CN113269138A - FeO content detection method based on deep multi-source information fusion - Google Patents
FeO content detection method based on deep multi-source information fusion Download PDFInfo
- Publication number
- CN113269138A CN113269138A CN202110676983.0A CN202110676983A CN113269138A CN 113269138 A CN113269138 A CN 113269138A CN 202110676983 A CN202110676983 A CN 202110676983A CN 113269138 A CN113269138 A CN 113269138A
- Authority
- CN
- China
- Prior art keywords
- image
- feo
- content
- deep
- source information
- 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
Images
Classifications
-
- 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
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Biology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for detecting FeO content based on deep multi-source information fusion, which relates to the field of steel production and mainly comprises the following steps: extracting a dropped frame image sequence, image preprocessing, deep neural network and artificial image feature extraction, multi-source information feature fusion, and model training and testing. Compared with the prior art, the method combines image information and working condition parameter information, utilizes multi-source information and carries out quantitative prediction on the FeO content through neural network training.
Description
Technical Field
The invention relates to the field of steel production, in particular to a method for detecting FeO content based on deep multi-source information fusion.
Background
In the iron and steel industry, sintered ore is one of important furnace-entering components in high-aluminum smelting, the sintered ore is more economical and practical compared with pellet ore and lump ore, and the proportion of the sintered ore in the furnace-entering components is gradually increased due to the recent rise of the import price of iron and steel raw materials. The quality of the impact of the physicochemical property of the sintered mineral on the smelting performance of the blast furnace is particularly outstanding, and the sintered mineral which meets the standard has great importance for the blast furnace smelting. The specific examples include the drum strength, alkalinity, low-temperature reduction pulverization property, FeO content and the like of the sintered ore. Among these indexes, the FeO content of the sintered ore is one of the most critical indexes. Therefore, the detection of the FeO content is directly related to the quality of blast furnace ironmaking.
At present, a chemical analysis method is more accurate in the determination of the content of FeO, but a time lag exists from sampling to testing, and the change of the content of FeO cannot be reflected in real time. The research on the sintering machine tail image mostly focuses on judging the visual image characteristic of a sintering red layer, and a sintering mineral mass weighing method for comprehensively analyzing the sintering machine tail video by combining the characteristics of sintering site working condition parameters and the like under the severe smoke condition of the sintering machine tail is not provided.
The search of the existing literature finds that the most similar implementation scheme is the Chinese patent application number: 201410307470.2, the name is: a method for controlling FeO content of a sintering ore machine tail section comprises the following specific steps: and simultaneously taking the visible light image and the infrared image as input, extracting the characteristics, and sending the characteristics into a fuzzy clustering system and a neural network system to obtain the FeO content grade. However, the method has simple image characteristics, does not consider working condition parameter information, has a simple network structure, can only carry out qualitative prediction on the FeO content, and cannot carry out quantitative prediction. The patent application numbers are: 201910642094.5, the name is: a method and a system for detecting the FeO content of sinter are specifically prepared as follows: and extracting a key frame image by utilizing the infrared image and combining with a dust change rule at the tail part of the sintering machine, and then carrying out classification judgment on the FeO content through a multiphase thermodynamic model and the like. However, the method only considers the key frame image, but only extracts the artificial feature information of the image, and does not use the depth feature information to fully extract the image features, so that the method has limitation in representing the image information.
In summary, the prior art has the following disadvantages:
(1) the network structure is simple, and the quantitative prediction of the FeO content of the sinter is difficult to carry out.
(2) Only the artificial feature information is considered, the depth information features are not extracted, and the image information is fully represented.
(3) And the artificial characteristics and the on-site working condition parameter characteristics are not fused, and the FeO content of the sintering ore is detected by using multi-source information.
Therefore, those skilled in the art are devoted to developing a method for detecting the content of FeO based on deep multi-source information fusion.
Disclosure of Invention
In view of the above defects in the prior art, the technical problem to be solved by the present invention is how to more comprehensively extract information features to characterize the quality of the sintered ore, and construct a model to accurately and quantitatively predict the FeO content.
In order to achieve the aim, the invention provides a method for detecting the content of FeO based on deep multi-source information fusion, which is characterized by comprising the following steps:
step 1, extracting a falling frame image sequence;
step 2, preprocessing an image;
step 3, extracting depth image features and artificial image features;
step 4, multi-source information characteristic fusion;
and 5, training and testing the model.
