CN109685038A - A kind of article clean level monitoring method and its device - Google Patents
A kind of article clean level monitoring method and its device Download PDFInfo
- Publication number
- CN109685038A CN109685038A CN201910019824.6A CN201910019824A CN109685038A CN 109685038 A CN109685038 A CN 109685038A CN 201910019824 A CN201910019824 A CN 201910019824A CN 109685038 A CN109685038 A CN 109685038A
- Authority
- CN
- China
- Prior art keywords
- image
- cleanliness
- article
- numerical value
- clean
- 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
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000012544 monitoring process Methods 0.000 title claims abstract description 17
- 230000003749 cleanliness Effects 0.000 claims abstract description 47
- 238000012549 training Methods 0.000 claims abstract description 19
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000013528 artificial neural network Methods 0.000 claims abstract description 9
- 238000004364 calculation method Methods 0.000 claims abstract description 9
- 238000011109 contamination Methods 0.000 claims description 6
- 238000004452 microanalysis Methods 0.000 claims description 6
- 238000012806 monitoring device Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 239000000284 extract Substances 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000012706 support-vector machine Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 2
- 238000010801 machine learning Methods 0.000 abstract description 4
- 235000013399 edible fruits Nutrition 0.000 description 5
- 238000001514 detection method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 239000003086 colorant Substances 0.000 description 2
- 235000013305 food Nutrition 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of article clean level monitoring method and its device, several article surfaces are acquired first completely with sordid image information, and the clean image in surface is demarcated;Gray processing processing is carried out respectively to image information is obtained, and its corresponding HOG character numerical value is calculated separately by variance and histogram of gradients;The HOG feature for extracting image, based on HOG feature training SVM classifier;Calculate the cleanliness numerical value n of article to be measured;Image dividing processing is carried out to image, a neural network classifier is gone out based on FASTER-RCNN model training;Calculate the cleanliness numerical value m of article to be measured;Based on cleanliness numerical value n and cleanliness numerical value m, final cleanliness result is obtained according to weight calculation;Cleanliness is carried out according to cleanliness result and threshold value is compared judgement, if clean this method and its device realize the monitoring that clean level is carried out to the surface of article based on computer vision and machine learning algorithm.
Description
Technical field
The invention belongs to image identification technical fields;More particularly to a kind of article clean level monitoring method and its device.
Background technique
With the development of the times, people increasingly pay close attention to the quality of life level of oneself, and the pursuit clean to health also exists
Constantly being promoted, can many people usually suspect oneself, whether the food eaten completely or not? is life appliance completely or not? clothes is washed
It is clean? this is to solve this problem it is necessary to a kind of clean monitoring method, to tool used in everyday, clothes or food
Etc. being monitored, allow people being capable of clean article or things safe to use.
Summary of the invention
The present invention provides a kind of article clean level monitoring method and its devices;This method be based on computer vision and
Machine learning algorithm realizes the monitoring that clean level is carried out to the surface of article.
The technical scheme is that a kind of article clean level monitoring method, includes the following steps;
S1, acquisition several surfaces of article and are demarcated the clean image in surface completely with sordid image information;
S2 carries out gray processing processing to the clean and sordid image information obtained in S1 respectively, and passes through variance
Its corresponding HOG characteristic value is calculated separately with histogram of gradients;
S3 extracts in S2 completely with the HOG feature of sordid image, is then based on HOG feature training SVM classifier;
S4, the SVM classifier based on training in S3 test the image of article surface to be measured, calculate article to be measured
Cleanliness numerical value n;
S5 carries out image dividing processing to the image obtained in S1, goes out a nerve based on FASTER-RCNN model training
Network classifier;
S6 tests the image of article surface to be measured based on the obtained neural network classifier of S5, calculates to be measured
The cleanliness numerical value m of article;
S7, based on cleanliness numerical value n and cleanliness numerical value m that S4 and S6 are obtained, according to cleanliness numerical value n and clean degree
The weight calculation of value m obtains final cleanliness result;
S8 carries out cleanliness judgement, given threshold, when numerical value is greater than 1.0 according to cleanliness result final obtained by S7
It is judged as clean, is otherwise judged as unclean.
The detailed process that image HOG feature is extracted in the S3 is:
S31 converts the image into polar coordinates;
S32 calculates the HOG feature of each pixel in image;
S33 traverses all pixels point, and the HOG feature vector of each pixel is formed multidimensional HOG feature vector group;
S34 calculates the HOG feature of image: the multidimensional HOG feature vector group in S3.3 being normalized by L2 norm, is turned
It is changed to a HOG feature vector.
