire
Flame profile changes with time frequency.
4. the converter splash prediction technique according to claim 1 based on video intelligent algorithm, it is characterised in that: further include
Have before constructing converter splash prediction model, the fire door flame consecutive image information marked frame by frame to completion pre-processes, institute
Stating pretreatment includes:
According to the flame brightness in fire door flame consecutive image information in splash section, is divided using figure and extract flame luminance graph
The region for being greater than a pre-set image gray threshold as gray value and flame luminance picture gray value are greater than the pre-set image gray scale
Enclosure rectangle corresponding to the region of value;
And
The fire door flame image information storing data that the splash period occurs is zoomed in the range of 0 to 1.
5. the converter splash prediction technique according to claim 1 based on video intelligent algorithm, it is characterised in that: the volume
Product neural network CNN is connect with the Recognition with Recurrent Neural Network RNN by full Connection Neural Network FC;
And whether converter splash prediction model prediction converter steelmaking process occurs the result of splash and passes through full Connection Neural Network FC
Output.
6. the converter splash prediction technique according to claim 1 based on video intelligent algorithm, it is characterised in that: acquisition furnace
Video frame rate when mouth flame consecutive image information is located in a frame per second section, and the minimum value in the frame per second section is greater than or waits
In a preset frame rate threshold value.
7. a kind of converter splash forecasting system based on video intelligent algorithm, which is characterized in that the system comprises have:
Acquisition module, for acquiring fire door flame consecutive image information in converter steelmaking process;
Mark module is connect with acquisition module, for in fire door flame consecutive image information occur splash section carry out by
Frame flag;
Prediction module is connect with mark module, is carried out for constructing converter splash prediction model, and to converter splash prediction model
Whether training optimization occurs splash according to the converter splash prediction model prediction converter steelmaking process after training optimization;
Wherein, prediction module building converter splash prediction model includes:
According to the preceding N frame consecutive image information of any point-in-time H in fire door flame consecutive image information as input, rear M frame connects
Whether there is splash label in continuous image information as the form of output and establishes sample data;Wherein, N and M is natural number;
Extract the flame still image feature in sample data in preceding N frame consecutive image information;
Extract the flame dynamic features that there are timing variations with the flame still image feature.
8. the converter splash forecasting system according to claim 7 based on video intelligent algorithm, it is characterised in that: further include
There is preprocessing module, the preprocessing module is connect with mark module and prediction module respectively, for turning in mark module building
Before furnace splash prediction model, the fire door flame consecutive image information for completing to mark frame by frame to mark module is pre-processed, described
Pretreatment includes:
According to the flame brightness in fire door flame consecutive image information in splash section, is divided using figure and extract flame luminance graph
The region for being greater than a pre-set image gray threshold as gray value and flame luminance picture gray value are greater than the pre-set image gray scale
Enclosure rectangle corresponding to the region of threshold value;
And
The fire door flame image information storing data that the splash period occurs is zoomed in the range of 0 to 1.
9. the converter splash forecasting system according to claim 7 based on video intelligent algorithm, it is characterised in that: described pre-
It surveys during module building converter splash prediction model, preceding N frame sequential chart in sample data is extracted by convolutional neural networks CNN
As the flame still image feature in information;
The flame dynamic features that there are timing variations with the flame still image feature are extracted by Recognition with Recurrent Neural Network RNN;
Convolutional neural networks CNN and the Recognition with Recurrent Neural Network RNN in the prediction module pass through full Connection Neural Network FC
Connection;
And whether converter splash prediction model prediction converter steelmaking process occurs the result of splash and passes through full Connection Neural Network FC
Output.
