CN108846831A - The steel strip surface defect classification method combined based on statistical nature and characteristics of image - Google Patents
The steel strip surface defect classification method combined based on statistical nature and characteristics of image Download PDFInfo
- Publication number
- CN108846831A CN108846831A CN201810524655.7A CN201810524655A CN108846831A CN 108846831 A CN108846831 A CN 108846831A CN 201810524655 A CN201810524655 A CN 201810524655A CN 108846831 A CN108846831 A CN 108846831A
- Authority
- CN
- China
- Prior art keywords
- defect
- sample
- image
- statistical nature
- vector
- 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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
Based on the steel strip surface defect classification method that statistical nature and characteristics of image combine, including step:Collect the training sample set of one group of tape label;S2, the angle point and each corner description for extracting each defect sample image;The corner description of whole defect sample images is subjected to unsupervised learning cluster in K dimension space;The M of defect sample is tieed up into image feature vector and N-dimensional statistical nature vector merges, forms the M+N dimensional feature vector of defect sample;Supervised learning training is carried out to M+N dimensional feature vector using adaptive boosting tree training method, self study classifier B is trained, exports the classification results of defect.The present invention distinguishes different defects, improves the classification accuracy rate to steel strip surface defect;Real-time online detects classification when surface defect using self study classifier to the defect progress detected automatically and accurately, and classifying rules is manually entered without artificial rely on, but machine learning techniques is used to pass through supervised learning acquisition.
Description
Technical field
The invention belongs to metallurgy industry steel strip surface defect detection system field, especially the classification field of surface defect,
More particularly to a kind of steel strip surface defect classification method combined based on statistical nature and characteristics of image.
Background technique
Steel strip surface defect is a key factor for influencing Cold-strip Steel Surface quality, directly affects final products
Appearance and service performance.Surface detecting system is scanned belt steel surface using camera sensor, obtains the two of belt steel surface
Image is tieed up, and surface defect is detected and classified using machine vision technique, relies on various advanced algorithm errors at present
The accuracy rate of detection has been up to 98%, has substantially met production requirement, but the classification accuracy rate of defect is barely satisfactory always,
General only 80% or so, it is not able to satisfy the requirement that iron and steel enterprise promotes control of product quality and production technology.
Summary of the invention
The technical problem to be solved by the present invention is to provide for existing steel strip surface defect classification above shortcomings
A kind of steel strip surface defect classification method combined based on statistical nature and characteristics of image improves strip surface quality detection system
The classification accuracy rate united to steel strip surface defect;Real-time online detects when surface defect using self study classifier to detecting
Defect carry out classification automatically and accurately, classifying rules is without artificial by being manually entered.
Used technical solution is the present invention to solve above-mentioned technical problem:
Based on the steel strip surface defect classification method that statistical nature and characteristics of image combine, specifically comprise the following steps:
S1, defect sample is collected, and according to the statistical nature of defect sample, including area, perimeter, length-width ratio, position, with
And the appearance of defect sample image classifies to each defect sample, generates the training sample set of one group of tape label, to certainly
The training of Study strategies and methods just closes progress in above-mentioned training sample set, it is assumed that includes W defect sample and C class defect;
S2, the angle point and each corner description for extracting each defect sample image, corner description reflect angle point and neighbour
The relationship of domain pixel, it is assumed that corner description is the vector of K dimension;
S3, the corner description of whole defect sample images is subjected to unsupervised learning cluster in K dimension space, cluster numbers are
M, these classes are L (1), L (2), L (3) ... L (M), and class center is LC (1), LC (2), LC (3) ... LC (M);
S4, probability distribution (characteristics of image) of the corner description of each defect sample in above-mentioned M class is calculated, that is, counted
The corner description of each defect sample belong to L (1), L (2), L (3) ... L (M) these classes number, and be arranged in a M tie up to
Amount, the M for thus constituting a defect sample tie up image feature vector;
S5, the N number of statistical nature (such as area, perimeter, position) for extracting each defect sample, these statistical natures constitute
The N-dimensional statistical nature vector of one defect sample;
S6, the M of defect sample is tieed up to image feature vector and the merging of N-dimensional statistical nature vector, forms the M+N of defect sample
Dimensional feature vector;
S7, supervision has been carried out to the W M+N dimensional feature vectors with classification marker using adaptive boosting tree training method
Learning training trains self study classifier B, and self study classifier B classifies automatically to the defect detected (will be different
Defect distinguishes), export the classification results of defect.