In S3, using the resulting image HOG characteristic value of S34 as input sample, SVM classifier is obtained based on the training of HOG feature,
In-between node layer is image HOG feature and support vector machines nonlinear operation, export each intermediate node it is corresponding support to
The linear combination of amount.
The process for going out a neural network classifier based on FASTER-RCNN model training in S5 is:
S51 carries out manual segmentation to the sordid image in part, is partitioned into the region of contamination with non-contamination, and uses respectively
Different colours are marked;
S52 selects most suitable parameter training to go out the optimal neural network classifier of effect.
In S4, cleanliness numerical value n is 0 or 1;In S6, cleanliness numerical value m is 0~1.
In S7, by calculation formula: 0.5* cleanliness numerical value n+0.5* cleanliness numerical value m calculates final cleanliness knot
Fruit, cleanliness result are that threshold value is set as 1.0 in 0~1.5, S8.
It is of the invention another solution is that a kind of article clean level monitoring device, including master control borad, master control borad pass through
Onboard USB interface connects digital microimaging head, is additionally provided with wireless transport module on master control borad;Master control borad is integrated and can
Implement above-mentioned method.
Further, the features of the present invention also characterized in that:
Wherein master control borad directly acquires article surface image to be measured by digital micro-analysis camera or passes through wireless transmission mould
Block indirection obtains article surface image to be measured.
Wherein wireless transport module is Wi-Fi, ZigBee, bluetooth or infrared.
Wherein master control borad is also connected with storage unit.
Compared with prior art, the beneficial effects of the present invention are: this method is based on image processing techniques and machine learning skill
Art is handled by measuring targets surface image, then by training SVM classifier, and based on SVM classifier reality
The analysis of detection images of items is now treated, to judge that article surface to be measured is clean or not completely.
Further, HOG feature be describe image local area histogram of gradients, be applied to target object detection with
Signature analysis;Based on HOG feature training SVM classifier, the judgement to article surface to be measured is realized by machine learning.
Beneficial effects of the present invention also reside in: directly acquiring image information by digital micro-analysis camera in the device, number
Word microimaging head can directly acquire the data image signal of image, and export to master control borad, and master control borad passes through above-mentioned side
Method realizes the monitoring to its clean degree.
Further, determinand can be obtained indirectly by other equipment by setting wireless transport module in the device
The image on product surface improves the flexibility that the device uses;Meanwhile the device can be read from storage unit image or its
His data.
Detailed description of the invention
Fig. 1 is the clean image of the sample surfaces of input;
Fig. 2 is the sordid image of sample surfaces of input;
Fig. 3 is that Fig. 2 is changed into the polar diameter after polar coordinate system and polar angle hum pattern;
Fig. 4 is clean image and its corresponding HOG feature;
Fig. 5 is unclean image and its corresponding HOG feature;
Fig. 6 is the device of the invention structural schematic diagram.
Specific embodiment
Technical solution of the present invention is further illustrated in the following with reference to the drawings and specific embodiments.
The present invention provides a kind of article clean level monitoring methods, comprising the following steps:
S1 obtains article surface completely with sordid image information, and demarcates obtained table as depicted in figs. 1 and 2
The clean image in face.
S2 carries out gray processing processing to two kinds of images obtained in S1 respectively, then reuses variance and histogram calculation
Its corresponding HOG characteristic value out.
S3, extracts the HOG feature of two images in S2, and two HOG features of acquisition are compared, and is based on
The training of HOG feature simultaneously obtains SVM classifier.
The detailed process for extracting image HOG feature is:
S31, as shown in figure 3, converting the image into polar coordinates first;Specifically, according to pixel X-direction each in image
With the gradient in Y-direction, realize cartesian coordinate and polar conversion, after being changed into polar coordinates, angular range [0,
360], convenient for calculating the histogram at all directions angle later.
S32 calculates the HOG feature of each pixel;The gradient direction and gradient magnitude of each pixel are calculated first, are counted
Calculate formula are as follows:
It is simple in order to calculate in tangible 0 ° -360 ° of the arc range of the gradient direction that the calculation formula obtains, by gradient to
Range constraint be 0 ° -180 °, and be divided into 9 directions, each direction is 20 °, then the range of gradient direction angle value becomes
For [0,9], the gradient magnitude inside each small Cell is counted according to this 9 directions, after having been calculated, it will generate one
A abscissa X is gradient direction, and ordinate Y is the histograms of oriented gradients of gradient magnitude;Then quick by 9 width integral images
Realize the calculating of HOG feature, the histogram of HOG feature has 9 bins, the corresponding integral image of each bins.