10. the converter splash forecasting system according to claim 7 based on video intelligent algorithm, it is characterised in that: also wrap
Alarm modules are included, the alarm modules are connect with prediction module, alert for issuing splash according to the prediction result of prediction module
Report.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910583656.3A CN110309973B (en) | 2019-07-01 | 2019-07-01 | Converter splash prediction method and system based on video intelligent algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910583656.3A CN110309973B (en) | 2019-07-01 | 2019-07-01 | Converter splash prediction method and system based on video intelligent algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110309973A true CN110309973A (en) | 2019-10-08 |
CN110309973B CN110309973B (en) | 2023-06-02 |
Family
ID=68079631
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910583656.3A Active CN110309973B (en) | 2019-07-01 | 2019-07-01 | Converter splash prediction method and system based on video intelligent algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110309973B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111104856A (en) * | 2019-11-18 | 2020-05-05 | 中冶赛迪技术研究中心有限公司 | Converter smelting splash monitoring method, system, storage medium and equipment |
CN111126206A (en) * | 2019-12-12 | 2020-05-08 | 创新奇智(成都)科技有限公司 | Smelting state detection system and method based on deep learning |
CN111340116A (en) * | 2020-02-27 | 2020-06-26 | 中冶赛迪重庆信息技术有限公司 | Converter flame identification method and system, electronic equipment and medium |
CN111598905A (en) * | 2020-05-13 | 2020-08-28 | 云垦智能科技(上海)有限公司 | Method for identifying type of blast furnace flame by using image segmentation technology |
CN113435037A (en) * | 2021-06-25 | 2021-09-24 | 马鞍山钢铁股份有限公司 | Method for judging slag pouring time in early stage of cast residue smelting of 300t converter |
CN116083676A (en) * | 2021-11-08 | 2023-05-09 | 清华大学 | Method, device, equipment and system for monitoring converter steelmaking process |
CN117173141A (en) * | 2023-09-11 | 2023-12-05 | 山东博昂信息科技有限公司 | Smelting observation system based on flame image characteristics |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140327568A1 (en) * | 2011-12-08 | 2014-11-06 | Thales Nederland B.V. | Method for determining the impact point of a projectile fired at a target above sea surface, and radar system implementing such method |
CN105678332A (en) * | 2016-01-08 | 2016-06-15 | 昆明理工大学 | Converter steel-making endpoint determination method and system based on flame image CNN recognizing and modeling process |
CN105956618A (en) * | 2016-04-27 | 2016-09-21 | 云南昆钢集团电子信息工程有限公司 | Converter steelmaking blowing state recognition system and method based on image dynamic and static characteristics |
CN107610124A (en) * | 2017-10-13 | 2018-01-19 | 中冶赛迪技术研究中心有限公司 | A kind of fire door image pre-processing method |
CN108985376A (en) * | 2018-07-17 | 2018-12-11 | 东北大学 | It is a kind of based on convolution-Recognition with Recurrent Neural Network rotary kiln sequence operating mode's switch method |
-
2019
- 2019-07-01 CN CN201910583656.3A patent/CN110309973B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140327568A1 (en) * | 2011-12-08 | 2014-11-06 | Thales Nederland B.V. | Method for determining the impact point of a projectile fired at a target above sea surface, and radar system implementing such method |
CN105678332A (en) * | 2016-01-08 | 2016-06-15 | 昆明理工大学 | Converter steel-making endpoint determination method and system based on flame image CNN recognizing and modeling process |
CN105956618A (en) * | 2016-04-27 | 2016-09-21 | 云南昆钢集团电子信息工程有限公司 | Converter steelmaking blowing state recognition system and method based on image dynamic and static characteristics |
CN107610124A (en) * | 2017-10-13 | 2018-01-19 | 中冶赛迪技术研究中心有限公司 | A kind of fire door image pre-processing method |
CN108985376A (en) * | 2018-07-17 | 2018-12-11 | 东北大学 | It is a kind of based on convolution-Recognition with Recurrent Neural Network rotary kiln sequence operating mode's switch method |
Non-Patent Citations (2)
Title |
---|
AMAL KHOUFECH等: "Influence of liquid formulation and impact conditions on the wetting of hydrophobic surfaces by aqueous polymetric solutions", 《CHEMICAL ENGINEERING RESEARCH AND DESIGN》 * |
罗特 等: "基于卷积神经网络的转炉炼钢终点碳含量预测方法", 《信息技术》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111104856A (en) * | 2019-11-18 | 2020-05-05 | 中冶赛迪技术研究中心有限公司 | Converter smelting splash monitoring method, system, storage medium and equipment |