Further, the step 1 is to determine the drop frame position through the shallow shape feature and the deep statistical feature of the image, and construct the drop frame image sequence.
Further, the step 2 comprises the following steps:
2.1, filtering the image by using dual-channel Gaussian filtering so as to complete the denoising treatment of the image;
and 2.2, carrying out image defogging processing on the denoised image based on a dark channel.
Further, the denoising process of step 2 uses a two-channel gaussian filter.
Further, the defogging process of the step 2 uses a defogging method based on a dark channel.
Further, the step 3 comprises:
3.1, for the artificial image features, extracting image features of the de-noised and defogged image;
and 3.2, inputting the denoised image into a convolutional neural network for the depth image characteristics, and extracting the depth image characteristics.
Further, the information extracted in step 3.1 mainly includes thickness information and brightness information of the red layer, and thickness information and brightness information of the air holes.
Further, the step 4 comprises:
step 4.1, inputting image data of a model to be input into the convolutional neural network structure channel and carrying out convolution to obtain the output of a first convolutional layer, namely the depth image characteristic;
and 4.2, inputting all feature information at the last full-connection layer for fusion.
Further, the feature information fused in the step 4.2 mainly includes the artificial image feature, the depth image feature and the working condition feature information of the sintering site.
Further, in the step 5, the output of the last hidden layer of the convolutional neural network is input into a regression layer, the function output target is the assay value of the FeO content, and the training and testing of the prediction model of the FeO content are performed through the regression layer.
Compared with the prior art, the beneficial technical effects of the invention comprise:
1) the method uses the dual-channel Gaussian to perform image denoising, can not only denoise the image, but also extract the red layer image and the pore image by selecting different threshold values, thereby being beneficial to the artificial feature extraction of the subsequent image.
2) The artificial features, the depth features and the field working condition features of the image are comprehensively considered, model construction based on multi-source information fusion is carried out, and the model is more accurate and comprehensive.
3) The image characteristics extracted manually are added into the full connection layer, so that the convergence speed of the network can be improved,
4) through a deep learning model, the FeO content in the video can be quantitatively predicted, and the evaluation is more accurate.
The conception, the specific structure and the technical effects of the present invention will be further described with reference to the accompanying drawings to fully understand the objects, the features and the effects of the present invention.
Drawings
FIG. 1 is a flow chart of a preferred embodiment of the present invention.
Detailed Description
The technical contents of the preferred embodiments of the present invention will be more clearly and easily understood by referring to the drawings attached to the specification. The present invention may be embodied in many different forms of embodiments and the scope of the invention is not limited to the embodiments set forth herein.
Fig. 1 is a schematic flow chart of a preferred embodiment of the present invention, which includes the following steps.
A FeO content detection method based on deep multi-source information fusion is characterized by comprising the following steps:
step 1, extracting a falling frame image sequence;
step 2, preprocessing an image;
step 3, extracting deep neural network and artificial image features;
step 4, multi-source information characteristic fusion;
step 5, training and testing the model;
further, the step 1 comprises:
and determining the position of the falling frame according to the shallow shape characteristic and the deep statistical characteristic of the image, and constructing a falling frame image sequence.
Further, the step 2 comprises:
and filtering the image by using dual-channel Gaussian filtering to finish the denoising processing of the image, obtaining the denoised image, and then performing image defogging processing based on a dark channel on the denoised image.
Further, the step 3 comprises:
3.1, for artificial features, carrying out image feature extraction on the de-noised and defogged image, firstly, extracting a red layer image and an air hole image by selecting different thresholds by utilizing double-channel Gaussian filtering again, and then carrying out artificial feature extraction, wherein the artificial feature extraction comprises thickness information and brightness information of a red layer, and thickness information and brightness information of an air hole;
3.2, inputting the denoised image into a convolutional neural network for the depth characteristics, and extracting the depth characteristics of the image through network training;
further, the step 4 comprises:
inputting the image data of the model to be input into a Convolutional Neural Network (CNN) for convolution, and vectorizing the image data to obtain depth characteristic information of the image; and inputting vectorized artificial image characteristics and working condition characteristic information vectors of a sintering site at the last full-connection layer for fusion.
Step 5, training and testing the model;
inputting the output of the last hidden layer into a regression layer, wherein the function output target is the test value of the FeO content of the sintering ore, and training and testing a prediction model of the FeO content through the regression layer;
in summary, the FeO content prediction algorithm based on the sintering machine tail section video provided by the invention performs multisource information characteristic fusion by extracting the falling frame image sequence, performing a series of image preprocessing on the falling frame image sequence, extracting the artificial image characteristic and the depth image characteristic, considering the sintering condition parameters, constructing a deep learning model, and performing quantitative detection on the FeO content of the sintering ore through model training.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.