The HOG characteristic value of image is calculated in S33;Firstly, by the HOG feature vector of each pixel obtained in step 3
It is combined into the HOG feature vector group of various dimensions;Then the HOG feature vector group of the various dimensions is subjected to the normalization of L2 norm, made
Its HOG feature vector for being changed into a Block, and can be according to the HOG to get the HOG feature vector for arriving whole image
Feature vector draws the visual representation of HOG characteristic pattern;As shown in figure 4, being clean image and its corresponding HOG characteristic value;Fig. 5 is not
Clean image and its corresponding HOG characteristic value.
Wherein based on the training of HOG feature and obtain the detailed process of SVM classifier and be: the SVM classifier is two classification
Classifier, i.e., simple neural network;Input is sample data, i.e. the image HOG characteristic value that S3 is obtained, middle layer node is figure
As HOG feature and support vector machines nonlinear operation, the linear combination of the corresponding supporting vector of each intermediate node is exported.
The case where can wherein dividing for general linear, corresponding classifying face equation are as follows: g (x)=wT·x+w0, w is hyperplane
Normal direction, resulting classifying face is known as hyperplane, and so-called optimal classification line is exactly that classification line is required to be not only able to two classes
It is faultless separated, and require classification gap sufficiently large.
S4 detects the image of article surface to be measured based on the obtained SVM classifier of S3, and passes through discrimination formula
And its decision principle judges article surface to be measured to be clean or unclean;Discrimination formula are as follows: g (x)=sgn (wT·x+w0);If
G (x) > 0 item returns the result 0, as unclean;1 is returned the result if g (x) < 0, as completely;Nothing is returned if g (x)=0
Return value;It returns the result as cleanliness numerical value n.
S5 carries out manual segmentation to the sordid image in part, is partitioned into contamination and non-contamination for the picture in S1
Region, and be marked respectively with different colours;Artificial image's segmentation is carried out using the open source annotation tool of Facebook company
It marks, is carried out in annotation process according to the rule of greasy dirt (blue), dust (black), clean part (white), distinguish multiple face
The purpose of color is to be more widely applied.
The image that S5 acceptance of the bid is poured in is put into FASTER-RCNN deep learning frame, carries out parameter optimization adjustment by S6
Afterwards, by obtaining training result model up to V100 image processor long-time numerical behavior tall and handsome, there is results model, by new figure
As input model, return is obtained as a result, the result is cleanliness numerical value m.
S7, based on the cleanliness numerical value n and cleanliness numerical value m obtained based on S4 and S6, according to cleanliness numerical value n and cleaning
The weight calculation of degree value m obtains final cleanliness result;Using the method that binode fruit weights be because during the test,
Binode fruit shows accuracy more higher than unijunction fruit, so the accuracy in order to guarantee result, by calculation formula: 0.5* is clean
Cleanliness numerical value n+0.5* cleanliness numerical value m calculates final cleanliness as a result, cleanliness result is 0~1.5;
S8, the calculating means weighted using binode fruit, and according to test result define one it is clean whether judgement
Threshold value, when numerical value is greater than threshold value be it is clean, otherwise to be unclean, cleanliness result is 0~1.5, and the threshold value is 1.0
The present invention also provides a kind of article clean level monitoring devices, as shown in fig. 6, including master control borad, master control borad is adopted
With systemonchip SoC, master control borad connects digital microimaging head by onboard USB interface, is additionally provided on master control borad wireless
Transmission module, and the master control borad can implement above-mentioned article clean level detection method.
Preferred master control borad uses MT7688AN chip, which is current industry systemonchip least in power-consuming, branch
It holds linux system and 802.11Wi-Fi is online.It has the wireless performance of brilliant operational performance, high speed, can directly use
Mobile phone connection can be accomplished to scheme to pass in real time, while support USB2.0Host, can carry an embedded Linux system, can be with
Most of USB camera on the market is connected and driven, meets us and needs to connect a Portable USB microimaging head
Demand.Master control borad can directly acquire the image of article surface to be measured by digital micro-analysis camera, or by wirelessly passing
The image of defeated module indirect gain article surface to be measured.
Preferred digital micro-analysis camera can convert optical signals into electric signal, then be converted to by analog-digital converter
Data image signal, then exported by interface to master control borad after being handled by digital signal processing chip (DSP).Lead in the present invention
The widora core board for crossing MT7688AN carries out GPIO pin control and automatically opens digital micro-analysis camera switch, recycles onboard
USB Host function-driven micro-image sensor, image is uploaded in master control borad.
