CN109161932A - A kind of extracting method of aluminium cell acute conjunctivitis video behavioral characteristics - Google Patents
A kind of extracting method of aluminium cell acute conjunctivitis video behavioral characteristics Download PDFInfo
- Publication number
- CN109161932A CN109161932A CN201811231426.2A CN201811231426A CN109161932A CN 109161932 A CN109161932 A CN 109161932A CN 201811231426 A CN201811231426 A CN 201811231426A CN 109161932 A CN109161932 A CN 109161932A
- Authority
- CN
- China
- Prior art keywords
- acute conjunctivitis
- picture
- video
- behavioral characteristics
- sample
- 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
-
- C—CHEMISTRY; METALLURGY
- C25—ELECTROLYTIC OR ELECTROPHORETIC PROCESSES; APPARATUS THEREFOR
- C25C—PROCESSES FOR THE ELECTROLYTIC PRODUCTION, RECOVERY OR REFINING OF METALS; APPARATUS THEREFOR
- C25C3/00—Electrolytic production, recovery or refining of metals by electrolysis of melts
- C25C3/06—Electrolytic production, recovery or refining of metals by electrolysis of melts of aluminium
- C25C3/20—Automatic control or regulation of cells
Landscapes
- Chemical & Material Sciences (AREA)
- Engineering & Computer Science (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Electrochemistry (AREA)
- Materials Engineering (AREA)
- Metallurgy (AREA)
- Organic Chemistry (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of extracting methods of aluminium cell acute conjunctivitis video behavioral characteristics, first by the image data of continuous 5 frame of video data acquiring, then segmentation obtains acute conjunctivitis image, image gray processing is combined into sample, substitute into improved VGG16 model, using the weight of transfer learning initialization VGG16, feature is obtained finally by model.The present invention can acquire the not getable deeper acute conjunctivitis behavioral characteristics of artificial experience institute, and acquisition speed is quickly, and accuracy rate is higher.
Description
Technical field
The present invention relates to industrial control fields, and in particular to a kind of extraction side of aluminium cell acute conjunctivitis video behavioral characteristics
Method.
Background technique
The degree of superheat of aluminum cell electrolyte refers to the difference of electrolyte temperature and liquidus temperature.The degree of superheat directly affects aluminium
The current efficiency of electrolysis, while the burner hearth shape and stability of electrolytic cell are influenced, and then influence the service life of electrolytic cell.Therefore, have
The size of the detection degree of superheat of effect has critically important directive function for the production process of aluminium electroloysis.Existing degree of superheat measurement
Method needs to measure electrolyte temperature and liquidus temperature respectively, and electrolyte temperature generally uses thermocouple or infrared survey
Wen Yi is measured, and the measurement of liquidus temperature needs offline progress, first has to be sampled electrolyte, is placed on measurement furnace
It is heated in body, makes its fusing, then integrally cool down heating furnace, while acquiring the sample temperature curve of crystallization process to obtain just
Brilliant temperature.Entire measurement process is complex, and measurement expense is high, it is most important that time of measuring length can not on-line measurement.Therefore
Acute conjunctivitis information is mostly observed with experienced worker to differentiate the degree of superheat in the actual production process.Due to artificial randomness and master
The property seen causes to be electrolysed state labile, it is difficult to adjust and arrive optimal state, cause electrolytic efficiency not high, result in waste of resources.
Summary of the invention
It is an object of the invention to overcome defect existing in the prior art, a kind of aluminium cell acute conjunctivitis video dynamic is provided
The extracting method of feature.
The invention adopts the following technical scheme:
A kind of extracting method of aluminium cell acute conjunctivitis video behavioral characteristics, which comprises
S1, continuous 5 frame picture is acquired from some regular time node to an acute conjunctivitis video, the picture is carried out
Threshold segmentation and denoising;
S2, to the picture after Threshold segmentation and denoising, Key dithering is carried out using Gaussian Blur method;
It is extracted centered on acute conjunctivitis center in S3, picture after de-jittering, the acute conjunctivitis figure for the square that side length is 400p
Picture;
S4, gray processing processing is carried out to the acute conjunctivitis image;
S5, by gray processing, treated that 5 frame acute conjunctivitis images are combined into a sample;
S6, the behavioral characteristics for extracting the acute conjunctivitis video from sample data using transfer learning and VGG16 model.
It is further, described that Threshold segmentation is carried out to picture specifically:
Acute conjunctivitis segmentation is carried out using threshold method to the channel r in the picture collected.
Further, the denoising are as follows:
It is denoised using the acute conjunctivitis part that " corrosion ", " expansion " method extract segmentation.
Further, the Gaussian Blur method is to distribute weight using being just distributed very much.
