CN108182674A - Granularity Detection analysis method based on U-Net deep learning networks - Google Patents
Granularity Detection analysis method based on U-Net deep learning networks Download PDFInfo
- Publication number
- CN108182674A CN108182674A CN201711338870.XA CN201711338870A CN108182674A CN 108182674 A CN108182674 A CN 108182674A CN 201711338870 A CN201711338870 A CN 201711338870A CN 108182674 A CN108182674 A CN 108182674A
- Authority
- CN
- China
- Prior art keywords
- granularity
- deep learning
- image
- grain block
- standardized images
- 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
Classifications
-
- 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
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of granularity Detection analysis methods based on U Net deep learning networks, include the following steps：A, a large amount of grain block diagram pictures of object to be detected are collected；B, grain block diagram picture is pre-processed, obtains standardized images；C, standardized images are manually demarcated, the edge for drawing out a block obtains artificial uncalibrated image, using artificial uncalibrated image as label image；D, the deep learning network based on U Net is built, using the standardized images in step B as the label image in input picture, step C as getImage is exported to deep learning network, starts to train, obtains trained granularity model.The granularity model for the grain block for obtaining object to be detected is trained, when carrying out grain size analysis for object to be detected later, the standardized images that shooting is handled well need to be only input in granularity model, can particle size distribution analysis quickly be completed to grain block or powder.
Description
Technical field
The invention belongs to field of visual inspection, more particularly to a kind of granularity Detection based on U-Net deep learning networks point
Analysis method.
Background technology
Granularity Detection is the experimental work characterized by specific instrument and method to grain block particle property.Grain
Block being widely used in daily life and industrial and agricultural production is general, such as flour, cement, plastics, papermaking, rubber, ceramics, drug
Deng.In different field, there is different requirements to grain block characteristic, in the index of all reflection grain block characteristics, the size of grain block is
It is shown by granularity, so size distribution is an index of greatest concern in all application fields.Size distribution is to grain block
Using effect have important influence, if size distribution be coal cinder, cement grade classification in an important ring, so objective reality
Ground reflects that the size distribution of grain block or powder is a very important job.The grain block granularity Detection of view-based access control model is a kind of new
Emerging technology, it shoots grain block diagram picture using camera, can quickly analyze to obtain size distribution by image procossing, has non-connect
The characteristics of touch, telemeasurement, intellectual analysis.But need synthesis more using conventional digital image method in the prior art
Kind image processing algorithm, such as image equilibration, Sobel edge detections, Threshold segmentation, range conversion, morphology correction, stream
Journey is complicated, has quantity of parameters to need to adjust, and in real image analytic process, and the edge detection and segmentation for being overlapped grain block are
The problem that need to primarily solve.
Invention content
The purpose of the present invention is to provide a kind of granularity Detection analysis methods based on U-Net deep learning networks, obtain
The granularity model of object to be detected quickly completes particle size distribution analysis to grain block.
To achieve these goals, the present invention takes following technical scheme：A kind of grain based on U-Net deep learning networks
Determination method is spent, is included the following steps：A, a large amount of grain block diagram pictures of object to be detected are collected；B, grain block diagram picture is located in advance
Reason, obtains standardized images；C, standardized images are manually demarcated, the edge for drawing out a block obtains artificial calibration maps
Picture, using artificial uncalibrated image as label image；D, the deep learning network based on U-Net is built, by the standardization in step B
Image, as getImage is exported to deep learning network, starts to train, obtain as the label image in input picture, step C
To trained granularity model.
In above-mentioned technical proposal, a large amount of grain block diagram pictures for treating detectable substance are pre-processed to obtain standardized images, are first adopted
It is manually carried out drawing grain block edge according to standardized images, obtains label image, then by standardized images, label image point
Not as in input, output getImage to the deep learning network based on U-Net, it is trained the grain block for obtaining object to be detected
Granularity model, when carrying out grain size analysis for object to be detected later, need to will only shoot the standardized images handled well and be input to
In granularity model, can particle size distribution analysis quickly be completed to grain block or powder.
Description of the drawings
Fig. 1 is flow chart of the present invention.