Preferred wireless transport module is Wi-Fi, ZigBee, bluetooth or infrared.In the present invention by mobile phone etc. other
External equipment gets article surface image to be measured and then sends an image to master control borad by wireless transport module.
It is also connected with storage unit on preferred master control borad, storage unit can store image or other data.
Claims (10)
1. a kind of article clean level monitoring method, which is characterized in that include the following steps;
S1, acquisition several surfaces of article and are demarcated the clean image in surface completely with sordid image information;
S2 carries out gray processing processing to the clean and sordid image information obtained in S1 respectively, and passes through variance and ladder
Degree histogram calculates separately its corresponding HOG characteristic value;
S3 extracts in S2 completely with the HOG feature of sordid image, is then based on HOG feature training SVM classifier;
S4, the SVM classifier based on training in S3 test the image of article surface to be measured, calculate the clean of article to be measured
Cleanliness numerical value n;
S5 carries out image dividing processing to the image obtained in S1, goes out a neural network based on FASTER-RCNN model training
Classifier;
S6 tests the image of article surface to be measured based on the obtained neural network classifier of S5, calculates article to be measured
Cleanliness numerical value m;
S7, based on cleanliness numerical value n and cleanliness numerical value m that S4 and S6 are obtained, according to cleanliness numerical value n and cleanliness numerical value m
Weight calculation obtain final cleanliness result;
S8 carries out cleanliness judgement according to cleanliness result final obtained by S7, and given threshold judges when numerical value is greater than 1.0
To be clean, otherwise it is judged as unclean.
2. article clean level monitoring method according to claim 1, which is characterized in that extract image HOG in the S3
The detailed process of feature is:
S31 converts the image into polar coordinates;
S32 calculates the HOG feature of each pixel in image;
S33 traverses all pixels point, and the HOG feature vector of each pixel is formed multidimensional HOG feature vector group;
S34 calculates the HOG feature of image: the multidimensional HOG feature vector group in S3.3 being normalized by L2 norm, is converted to
One HOG feature vector.
3. article clean level monitoring method according to claim 1, which is characterized in that in S3, with the resulting image of S34
HOG characteristic value is input sample, obtains SVM classifier based on the training of HOG feature, in-between node layer be image HOG feature with
Support vector machines nonlinear operation exports the linear combination of the corresponding supporting vector of each intermediate node.
4. article clean level monitoring method according to claim 1, which is characterized in that be based on FASTER- in the S5
The process that RCNN model training goes out a neural network classifier is:
S51 carries out manual segmentation to the sordid image in part, is partitioned into the region of contamination with non-contamination, and respectively with difference
Color mark comes out;
S52 selects most suitable parameter training to go out the optimal neural network classifier of effect.
5. article clean level monitoring method according to claim 1, which is characterized in that in S4, cleanliness numerical value n is 0
Or 1;In S6, cleanliness numerical value m is 0~1.
6. article clean level monitoring method according to claim 1, which is characterized in that in S7, by calculation formula: 0.5*
Cleanliness numerical value n+0.5* cleanliness numerical value m calculates final cleanliness as a result, cleanliness result is threshold in 0~1.5, S8
Value is set as 1.0.
7. a kind of article clean level monitoring device, which is characterized in that including master control borad, master control borad is connected by onboard USB interface
Digital micro-analysis camera is connect, is additionally provided with wireless transport module on master control borad;Master control borad integrates and can implement right such as and wants
Method described in asking 1.
8. article clean level monitoring device according to claim 7, which is characterized in that the master control borad is aobvious by number
Micro- camera directly acquires article surface image to be measured or by wireless transport module indirect gain article surface image to be measured.
9. article clean level monitoring device according to claim 7 or 8, which is characterized in that the wireless transport module
For Wi-Fi, ZigBee, bluetooth or infrared.