Further, gray processing handles formula are as follows:
Gray_1=R*0.687+G*0.199+B*0.114
Wherein, R, G, B respectively indicate the pixel value in tri- channels R, G, B, and Gray_1 is gray value.
Further, the S5 includes:
Difference before and after 5 frame acute conjunctivitis gray level images and image is combined into the sample in 9 channels.
Further, the S6 includes:
The port number of the sample is transformed to 3 channels with convolutional layer;
Utilize the weight of transfer learning initialization VGG16;
The acute conjunctivitis data in the sample are brought into VGG16 model as training set to be trained to obtain the dynamic spy
Sign.
The advantages and beneficial effects of the present invention are:
The present invention provides a kind of extracting method of aluminium cell acute conjunctivitis video behavioral characteristics, first connects video data acquiring
The image data of 5 continuous frames, then segmentation obtains acute conjunctivitis image, and image gray processing is combined into sample, is substituted into improved
VGG16 model obtains feature finally by model using the weight of transfer learning initialization VGG16.The present invention can acquire people
The not getable deeper acute conjunctivitis behavioral characteristics of work experience institute, and acquisition speed is quickly, accuracy rate is higher.
Detailed description of the invention
Fig. 1 is a kind of extracting method flow diagram of aluminium cell acute conjunctivitis video behavioral characteristics provided by the invention.
Specific embodiment
With reference to the accompanying drawings and examples, further description of the specific embodiments of the present invention.Following embodiment is only
For clearly illustrating technical solution of the present invention, and not intended to limit the protection scope of the present invention.
Deep learning and the maximum difference of traditional mode recognition methods are that it is the automatic learning characteristic from big data, and
The non-feature using hand-designed.And the performance of pattern recognition system can be greatly improved in good feature.In past pattern-recognition
Various applications in, the feature of hand-designed is in dominant position.It relies primarily on the priori knowledge of designer, is difficult with big
The advantage of data.Parameter is adjusted by hand due to relying on, the parameter for only allowing to occur a small amount of in the design of feature.Deep learning can be from
The expression of automatic learning characteristic in big data, wherein may include thousands of parameter.Hand-designed is effectively characterized in out
One quite very long process, and deep learning can for new application from training data quickly study obtain it is new effective
Character representation.
Since the recognition methods based on single frames can only get the static nature of acute conjunctivitis image, and video has shake, and
The pixel of acute conjunctivitis image is 1080*1920, and wherein most is all background, directly with the recognition methods based on CNN extended network
Meeting is so that the training time is long and effect will not be fine.The present invention in the degree of superheat short time in view of that will not change, and one
The extracting method of acute conjunctivitis video behavioral characteristics of the kind based on deep learning.
Fig. 1 is a kind of extracting method flow diagram of aluminium cell acute conjunctivitis video behavioral characteristics provided by the invention, institute
The method of stating includes:
S1, continuous 5 frame picture is acquired from some regular time node to an acute conjunctivitis video, the picture is carried out
Threshold segmentation and denoising;
S2, to the picture after Threshold segmentation and denoising, Key dithering is carried out using Gaussian Blur method;
It is extracted centered on acute conjunctivitis center in S3, picture after de-jittering, the acute conjunctivitis figure for the square that side length is 400p
Picture;
S4, gray processing processing is carried out to the acute conjunctivitis image;
S5, by gray processing, treated that 5 frame acute conjunctivitis images are combined into a sample;
S6, the behavioral characteristics for extracting the acute conjunctivitis video from sample data using transfer learning and VGG16 model.
Further details of elaboration is done to the extracting method process of acute conjunctivitis video behavioral characteristics of the invention below.
Step 1: to the continuous 5 frame picture of a video acquisition, pass through the observation to tri- channel pictures of rgb, it is known that r
The brightness highest in channel, therefore threshold method, that is, divisible acute conjunctivitis is used to the channel r, then can set by the channel r pixel value histogram
Threshold value is 130.There is certain noise by the acute conjunctivitis that threshold method is split, it is " rotten using traditional image processing method here
Erosion ", " expansion " method denoise acute conjunctivitis.
Step 2: since the acute conjunctivitis video of acquisition is manually shot, shake is had in shooting process, so needing here
It will be to acute conjunctivitis video stabilization.Since here using continuous 5 frame picture as training data, and continuous 5 frame picture one
As do not have too big shake, therefore here using Gaussian Blur carry out Key dithering.The principle of Gaussian Blur is as follows, using just dividing very much
Cloth distributes weight.On figure, normal distribution is a kind of bell curve, and closer to center, value is bigger, further away from center,
Value is smaller.When calculating average value, we only need " central point " as origin, other points are according to it in normal curve
On position, distribute weight, so that it may obtain a weighted average:
Step 3: it influences, only takes centered on acute conjunctivitis center, side length 400p in order to solve image shape difference bring
Square acute conjunctivitis image can be directly obtained the image lower left corner (xl, yl) of acute conjunctivitis part since image is binary map
With the coordinate (xr, yr) in the upper right corner.Then it can get in acute conjunctivitis according to formula (xc, yc)=((xl+xr)/2, (yl+yr)/2)
Heart coordinate, the lower-left angular coordinate and upper right angular coordinate for finally extracting image are (xl-100, yl-100), (xl+100, yl+100).