Claims (10)
1. A FeO content detection method based on deep multi-source information fusion is characterized by comprising the following steps:
step 1, extracting a falling frame image sequence;
step 2, preprocessing an image;
step 3, extracting depth image features and artificial image features;
step 4, multi-source information characteristic fusion;
and 5, training and testing the model.
2. The method for detecting the FeO content based on the depth multi-source information fusion of claim 1, wherein the step 1 is to determine the position of the dropped frame through a shallow shape feature and a deep statistical feature of the image, and construct the image sequence of the dropped frame.
3. The method for detecting the content of FeO based on deep multi-source information fusion according to claim 2, wherein the step 2 comprises the following steps:
2.1, filtering the image by using dual-channel Gaussian filtering so as to complete the denoising treatment of the image;
and 2.2, carrying out image defogging processing on the denoised image based on a dark channel.
4. The method for detecting the content of FeO based on deep multi-source information fusion of claim 3, wherein the denoising process in the step 2 uses two-channel Gaussian filtering.
5. The method for detecting the content of FeO based on deep multisource information fusion of claim 4, wherein the defogging process in the step 2 is a defogging method based on a dark channel.
6. The method for detecting the content of FeO based on deep multi-source information fusion according to claim 5, wherein the step 3 comprises:
3.1, for the artificial image features, extracting image features of the de-noised and defogged image;
and 3.2, inputting the denoised image into a convolutional neural network for the depth image characteristics, and extracting the depth image characteristics.
7. The method for detecting the content of FeO based on deep multi-source information fusion of claim 6, wherein the information extracted in step 3.1 mainly comprises thickness information and brightness information of a red layer, and thickness information and brightness information of pores.
8. The method for detecting the content of FeO based on deep multi-source information fusion according to claim 7, wherein the step 4 comprises:
step 4.1, inputting image data of a model to be input into the convolutional neural network structure channel and carrying out convolution to obtain the output of a first convolutional layer, namely the depth image characteristic;
and 4.2, inputting all feature information at the last full-connection layer for fusion.
9. The method for detecting the content of FeO based on the deep multi-source information fusion of claim 8, wherein the feature information fused in the step 4.2 mainly includes the artificial image feature, the deep image feature and the working condition feature information of a sintering site.
10. The method for detecting the content of FeO based on deep multisource information fusion of claim 9, wherein in the step 5, the output of the last hidden layer of the convolutional neural network is input into a regression layer, the function output target is the assay value of the content of FeO, and training and testing of a prediction model of the content of FeO are performed through the regression layer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110676983.0A CN113269138A (en) | 2021-06-18 | 2021-06-18 | FeO content detection method based on deep multi-source information fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110676983.0A CN113269138A (en) | 2021-06-18 | 2021-06-18 | FeO content detection method based on deep multi-source information fusion |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113269138A true CN113269138A (en) | 2021-08-17 |
Family
ID=77235296
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110676983.0A Pending CN113269138A (en) | 2021-06-18 | 2021-06-18 | FeO content detection method based on deep multi-source information fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113269138A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101339177A (en) * | 2008-08-27 | 2009-01-07 | 攀枝花新钢钒股份有限公司 | Sintering mine FeO prediction system based on nerval net |
| CN102156405A (en) * | 2011-01-24 | 2011-08-17 | 王爱民 | Sintered ore chemical component prediction and intelligent control system under small sample poor information |
| CN105276988A (en) * | 2014-06-30 | 2016-01-27 | 宝钢不锈钢有限公司 | Method for controlling FeO content in sintered ore endmost section |
| WO2019124768A1 (en) * | 2017-12-18 | 2019-06-27 | 주식회사 포스코 | Sintering operation control apparatus and method therefor |