10. article clean level monitoring device according to claim 7, which is characterized in that the master control borad is also connected with
Storage unit.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910019824.6A CN109685038A (en) | 2019-01-09 | 2019-01-09 | A kind of article clean level monitoring method and its device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910019824.6A CN109685038A (en) | 2019-01-09 | 2019-01-09 | A kind of article clean level monitoring method and its device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109685038A true CN109685038A (en) | 2019-04-26 |
Family
ID=66192771
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910019824.6A Pending CN109685038A (en) | 2019-01-09 | 2019-01-09 | A kind of article clean level monitoring method and its device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109685038A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112084851A (en) * | 2020-08-04 | 2020-12-15 | 珠海格力电器股份有限公司 | Hand hygiene effect detection method, device, equipment and medium |
WO2021018175A1 (en) * | 2019-07-29 | 2021-02-04 | Goldway Technology Limited | Process and system for diamond clarity measurement |
CN116337722A (en) * | 2023-05-31 | 2023-06-27 | 国网湖北省电力有限公司超高压公司 | Hydrophobicity monitoring method and system based on image processing |
CN116651847A (en) * | 2023-06-07 | 2023-08-29 | 江苏省人民医院(南京医科大学第一附属医院) | Ultrasonic cleaning effect detection system for medical instrument |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105445283A (en) * | 2016-02-01 | 2016-03-30 | 成都通甲优博科技有限责任公司 | Detection method for filthy conditions of insulator images |
CN108038513A (en) * | 2017-12-26 | 2018-05-15 | 北京华想联合科技有限公司 | A kind of tagsort method of liver ultrasonic |
CN108364006A (en) * | 2018-01-17 | 2018-08-03 | 超凡影像科技股份有限公司 | Medical Images Classification device and its construction method based on multi-mode deep learning |
-
2019
- 2019-01-09 CN CN201910019824.6A patent/CN109685038A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105445283A (en) * | 2016-02-01 | 2016-03-30 | 成都通甲优博科技有限责任公司 | Detection method for filthy conditions of insulator images |
CN108038513A (en) * | 2017-12-26 | 2018-05-15 | 北京华想联合科技有限公司 | A kind of tagsort method of liver ultrasonic |
CN108364006A (en) * | 2018-01-17 | 2018-08-03 | 超凡影像科技股份有限公司 | Medical Images Classification device and its construction method based on multi-mode deep learning |
Non-Patent Citations (1)
Title |
---|
文少波 等,, 东南大学出版社, pages: 173 - 52 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021018175A1 (en) * | 2019-07-29 | 2021-02-04 | Goldway Technology Limited | Process and system for diamond clarity measurement |
US11137355B2 (en) | 2019-07-29 | 2021-10-05 | Goldway Technology Limited | Process and system for diamond clarity measurement |
CN112084851A (en) * | 2020-08-04 | 2020-12-15 | 珠海格力电器股份有限公司 | Hand hygiene effect detection method, device, equipment and medium |
CN116337722A (en) * | 2023-05-31 | 2023-06-27 | 国网湖北省电力有限公司超高压公司 | Hydrophobicity monitoring method and system based on image processing |
CN116337722B (en) * | 2023-05-31 | 2023-10-13 | 国网湖北省电力有限公司超高压公司 | Hydrophobicity monitoring method and system based on image processing |
CN116651847A (en) * | 2023-06-07 | 2023-08-29 | 江苏省人民医院(南京医科大学第一附属医院) | Ultrasonic cleaning effect detection system for medical instrument |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109685038A (en) | A kind of article clean level monitoring method and its device | |
CN105574527B (en) | A kind of quick object detecting method based on local feature learning | |
JP6305171B2 (en) | How to detect objects in a scene | |
CN106023257B (en) | A kind of method for tracking target based on rotor wing unmanned aerial vehicle platform | |
CN106056597B (en) | Object visible detection method and device | |
CN110363088B (en) | Self-adaptive skin inflammation area detection method based on multi-feature fusion | |
Punn et al. | Classification of wheat grains using machine algorithms | |
CN103544499B (en) | The textural characteristics dimension reduction method that a kind of surface blemish based on machine vision is detected | |
CN107016353B (en) | A kind of integrated method and system of variable resolution target detection and identification | |
CN108629319B (en) | Image detection method and system | |
CN109409289A (en) | A kind of electric operating safety supervision robot security job identifying method and system | |
CN106886216A (en) | Robot automatic tracking method and system based on RGBD Face datections | |
Chen et al. | Facial expression recognition based on edge detection | |
CN110096980A (en) | Character machining identifying system | |
CN104197836A (en) | Vehicle lock assembly size detection method based on machine vision | |
CN112862757A (en) | Weight evaluation system based on computer vision technology and implementation method | |
CN109115775B (en) | Betel nut grade detection method based on machine vision | |
CN107358155A (en) | A kind of funny face motion detection method and device and vivo identification method and system | |
Ashmawi et al. | Fitme: Body measurement estimations using machine learning method | |
CN117197700A (en) | Intelligent unmanned inspection contact net defect identification system | |
CN116703895A (en) | Small sample 3D visual detection method and system based on generation countermeasure network | |
CN109635684A (en) | A kind of food traceability system | |
TWI749714B (en) | Method for defect detection, method for defect classification and system thereof | |
Han et al. | Detection for power line inspection | |
CN109360289B (en) | Power meter detection method fusing inspection robot positioning information |
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: 20190426 |