Step 4: bring is influenced to convert grayscale image for image in order to reduce flame, traditional formula is converted are as follows:
Gray=R*0.299+G*0.587+B*0.114 (2)
Gray_1=R*0.687+G*0.199+B*0.114 (3)
Because pixel value of the flame portion in the channel G is relatively high, therefore weaken flame by reducing the channel G institute's accounting column
It influences.
Step 5: training sample is generated, the difference before and after the acute conjunctivitis gray level image and image of 5 frames is combined into one 9 and is led to
The sample in road.
Step 6: extracting feature using VGG16, and due to the data that the input of VGG16 is 3 channels, and acute conjunctivitis video is handled
Data out are 9 channels, therefore port number is first transformed to 3 channels with convolutional layer.Again using the shared parameter concluded in migration
Method carry out training pattern, i.e., the parameter of model is first trained on imagNet with VGG16, then again regards acute conjunctivitis data
Training set is finely adjusted, in this way can not only more rapid convergence, and better effect can be reached.
The present invention is first by the image data of continuous 5 frame of video data acquiring, and then segmentation obtains acute conjunctivitis image, will scheme
As gray processing is combined into sample, improved VGG16 model is substituted into, using the weight of transfer learning initialization VGG16, is finally led to
It crosses model and obtains feature.The present invention can acquire the not getable deeper acute conjunctivitis behavioral characteristics of artificial experience institute, and obtain
Taking speed, accuracy rate is higher quickly.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (7)
1. a kind of extracting method of aluminium cell acute conjunctivitis video behavioral characteristics, which is characterized in that the described method includes:
S1, continuous 5 frame picture is acquired from some regular time node to an acute conjunctivitis video, threshold value is carried out to the picture
Segmentation and denoising;
S2, to the picture after Threshold segmentation and denoising, Key dithering is carried out using Gaussian Blur method;
It is extracted centered on acute conjunctivitis center in S3, picture after de-jittering, the acute conjunctivitis image for the square that side length is 400p;
S4, gray processing processing is carried out to the acute conjunctivitis image;
S5, by gray processing, treated that 5 frame acute conjunctivitis images are combined into a sample;
S6, the behavioral characteristics for extracting the acute conjunctivitis video from sample data using transfer learning and VGG16 model.
2. the method as described in claim 1, which is characterized in that described to carry out Threshold segmentation to picture specifically:
Acute conjunctivitis segmentation is carried out using threshold method to the channel r in the picture collected.
3. the method as described in claim 1, which is characterized in that the denoising are as follows:
It is denoised using the acute conjunctivitis part that " corrosion ", " expansion " method extract segmentation.
4. the method as described in claim 1, which is characterized in that the Gaussian Blur method is to distribute and weigh using being just distributed very much
Weight.
5. the method as described in claim 1, which is characterized in that gray processing handles formula are as follows:
Gray_1=R*0.687+G*0.199+B*0.114
Wherein, R, G, B respectively indicate the pixel value in tri- channels R, G, B, and Gray_1 is gray value.
6. the method as described in claim 1, which is characterized in that the S5 includes:
Difference before and after 5 frame acute conjunctivitis gray level images and image is combined into the sample in 9 channels.