According to the above scheme, it when real-time online detects, is taken when there is new defect to be detected using self study classifier B
Following steps carry out defect classification to new defect, mainly include step:
I) the K dimension description of the angle point and each angle point of the defect image is extracted, and calculates the K dimension description of angle point in L
(1), L (2), the probability distribution in L (3) ... L (M) these classifications, the M as defect tie up image feature vector;
Ii the N-dimensional statistical nature vector of the defect) is calculated;
Iii the M of defect) is tieed up into image feature vector and N-dimensional statistical nature vector merges, forms the M+N dimensional feature of defect
M+N dimensional feature vector is input to and is predicted to obtain the classification results of the defect in self study classifier B by vector.
Compared with prior art, the present invention has the advantages that:
1, defect statistics feature and characteristics of image are combined to the feature vector to form defect, instructed in this feature vector space
Practice self study classifier, different defects is distinguished, improves strip surface quality detection system and steel strip surface defect is divided
Class accuracy;
2, real-time online, which detects, carries out automatic and standard to the defect detected using self study classifier when surface defect
True classification, classifying rules are manually entered without artificial rely on, but are used machine learning techniques to pass through supervised learning and obtained.
Detailed description of the invention
Fig. 1 is that the present invention is based on the steel strip surface defect classification method flow charts that statistical nature and characteristics of image combine;
Fig. 2 is the assorting process figure that the embodiment of the present invention detects new defect.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.
Traditional steel strip surface defect classification relies primarily on the statistical nature for defect and sets rule manually to realize, than
It is such as directed to punching defect, can usually be set as like circle property<1.1 and gray value mean value<40 and width<Rule as 50mm.
But it is difficult to describe the defects of specific features for comparing, such as phosphorus spot becomes extremely difficult if sticking up skin etc. and manually laying down a regulation.With
The progress of machine learning techniques, carries out defect classification by the way of sample self study at this stage, that is, provides lineup's work label
The sample of classification results is crossed, then using machine learning techniques by computer Automatic-searching to the rule classified in this way.It is general
All defect is described with the statistical nature of defect, then training classifier, statistical nature in the dimension of statistical nature
It is some features relevant to the position of defect, shape, texture, gray scale, topological structure etc., mainly includes:Defect length and width, face
Product, perimeter, length-width ratio, compactness, the distance to strip steel boundary, position of centre of gravity, seemingly round, gray average, gray variance, Euler
Number etc..Statistical nature is the summary and refinement to defect information, it is largely the macroscopic view embodiment of defect characteristic.It uses
Statistical nature classifying quality when gross feature differs greatly between defect classification is preferable, for example distinguishes hickie and stain, distinguishes weldering
Seam and hole.When the gross feature difference very little of defect, classification accuracy will be greatly reduced, as distinguish scab and dampen it is this kind of
Defect.The corner feature of image preferably describes the details and texture information of image, and the angle of image is often utilized in practical application
Point feature carries out Face datection and pedestrian's identification, corner feature is also with the visual signature of the mankind closer to being more concerned about image
The point of upper gray scale mutation.But if only image corner feature taxonomic defficiency is also inadequate, because the classification of many defects is not
It is only related to its appearance, and have very big relationship with its statistical property, therefore the present invention considers the statistical nature of defect
It is combined with the corner feature of defect image, takes their union as the eigen vector of defect, and use machine learning algorithm
Automatic training classifier classifies to defect.