CN111104856B (en) * | 2019-11-18 | 2023-06-30 | 中冶赛迪技术研究中心有限公司 | Converter smelting splash monitoring method, system, storage medium and equipment |
CN111126206A (en) * | 2019-12-12 | 2020-05-08 | 创新奇智(成都)科技有限公司 | Smelting state detection system and method based on deep learning |
CN111126206B (en) * | 2019-12-12 | 2023-04-07 | 创新奇智(成都)科技有限公司 | Smelting state detection system and method based on deep learning |
CN111340116A (en) * | 2020-02-27 | 2020-06-26 | 中冶赛迪重庆信息技术有限公司 | Converter flame identification method and system, electronic equipment and medium |
CN111598905A (en) * | 2020-05-13 | 2020-08-28 | 云垦智能科技(上海)有限公司 | Method for identifying type of blast furnace flame by using image segmentation technology |
CN113435037A (en) * | 2021-06-25 | 2021-09-24 | 马鞍山钢铁股份有限公司 | Method for judging slag pouring time in early stage of cast residue smelting of 300t converter |
CN113435037B (en) * | 2021-06-25 | 2022-08-26 | 马鞍山钢铁股份有限公司 | Method for judging slag pouring time in early stage of cast residue smelting of 300t converter |
CN116083676A (en) * | 2021-11-08 | 2023-05-09 | 清华大学 | Method, device, equipment and system for monitoring converter steelmaking process |
CN117173141A (en) * | 2023-09-11 | 2023-12-05 | 山东博昂信息科技有限公司 | Smelting observation system based on flame image characteristics |
Also Published As
Publication number | Publication date |
---|---|
CN110309973B (en) | 2023-06-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110309973A (en) | A kind of converter splash prediction technique and system based on video intelligent algorithm | |
Wang et al. | Machine vision for natural gas methane emissions detection using an infrared camera | |
CN111047818A (en) | Forest fire early warning system based on video image | |
CN111242025B (en) | Real-time action monitoring method based on YOLO | |
CN113903081A (en) | Visual identification artificial intelligence alarm method and device for images of hydraulic power plant | |
CN104966304A (en) | Kalman filtering and nonparametric background model-based multi-target detection tracking method | |
CN107729866A (en) | Ship based on timing diagram picture touches mark automatic detection device and method | |
CN111985112A (en) | Blast furnace digital twin system based on Unity3D | |
CN109978215B (en) | Patrol management method and device | |
CN103945197B (en) | Electric power facility external force damage prevention early warning scheme based on Video Motion Detection technology | |
CN108182416A (en) | A kind of Human bodys' response method, system and device under monitoring unmanned scene | |
CN111079722A (en) | Hoisting process personnel safety monitoring method and system | |
CN115965578A (en) | Binocular stereo matching detection method and device based on channel attention mechanism | |
CN116846059A (en) | Edge detection system for power grid inspection and monitoring | |
CN108053422A (en) | Mobile target monitoring method | |
CN113361364B (en) | Target behavior detection method, device, equipment and storage medium | |
CN104023205B (en) | Intelligent safety defense monitoring system | |
CN105095891A (en) | Human face capturing method, device and system | |
CN104079872A (en) | Video image processing and human-computer interaction method based on content | |
CN117612249A (en) | Underground miner dangerous behavior identification method and device based on improved OpenPose algorithm | |
CN112597902A (en) | Small target intelligent identification method based on nuclear power safety | |
CN116665305A (en) | Method and system for detecting worker behaviors based on computer vision and knowledge graph | |
CN115829324A (en) | Personnel safety risk silent monitoring method | |
CN111881863B (en) | Regional group abnormal behavior detection method | |
CN114708544A (en) | Intelligent violation monitoring helmet based on edge calculation and monitoring method thereof |
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 | ||
CB02 | Change of applicant information |
Address after: 401329 No. 5-6, building 2, No. 66, Nongke Avenue, Baishiyi Town, Jiulongpo District, Chongqing Applicant after: MCC CCID information technology (Chongqing) Co.,Ltd. Address before: 401122 No. 11 Huijin Road, North New District of Chongqing Applicant before: CISDI CHONGQING INFORMATION TECHNOLOGY Co.,Ltd. |
|
CB02 | Change of applicant information | ||
GR01 | Patent grant | ||
GR01 | Patent grant |