Specific embodiment
The present invention is made further instructions with reference to attached drawing 1：
A kind of granularity Detection analysis method based on U-Net deep learning networks, includes the following steps：A, it collects to be detected
A large amount of grain block diagram pictures of object；B, grain block diagram picture is pre-processed, obtains standardized images；C, standardized images are carried out artificial
Calibration, the edge for drawing out a block obtains artificial uncalibrated image, using artificial uncalibrated image as label image；D, it builds based on U-
The deep learning network of Net, using the standardized images in step B as the label image in input picture, step C as output
GetImage starts to train, obtains trained granularity model to deep learning network.Treat a large amount of grain block diagram pictures of detectable substance
It is pre-processed to obtain standardized images, is first carried out drawing grain block edge according to standardized images using manually, obtain label figure
Picture, then using standardized images, label image as input, output getImage to the deep learning net based on U-Net
In network, the granularity model for the grain block for obtaining object to be detected is trained, when carrying out grain size analysis for object to be detected later, is only needed
The standardized images that shooting is handled well are input in granularity model, can size distribution point quickly be completed to grain block or powder
Analysis.
Following steps are further included after the step D：E, it re-shoots to obtain new grain block diagram picture, and pre-process
To standardized images, granularity model standardized images being input in step D carries out grain block edge detection；It F, will be in step E
Granularity edge detection results output carry out image procossing, using morphology operations enhance image border, use watershed algorithm
It is partitioned into a block region；G, bianry image statistics is carried out to the grain block region divided in step E, obtain a block radius, area,
The statistical parameters such as quantity finally draw the size distribution histogram curve of grain block.After obtaining the granularity model in step D, in reality
Only the object standardized images to be detected re-shoot need to be inputted granularity model in the application of border, you can be rapidly performed by a block
Edge detection, and image procossing is carried out to result and is partitioned into a block region, the parameters such as statistics grain block radius, area, quantity obtain
Go out intuitive size distribution histogram curve, convenient for personnel's reference analysis.
The step A includes the following steps：A1, object to be detected is uniformly paved in the horizontal plane；A2, shooting position is determined
It puts, fixed camera, adjusts camera direction and lens focus, make the grain block of the object to be detected blur-free imaging on camera；A3, switching phase
Seat in the plane is put and angle treats detectable substance and carries out a large amount of grain block diagram pictures of shooting collection.It is true before U-Net deep learning networks are carried out
Guarantor largely stablized, clearly object grain block diagram picture to be detected.
Grain block diagram picture is pre-processed as follows in the step B：Picture noise is reduced using mean filter, using certainly
Adaptive histogram equalizationization promotes picture contrast, and standardized images are finally obtained using whitening processing.
Include the following steps in the step D：D1, by the label image in the standardized images in step B and step C
It is corresponding to carry out zonule random division, according to mutual corresponding zonule standardized images and zonule in deep learning network
Label image is trained, and obtains granularity model；D2, after obtaining granularity model, deep learning network is automatically by the mark in step B
Standardization image is input in granularity model, obtains detection image；D3, will label image corresponding with step B Plays images
It is compared with detection image, checks whether similarity reaches 98%, be, then granularity model is final detection granularity model；
It is no, then continue deep learning network training.Quickly learnt for the ease of deep learning network, by standardized images
Carry out zonule random division, such as by the standardized images of 1000 × 1000 pixels and label image random division be several 48
× 48 cell area image, is then trained again, fast and accurately obtains granularity model, then automatically to granularity model into
Row detection if detection image is not achieved 98% with label image similarity, proves the not accurate enough of this granularity model training,
It need to continue to train.
Claims (5)
1. a kind of granularity Detection analysis method based on U-Net deep learning networks, includes the following steps：
A, a large amount of grain block diagram pictures of object to be detected are collected；
B, grain block diagram picture is pre-processed, obtains standardized images；
C, standardized images are manually demarcated, the edge for drawing out a block obtains artificial uncalibrated image, by artificial calibration maps
As being used as label image；
D, the deep learning network based on U-Net is built, using the standardized images in step B as in input picture, step C
Label image, to deep learning network, starts to train, obtains trained granularity model as output getImage.
2. the granularity Detection analysis method according to claim 1 based on U-Net deep learning networks, it is characterised in that：
Following steps are further included after the step D：
E, it re-shoots to obtain new grain block diagram picture, and pre-processed to obtain standardized images, standardized images is input to
Granularity model in step D carries out grain block edge detection；
F, the granularity edge detection results output in step E is subjected to image procossing, enhances image border using morphology operations,
A block region is partitioned into using watershed algorithm；
G, bianry image statistics is carried out to the grain block region divided in step E, obtains the statistics such as block radius, area, quantity
Parameter finally draws the size distribution histogram curve of grain block.