Shown in referring to Fig.1, the present invention is based on the steel strip surface defect classification method that statistical nature and characteristics of image combine,
Defect statistics information and image information are combined to train self study classifier, and using self study classifier to detecting
Defect is classified automatically, is specifically comprised the following steps:
S1, a large amount of typical defect sample is collected, and by professional according to the statistical nature of defect sample, such as area,
The appearance of perimeter, length-width ratio, position etc. and defect sample image classifies to each defect sample, to self study classifier
Training just close progress in above-mentioned training sample set, it is assumed that include W defect sample and C class defect, this step is for producing
The training sample set of raw one group of tape label, not only needs defect sample image to be also required to when manually carrying out defect classification
Statistical nature could correctly determine the classification of defect;
The K of S2, the angle point for extracting each defect sample image and each angle point tie up description, this step may be selected to use
FAST angle point is as angular-point detection method, and angle point is a kind of important local feature of image, it determines the shape of target in image
Shape, so there is important application in images match, goal description and identification, angle point is image grey scale change in two-dimensional space
Violent position is the pixel for having notable difference with the adjoint point of surrounding;Since camera is that fixed position is clapped in defects detection
It takes the photograph, bias light interference is also smaller, and not needing angle point has rotation and scaling invariance, it is possible to fastest using detecting
FAST angle point;
S3, the corner description of whole defect sample images is subjected to unsupervised learning cluster in K dimension space, cluster numbers are
M, these classes are respectively L (1), L (2), L (3) ... L (M), and class center is LC (1), LC (2), LC (3) ... LC (M);
S4, probability distribution of the corner description for generating each defect sample in above-mentioned M class is calculated, that is, counts each lack
Fall into sample corner description belong to L (1), L (2), L (3) ... L (M) these classes number, thus constitute a defect sample M
Tie up image feature vector;It must be a vector by these Feature Compressions because the angle point number of different samples is inconsistent
It is described, step S3, S4 clusters the description vectors of all angle points in description vectors dimension, and obtains M feature
Thus the corner description of multiple vectors of image is converted to probability distribution of its corner description on M cluster centre by center
Single vector description, this step can be considered Feature Dimension Reduction process;
S5, N number of statistical nature that defect is just calculated when each defect sample is detected, these statistical natures constitute
The N-dimensional statistical nature vector of one defect sample;This step is actually just to complete in real time in defects detection, because such as position
Setting when feature etc. only has defects detection could obtain, and location information can not be only obtained from defect sample image;
S6, the M of defect sample is tieed up to image feature vector and the merging of N-dimensional statistical nature vector, forms the M+N of defect sample
Dimensional feature vector;The statistical nature and characteristics of image of this step defect sample, by defect with one comprising statistical nature and
The M+N dimensional feature vector of image feature information is characterized;
S7, supervision has been carried out to the W M+N dimensional feature vectors with classification marker using adaptive boosting tree training method
Learning training trains self study classifier B, and self study classifier B classifies automatically to the defect detected, will be different
Defect distinguishes, and exports the classification results of defect;Commonly classifier training method includes:Random forest, adaptive boosting tree
With support vector machines etc.;Here the feature used of classifying includes statistical nature and characteristics of image, the two belong to different physical quantities and
Numerical value is widely different, belongs to blended data, using when random forest, adaptive boosting tree it is not necessary that place is normalized in data
Reason, and then need first will just to can be carried out classification using support vector machines after data normalization, data prediction is more complex;Actually adopt
When being assessed with generalization ability of the Cross-Validation technique to classifier, classification accuracy rate of the adaptive boosting tree on test set
Highest, about 85%, so carrying out classifier training used here as adaptive boosting tree training method.
Referring to shown in Fig. 2, when real-time online detects, just needed according to above-mentioned step after thering is new defect to be detected
Defect is described as M+N dimensional feature vector and is classified using trained self study classifier B to new defect, it is main
Including step:
I) the K dimension description of the angle point and each angle point of the defect image is extracted, and calculates it in L (1), L (2), L
(3) ... the probability distribution in L (M) these classifications, the feature for calculating each angle point describe C to M cluster centre LC (1), LC
(2), then the distance of (3) LC ... LC (M) finds out classification corresponding apart from the smallest center, if C is apart from classification LC (x)
Central point it is nearest, then the angle point belongs to L (x) class, the value of image feature vector I (x) adds 1, traverses each of defect image
The M dimension image feature vector of the defect is obtained after the feature description of a angle point;
Ii the N-dimensional statistical nature vector of the defect) is calculated;
Iii the M of defect) is tieed up into image feature vector and N-dimensional statistical nature vector merges, forms the M+N dimensional feature of defect
Vector inputs into self study classifier B and is predicted to obtain the classification results of the defect.
The above is presently preferred embodiments of the present invention, but the present invention should not be limited to the embodiment and attached drawing institute
Disclosure.So all do not depart from the lower equivalent or modification completed of spirit disclosed in this invention, guarantor of the present invention is both fallen within
The range of shield.