| CN111128313A (en) * | 2019-07-16 | 2020-05-08 | 中南大学 | Method and system for detecting FeO content of sinter |
| US20200202176A1 (en) * | 2017-05-12 | 2020-06-25 | Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirketi | Method for correlation filter based visual tracking |
| CN112329558A (en) * | 2020-10-22 | 2021-02-05 | 上海交通大学 | FeO content prediction method based on sintering machine tail section video |
-
2021
- 2021-06-18 CN CN202110676983.0A patent/CN113269138A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101339177A (en) * | 2008-08-27 | 2009-01-07 | 攀枝花新钢钒股份有限公司 | Sintering mine FeO prediction system based on nerval net |
| CN102156405A (en) * | 2011-01-24 | 2011-08-17 | 王爱民 | Sintered ore chemical component prediction and intelligent control system under small sample poor information |
| CN105276988A (en) * | 2014-06-30 | 2016-01-27 | 宝钢不锈钢有限公司 | Method for controlling FeO content in sintered ore endmost section |
| US20200202176A1 (en) * | 2017-05-12 | 2020-06-25 | Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirketi | Method for correlation filter based visual tracking |
| WO2019124768A1 (en) * | 2017-12-18 | 2019-06-27 | 주식회사 포스코 | Sintering operation control apparatus and method therefor |
| CN111128313A (en) * | 2019-07-16 | 2020-05-08 | 中南大学 | Method and system for detecting FeO content of sinter |
| CN112329558A (en) * | 2020-10-22 | 2021-02-05 | 上海交通大学 | FeO content prediction method based on sintering machine tail section video |
Non-Patent Citations (6)
| Title |
|---|
| RUBEN USAMENTIAGA ET AL.: ""Monitoring Sintering Burn-Through Point Using Infrared Thermography"", 《SENSORS》 * |
| 吴晶: ""基于烧结机尾断面图像的烧结质量人工智能检测方法的研究"", 《中国优秀博硕士学位论文全文数据库(硕士) 工程科技I辑》 * |
| 张毅 等: "《移动机器人技术基础与制作》", 31 January 2013, 哈尔滨工业大学出版社 * |
| 张海波: "《智慧图书馆技术及应用》", 31 May 2020, 河北科学技术出版社 * |
| 王翊: ""烧结矿FeO含量实时预测***的关键技术研究"", 《中国优秀博硕士学位论文全文数据库(硕士) 工程科技I辑》 * |
| 邓超 等: "《数字图像处理与模式识别研究》", 30 June 2018, 地质出版社 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103984951B (en) | A kind of Magnetic testing defect inspection method and system | |
| CN110969130B (en) | Driver dangerous action identification method and system based on YOLOV3 | |
| CN109409315B (en) | Method and system for detecting remnants in panel area of ATM (automatic Teller machine) | |
| CN114549522A (en) | Textile quality detection method based on target detection | |
| CN105678332A (en) | Converter steel-making endpoint determination method and system based on flame image CNN recognizing and modeling process | |
| CN112149512A (en) | Helmet wearing identification method based on two-stage deep learning | |
| CN112036755A (en) | Supervision method and system for building engineering quality detection | |
| CN112329558B (en) | FeO content prediction method based on sintering machine tail section video | |
| CN104077594A (en) | Image recognition method and device | |
| CN106934801A (en) | A kind of fluorescentmagnetic particle(powder) defect inspection method based on Laws texture filterings | |
| CN108846412A (en) | A kind of method of extensive zero sample learning | |
| CN110852364A (en) | Method and device for identifying water source of water burst in mine and electronic equipment | |
| CN116883394B (en) | Diamond quality detection method based on image data processing | |
| CN113935666A (en) | Building decoration wall tile abnormity evaluation method based on image processing | |
| Hu et al. | Accurate identification strategy of coal and gangue using infrared imaging technology combined with convolutional neural network | |
| CN110097140A (en) | A kind of data processing system for tunnel crusing robot | |
| Li et al. | Deep learning features in facial identification and the likelihood ratio bound | |
| CN113269138A (en) | FeO content detection method based on deep multi-source information fusion | |
| JP7301783B2 (en) | Composition analysis method for electronic/electrical equipment parts scrap, electronic/electrical equipment parts scrap processing method, electronic/electrical equipment parts scrap composition analysis device, and electronic/electrical equipment parts scrap processing equipment | |
| CN111563553A (en) | Method for predicting alkalinity of pellet ore phase based on PCA-CNN classification recognition | |
| CN112967335A (en) | Bubble size monitoring method and device | |
| CN116206155A (en) | Waste steel classification and identification method based on YOLOv5 network | |
| CN114882375A (en) | Intelligent identification method and device for tailing pond | |
| CN105354597B (en) | A kind of classification method and device of game articles | |
| CN111191575B (en) | Naked flame detection method and system based on flame jumping modeling |
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 |