3. the granularity Detection analysis method according to claim 1 based on U-Net deep learning networks, it is characterised in that：
The step A includes the following steps：
A1, object to be detected is uniformly paved in the horizontal plane；
A2, camera site, fixed camera are determined, adjusts camera direction and lens focus, make the grain block of object to be detected on camera
Blur-free imaging；
A3, switching camera position and angle treat detectable substance and carry out a large amount of grain block diagram pictures of shooting collection.
4. the granularity Detection analysis method according to claim 1 based on U-Net deep learning networks, it is characterised in that：
Grain block diagram picture is pre-processed as follows in the step B：Picture noise is reduced using mean filter, uses adaptive Nogata
Figure equalization promotes picture contrast, and standardized images are finally obtained using whitening processing.
5. the granularity Detection analysis method according to claim 1 based on U-Net deep learning networks, it is characterised in that：
Include the following steps in the step D：
D1, by the standardized images in step B it is corresponding with the label image in step C carry out zonule random division, in depth
It is trained in learning network according to mutual corresponding zonule standardized images and zonule label image, obtains granularity mould
Type；
D2, after obtaining granularity model, the standardized images in step B are input in granularity model by deep learning network automatically,
Obtain detection image；
D3, label image corresponding with step B Plays images and detection image are compared, checks whether similarity reaches
It is that then granularity model is final detection granularity model to 98%；It is no, then continue deep learning network training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711338870.XA CN108182674B (en) | 2017-12-14 | 2017-12-14 | Granularity detection analysis method based on U-Net deep learning network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711338870.XA CN108182674B (en) | 2017-12-14 | 2017-12-14 | Granularity detection analysis method based on U-Net deep learning network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108182674A true CN108182674A (en) | 2018-06-19 |
| CN108182674B CN108182674B (en) | 2021-11-23 |
Family
ID=62545982
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711338870.XA Active CN108182674B (en) | 2017-12-14 | 2017-12-14 | Granularity detection analysis method based on U-Net deep learning network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108182674B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109064477A (en) * | 2018-07-06 | 2018-12-21 | 陕西师范大学 | With the method for improved U-Net detection cell nuclear periphery |
| CN110089297A (en) * | 2019-05-18 | 2019-08-06 | 安徽大学 | Severity diagnostic method and device under the environment of wheat scab crop field |
| CN110211099A (en) * | 2019-05-18 | 2019-09-06 | 安徽大学 | The wheat scab detection method and device of the comprehensive wheat head front and back sides state of an illness |
| CN110443808A (en) * | 2019-07-04 | 2019-11-12 | 杭州深睿博联科技有限公司 | Medical image processing method and device, equipment, storage medium for the detection of brain middle line |
| CN112070722A (en) * | 2020-08-14 | 2020-12-11 | 厦门骁科码生物科技有限公司 | Fluorescence in situ hybridization cell nucleus segmentation method and system |
| CN113592829A (en) * | 2021-08-04 | 2021-11-02 | 清华大学 | Deep learning silt particle identification method and device based on segmentation and recombination |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006204160A (en) * | 2005-01-27 | 2006-08-10 | Raito Kogyo Co Ltd | Eelgrass holder |
| US20080036187A1 (en) * | 1992-05-05 | 2008-02-14 | Automotive Technologies International, Inc. | Vehicle-Mounted Monitoring Arrangement and Method Using Light-Regulation |
| CN104792674A (en) * | 2015-04-01 | 2015-07-22 | 东南大学 | Method for measuring particle concentration |
| CN107330900A (en) * | 2017-06-22 | 2017-11-07 | 成都品果科技有限公司 | A kind of automatic portrait dividing method |
-
2017
- 2017-12-14 CN CN201711338870.XA patent/CN108182674B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080036187A1 (en) * | 1992-05-05 | 2008-02-14 | Automotive Technologies International, Inc. | Vehicle-Mounted Monitoring Arrangement and Method Using Light-Regulation |
| JP2006204160A (en) * | 2005-01-27 | 2006-08-10 | Raito Kogyo Co Ltd | Eelgrass holder |
| CN104792674A (en) * | 2015-04-01 | 2015-07-22 | 东南大学 | Method for measuring particle concentration |
| CN107330900A (en) * | 2017-06-22 | 2017-11-07 | 成都品果科技有限公司 | A kind of automatic portrait dividing method |
Non-Patent Citations (2)
| Title |
|---|