Claims (2)
1. the steel strip surface defect classification method combined based on statistical nature and characteristics of image, which is characterized in that specifically include
Following steps:
S1, defect sample is collected, and according to the statistical nature of defect sample, including area, perimeter, length-width ratio, position, and lacked
The appearance for falling into sample image classifies to each defect sample, the training sample set of one group of tape label is generated, to self study
The training of classifier just closes progress in above-mentioned training sample set, it is assumed that includes W defect sample and C class defect;
S2, the angle point and each corner description for extracting each defect sample image, corner description reflect angle point and neighborhood picture
The relationship of vegetarian refreshments, it is assumed that corner description is the vector of K dimension;
S3, the corner description of whole defect sample images carries out to unsupervised learning cluster in K dimension space, cluster numbers M, this
A little classes are L (1), L (2), L (3) ... L (M), and class center is LC (1), LC (2), LC (3) ... LC (M);
S4, probability distribution of the corner description of each defect sample in above-mentioned M class is calculated, that is, counts each defect sample
Corner description belong to L (1), L (2), L (3) ... L (M) these classes number, and be arranged in a M dimensional vector, thus constitute one
The M of a defect sample ties up image feature vector;
S5, the N number of statistical nature for extracting each defect sample, these statistical natures constitute the N-dimensional statistics an of defect sample
Feature vector;
S6, the M of defect sample is tieed up to image feature vector and the merging of N-dimensional statistical nature vector, forms the M+N Wei Te of defect sample
Levy vector;
S7, supervised learning is carried out to the W M+N dimensional feature vectors with classification marker using adaptive boosting tree training method
Training, trains self study classifier B, and self study classifier B classifies automatically to the defect detected, exports point of defect
Class result.
2. the steel strip surface defect classification method combined as described in claim 1 based on statistical nature and characteristics of image,
It is characterized in that, when real-time online detects, takes following steps pair using self study classifier B when there is new defect to be detected
New defect carries out defect classification, mainly includes step:
I) the K dimension description of the angle point and each angle point of the defect image is extracted, and calculates the K dimension description of angle point in L (1), L
(2), the probability distribution in (3) L ... L (M) these classifications, the M as defect tie up image feature vector;
Ii the N-dimensional statistical nature vector of the defect) is calculated;
Iii the M of defect) is tieed up into image feature vector and N-dimensional statistical nature vector merges, forms the M+N dimensional feature vector of defect,
M+N dimensional feature vector is input to and is predicted to obtain the classification results of the defect in self study classifier B.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810524655.7A CN108846831B (en) | 2018-05-28 | 2018-05-28 | Band steel surface defect classification method based on combination of statistical characteristics and image characteristics |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810524655.7A CN108846831B (en) | 2018-05-28 | 2018-05-28 | Band steel surface defect classification method based on combination of statistical characteristics and image characteristics |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108846831A true CN108846831A (en) | 2018-11-20 |
CN108846831B CN108846831B (en) | 2021-09-28 |
Family
ID=64207911
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810524655.7A Active CN108846831B (en) | 2018-05-28 | 2018-05-28 | Band steel surface defect classification method based on combination of statistical characteristics and image characteristics |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108846831B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109657718A (en) * | 2018-12-19 | 2019-04-19 | 广东省智能机器人研究院 | SPI defect classification intelligent identification Method on a kind of SMT production line of data-driven |
CN110135477A (en) * | 2019-04-28 | 2019-08-16 | 湖北工业大学 | A kind of strip surface quality defect classifier and its classification method based on serial/parallel row integrated study frame |
CN111539938A (en) * | 2020-04-26 | 2020-08-14 | 中冶赛迪重庆信息技术有限公司 | Method, system, medium and electronic terminal for detecting curvature of rolled strip steel strip head |
CN113808136A (en) * | 2021-11-19 | 2021-12-17 | 中导光电设备股份有限公司 | Liquid crystal screen defect detection method and system based on nearest neighbor algorithm |
US11205260B2 (en) | 2019-11-21 | 2021-12-21 | International Business Machines Corporation | Generating synthetic defect images for new feature combinations |
CN113838043A (en) * | 2021-09-30 | 2021-12-24 | 杭州百子尖科技股份有限公司 | Machine vision-based quality analysis method in metal foil manufacturing |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8615125B2 (en) * | 2010-10-08 | 2013-12-24 | Omron Corporation | Apparatus and method for inspecting surface state |