7. the method as described in one of claim 1-6, which is characterized in that the S6 includes:
The port number of the sample is transformed to 3 channels with convolutional layer;
Utilize the weight of transfer learning initialization VGG16;
The acute conjunctivitis data in the sample are brought into VGG16 model as training set to be trained to obtain the behavioral characteristics.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811231426.2A CN109161932A (en) | 2018-10-22 | 2018-10-22 | A kind of extracting method of aluminium cell acute conjunctivitis video behavioral characteristics |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811231426.2A CN109161932A (en) | 2018-10-22 | 2018-10-22 | A kind of extracting method of aluminium cell acute conjunctivitis video behavioral characteristics |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109161932A true CN109161932A (en) | 2019-01-08 |
Family
ID=64878891
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811231426.2A Pending CN109161932A (en) | 2018-10-22 | 2018-10-22 | A kind of extracting method of aluminium cell acute conjunctivitis video behavioral characteristics |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109161932A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114959797A (en) * | 2022-07-04 | 2022-08-30 | 广东技术师范大学 | Aluminum electrolysis cell condition diagnosis method based on data amplification and SSKELM |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102348046A (en) * | 2010-07-30 | 2012-02-08 | 富士通株式会社 | Video shake removing method and video shake removing device |
CN104748793A (en) * | 2015-03-19 | 2015-07-01 | 中南大学 | Real-time combined measurement device and method for temperature and flow speed of aluminum electrolytic cell melt |
CN107204004A (en) * | 2017-04-26 | 2017-09-26 | 中南大学 | A kind of aluminium cell pinkeye video behavioral characteristics recognition methods and system |
CN107248157A (en) * | 2017-06-30 | 2017-10-13 | 中南大学 | A kind of aluminium cell sees the method and its device of fire automatically |
CN107330876A (en) * | 2017-06-12 | 2017-11-07 | 济南浪潮高新科技投资发展有限公司 | A kind of image automatic diagnosis method based on convolutional neural networks |
CN109871778A (en) * | 2019-01-23 | 2019-06-11 | 长安大学 | Lane based on transfer learning keeps control method |
-
2018
- 2018-10-22 CN CN201811231426.2A patent/CN109161932A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102348046A (en) * | 2010-07-30 | 2012-02-08 | 富士通株式会社 | Video shake removing method and video shake removing device |
CN104748793A (en) * | 2015-03-19 | 2015-07-01 | 中南大学 | Real-time combined measurement device and method for temperature and flow speed of aluminum electrolytic cell melt |
CN107204004A (en) * | 2017-04-26 | 2017-09-26 | 中南大学 | A kind of aluminium cell pinkeye video behavioral characteristics recognition methods and system |
CN107330876A (en) * | 2017-06-12 | 2017-11-07 | 济南浪潮高新科技投资发展有限公司 | A kind of image automatic diagnosis method based on convolutional neural networks |
CN107248157A (en) * | 2017-06-30 | 2017-10-13 | 中南大学 | A kind of aluminium cell sees the method and its device of fire automatically |
CN109871778A (en) * | 2019-01-23 | 2019-06-11 | 长安大学 | Lane based on transfer learning keeps control method |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114959797A (en) * | 2022-07-04 | 2022-08-30 | 广东技术师范大学 | Aluminum electrolysis cell condition diagnosis method based on data amplification and SSKELM |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109101865A (en) | A kind of recognition methods again of the pedestrian based on deep learning | |
CN105740945B (en) | A kind of people counting method based on video analysis | |
CN107564022B (en) | Saliency detection method based on Bayesian Fusion | |
CN109949341B (en) | Pedestrian target tracking method based on human skeleton structural features | |
CN104867144B (en) | IC element welding point defect detection methods based on mixed Gauss model | |
CN109166137A (en) | For shake Moving Object in Video Sequences detection algorithm | |
CN107909085A (en) | A kind of characteristics of image Angular Point Extracting Method based on Harris operators | |
CN107491767B (en) | End-to-end unconstrained face key point detection method | |
CN104537381B (en) | A kind of fuzzy image recognition method based on fuzzy invariant features | |
CN109816645A (en) | A kind of automatic testing method of coil of strip loose winding | |
CN108960142A (en) | Pedestrian based on global characteristics loss function recognition methods again | |
CN106570885A (en) | Background modeling method based on brightness and texture fusion threshold value | |
CN112364881B (en) | Advanced sampling consistency image matching method | |
CN109161932A (en) | A kind of extracting method of aluminium cell acute conjunctivitis video behavioral characteristics | |
CN103955673B (en) | Body recognizing method based on head and shoulder model | |
CN103093241B (en) | Based on the remote sensing image nonuniformity cloud layer method of discrimination of homogeneity process | |
CN109559323A (en) | A method of picture edge characteristic is enhanced based on improved prewitt operator | |
CN109978916A (en) | Vibe moving target detecting method based on gray level image characteristic matching | |
CN108230334A (en) | The method of high concentration dust storm image segmentation based on gray threshold | |
CN112580634A (en) | Air tightness detection light source adjusting method and system based on computer vision | |
CN109002745A (en) | A kind of lane line real-time detection method based on deep learning and tracking technique | |
CN116739986A (en) | Optical acoustic combination method for detecting typical defects of AUV in water conveyance tunnel | |
CN111160362A (en) | FAST feature homogenization extraction and IMU-based inter-frame feature mismatching removal method | |
CN109919964A (en) | The method that Gaussian Background modeling technique based on mathematical morphology carries out image procossing | |
CN112561949B (en) | Rapid moving object detection algorithm based on RPCA and support vector machine |
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: 20190108 |
|
RJ01 | Rejection of invention patent application after publication |