| WAFAA ALAKWAA 等: "Lung Cancer Detection and Classification with 3D Convolutional Neural Network (3D-CNN)", 《INTERNATIONAL JOURNAL OF ADVANCED COMPUTER SCIENCE AND APPLICATIONS》 * |
| 陈睿敏: "基于深度学习的红外遥感信息自动提取", 《红外》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109064477A (en) * | 2018-07-06 | 2018-12-21 | 陕西师范大学 | With the method for improved U-Net detection cell nuclear periphery |
| CN110089297A (en) * | 2019-05-18 | 2019-08-06 | 安徽大学 | Severity diagnostic method and device under the environment of wheat scab crop field |
| CN110211099A (en) * | 2019-05-18 | 2019-09-06 | 安徽大学 | The wheat scab detection method and device of the comprehensive wheat head front and back sides state of an illness |
| CN110089297B (en) * | 2019-05-18 | 2021-11-26 | 安徽大学 | Method and device for diagnosing severity of disease condition of wheat scab in field environment |
| CN110443808A (en) * | 2019-07-04 | 2019-11-12 | 杭州深睿博联科技有限公司 | Medical image processing method and device, equipment, storage medium for the detection of brain middle line |
| CN110443808B (en) * | 2019-07-04 | 2022-04-01 | 杭州深睿博联科技有限公司 | Medical image processing method and device for brain midline detection, equipment and storage medium |
| CN112070722A (en) * | 2020-08-14 | 2020-12-11 | 厦门骁科码生物科技有限公司 | Fluorescence in situ hybridization cell nucleus segmentation method and system |
| CN113592829A (en) * | 2021-08-04 | 2021-11-02 | 清华大学 | Deep learning silt particle identification method and device based on segmentation and recombination |
| CN113592829B (en) * | 2021-08-04 | 2024-05-17 | 清华大学 | Deep learning sediment particle identification method and device based on segmentation and recombination |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108182674B (en) | 2021-11-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108182674A (en) | Granularity Detection analysis method based on U-Net deep learning networks | |
| CN106651872B (en) | Pavement crack identification method and system based on Prewitt operator | |
| CN107909081B (en) | Method for quickly acquiring and quickly calibrating image data set in deep learning | |
| Aquino et al. | A new methodology for estimating the grapevine-berry number per cluster using image analysis | |
| CN109063559B (en) | Pedestrian detection method based on improved region regression | |
| CN106548160A (en) | A kind of face smile detection method | |
| CN108985169B (en) | Shop cross-door operation detection method based on deep learning target detection and dynamic background modeling | |
| US20230281265A1 (en) | Method for estimating body size and weight of pig based on deep learning | |
| Tong et al. | Skewness correction and quality evaluation of plug seedling images based on Canny operator and Hough transform | |
| CN108711149B (en) | Mineral rock granularity detection method based on image processing | |
| CN102096808B (en) | Method for automatically monitoring and reporting insect condition of rice planthopper | |
| CN108182706B (en) | Method and system for monitoring incinerated substances | |
| CN105184787A (en) | Identification camera capable of automatically carrying out portrait cutout and method thereof | |
| CN113408584B (en) | RGB-D multi-modal feature fusion 3D target detection method | |
| CN102637262B (en) | A kind of self-adaptation bacterial counting | |
| CN113706566B (en) | Edge detection-based perfuming and spraying performance detection method | |
| Sung et al. | Automatic grader for flatfishes using machine vision | |
| CN117253024A (en) | Industrial salt quality inspection control method and system based on machine vision | |
| Verma et al. | Vision based detection and classification of disease on rice crops using convolutional neural network | |
| CN102680488B (en) | Device and method for identifying massive agricultural product on line on basis of PCA (Principal Component Analysis) | |
| CN112686872B (en) | Wood counting method based on deep learning | |
| CN104616283B (en) | A kind of pest detection method based on low-rank representation algorithm | |
| Hua et al. | Image segmentation algorithm based on improved visual attention model and region growing | |
| CN103177244A (en) | Method for quickly detecting target organisms in underwater microscopic images | |
| CN108960094A (en) | A kind of driver's smoking motion detection algorithm based on histograms of oriented gradients |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP01 | Change in the name or title of a patent holder | ||
| CP01 | Change in the name or title of a patent holder |
Address after: No. 228, Wanshui Road, high tech Zone, Hefei City, Anhui Province, 230088 Patentee after: Hefei Jinxing Intelligent Control Technology Co.,Ltd. Address before: No. 228, Wanshui Road, high tech Zone, Hefei City, Anhui Province, 230088 Patentee before: HEFEI GOLD STAR MECHATRONICS TECHNICAL DEVELOPMENT Co.,Ltd. |