CN103631932A (en) * | 2013-12-06 | 2014-03-12 | 中国科学院自动化研究所 | Method for detecting repeated video |
CN103745234A (en) * | 2014-01-23 | 2014-04-23 | 东北大学 | Band steel surface defect feature extraction and classification method |
CN107392211A (en) * | 2017-07-19 | 2017-11-24 | 苏州闻捷传感技术有限公司 | The well-marked target detection method of the sparse cognition of view-based access control model |
-
2018
- 2018-05-28 CN CN201810524655.7A patent/CN108846831B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8615125B2 (en) * | 2010-10-08 | 2013-12-24 | Omron Corporation | Apparatus and method for inspecting surface state |
CN103631932A (en) * | 2013-12-06 | 2014-03-12 | 中国科学院自动化研究所 | Method for detecting repeated video |
CN103745234A (en) * | 2014-01-23 | 2014-04-23 | 东北大学 | Band steel surface defect feature extraction and classification method |
CN107392211A (en) * | 2017-07-19 | 2017-11-24 | 苏州闻捷传感技术有限公司 | The well-marked target detection method of the sparse cognition of view-based access control model |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109657718A (en) * | 2018-12-19 | 2019-04-19 | 广东省智能机器人研究院 | SPI defect classification intelligent identification Method on a kind of SMT production line of data-driven |
CN109657718B (en) * | 2018-12-19 | 2023-02-07 | 广东省智能机器人研究院 | Data-driven SPI defect type intelligent identification method on SMT production line |
CN110135477A (en) * | 2019-04-28 | 2019-08-16 | 湖北工业大学 | A kind of strip surface quality defect classifier and its classification method based on serial/parallel row integrated study frame |
CN110135477B (en) * | 2019-04-28 | 2023-03-24 | 湖北工业大学 | Strip steel surface quality defect classifier based on serial/parallel integrated learning framework and classification method thereof |
US11205260B2 (en) | 2019-11-21 | 2021-12-21 | International Business Machines Corporation | Generating synthetic defect images for new feature combinations |
CN111539938A (en) * | 2020-04-26 | 2020-08-14 | 中冶赛迪重庆信息技术有限公司 | Method, system, medium and electronic terminal for detecting curvature of rolled strip steel strip head |
CN111539938B (en) * | 2020-04-26 | 2022-12-16 | 中冶赛迪信息技术(重庆)有限公司 | Method, system, medium and electronic terminal for detecting curvature of rolled strip steel strip head |
CN113838043A (en) * | 2021-09-30 | 2021-12-24 | 杭州百子尖科技股份有限公司 | Machine vision-based quality analysis method in metal foil manufacturing |
CN113808136A (en) * | 2021-11-19 | 2021-12-17 | 中导光电设备股份有限公司 | Liquid crystal screen defect detection method and system based on nearest neighbor algorithm |
Also Published As
Publication number | Publication date |
---|---|
CN108846831B (en) | 2021-09-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108846831A (en) | The steel strip surface defect classification method combined based on statistical nature and characteristics of image | |
CN112287788A (en) | Pedestrian detection method based on improved YOLOv3 and improved NMS | |
CN103886589B (en) | Object-oriented automated high-precision edge extracting method | |
Lin et al. | Estimation of number of people in crowded scenes using perspective transformation | |
CN102117413B (en) | Method for automatically filtering defective image based on multilayer feature | |
CN105608441B (en) | Vehicle type recognition method and system | |
CN112861635B (en) | Fire disaster and smoke real-time detection method based on deep learning | |
CN108470354A (en) | Video target tracking method, device and realization device | |
CN104881671B (en) | A kind of high score remote sensing image Local Feature Extraction based on 2D Gabor | |
CN107230203A (en) | Casting defect recognition methods based on human eye vision attention mechanism | |
CN102903122A (en) | Video object tracking method based on feature optical flow and online ensemble learning | |
CN107292310A (en) | A kind of circular pointer dial plate vision positioning and automatic reading method | |
CN104464079A (en) | Multi-currency-type and face value recognition method based on template feature points and topological structures of template feature points | |
CN103136504A (en) | Face recognition method and device | |
CN102208038A (en) | Image classification method based on visual dictionary | |
CN109815859A (en) | A kind of inferior grade road automatically extracts and mutation analysis method | |
CN104463199A (en) | Rock fragment size classification method based on multiple features and segmentation recorrection | |
CN104036284A (en) | Adaboost algorithm based multi-scale pedestrian detection method | |
CN116205919A (en) | Hardware part production quality detection method and system based on artificial intelligence | |
CN107392929A (en) | A kind of intelligent target detection and dimension measurement method based on human vision model | |
CN101162503A (en) | Method for extracting and recognizing human ear characteristic by improved Hausdorff distance | |
CN110826408B (en) | Face recognition method by regional feature extraction | |
CN104077594A (en) | Image recognition method and device | |
Liu et al. | A classification method of glass defect based on multiresolution and information fusion | |
CN113221881B (en) | Multi-level smart phone screen defect detection method |
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 |