CN109614977A - A kind of hub type recognition methods - Google Patents
A kind of hub type recognition methods Download PDFInfo
- Publication number
- CN109614977A CN109614977A CN201811352681.2A CN201811352681A CN109614977A CN 109614977 A CN109614977 A CN 109614977A CN 201811352681 A CN201811352681 A CN 201811352681A CN 109614977 A CN109614977 A CN 109614977A
- Authority
- CN
- China
- Prior art keywords
- model
- wheel hub
- characteristic
- image
- point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- 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
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/757—Matching configurations of points or features
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of hub type recognition methods, comprising steps of (1) establishes model data library: feature being stored in model data library to its standard wheel hub image zooming-out feature for the wheel hub of each model；(2) online recognition: wheel hub image to be checked is obtained, after feature extraction, obtain inspected feature data, inspected feature data are successively matched with the characteristic of each model in model data library, postsearch screening is carried out by using the characteristic point of Flann algorithmic match wheel hub pair, then with RANSAC algorithm.The present invention also proposes that multiple rotary further knows method for distinguishing.The method of the present invention can preferably overcome illumination, hub positions, wheel hub overlap etc. to interfere, and have the advantages that detection accuracy height, using flexible, can satisfy the requirement of on-line checking.
Description
Technical field
The present invention relates to hub type identification research, and in particular to one kind passes through RANSAC algorithm and multiple rotary
It carries out hub type and knows method for distinguishing.
Background technique
There are many kinds of classes for the wheel hub of automobile or motorcycle, and labeled as different hub types, wheel hub is in automatic processing or certainly
The identification for carrying out hub type to the wheel hub being sent to first is required when dynamic detection, with Auto-matching process equipment or detection device
Parameter.
The demand of hub type automatic identification is more and more, therefore also higher and higher to the requirement of the accuracy rate of identification, in public affairs
The method proposed in the patent CN101079107A opened is to go background with subtractive method to real scene shooting wheel hub, then taken turns by binaryzation
Hub shape feature is compared to identify hub type with the boss shape feature in model data library.This method will receive
The very big influence of illumination, binarization threshold etc., and due to technological factors such as wheel hub overlaps it is difficult to ensure that the shape feature of wheel hub
Consistency, therefore accuracy rate is extremely difficult to actual production demand.
Summary of the invention
The purpose of the present invention is to overcome the shortcomings of the existing technology and deficiency, provides a kind of hub type recognition methods, should
Method can preferably overcome illumination, hub positions, wheel hub overlap etc. to interfere, and have the advantages that detection accuracy height, using flexible,
It can satisfy the requirement of on-line checking.
The purpose of the present invention is realized by the following technical solution: a kind of hub type recognition methods, comprising steps of
(1) model data library is established: will be special to its standard wheel hub image zooming-out feature for the wheel hub of each model
Sign is stored in model data library；
(2) online recognition: obtaining wheel hub image to be checked, after feature extraction, obtains inspected feature data, by spy to be checked
Sign data are successively matched with the characteristic of each model in model data library, and step is:
Using Flann, (fast library for approximate nearest neighbors, arest neighbors are quickly close
Like matching) algorithm matches inspected feature data with the characteristic of current versions in model data library, by successful match
Characteristic to being stored in the first screened container；
Using RANSAC (Random Sample Consensus, random sample consensus) algorithm to the first screened container
Interior characteristic further judges this feature data to whether matching, if it does, being deposited into postsearch screening is carried out
Second screened container；
Using the corresponding model of successful match number maximum value in the second screened container as the model of current wheel hub to be measured.
The present invention then with RANSAC algorithm carries out postsearch screening by using the characteristic point of Flann algorithmic match wheel hub pair,
It is greatly improved matched accuracy, enables the algorithm to the requirement for reaching online recognition.
Preferably, before to wheel hub image zooming-out feature, wheel hub outer profile is extracted first, method is:
(1-1) is filtered denoising and edge extracting to the wheel hub image of acquisition, obtains edge image；
(1-2) looks for circle with Hough algorithm on the wheel hub image of acquisition, by the central coordinate of circle and radius of each circle found
It is stored in the first array；
(1-3) according to the central coordinate of circle and radius of circle each in the first array in wheel hub image upper drawing circle, and one by one with side
Edge image compares, will be with the highest circle of edge image registration as wheel hub outer profile circle；
(1-4) removes all image informations outside wheel hub outer profile circle in original image, is stored as characteristic image.
Further, the filtering uses gaussian filtering method, and edge extracting uses canny operator.
Preferably, when establishing model data library, using ORB (Oriented FAST and Rotated BRIEF) algorithm
Feature point extraction is carried out to characteristic image；Characteristic point Criterion of Selecting therein are as follows: obtain pixel current in wheel hub image
The gray value of the pixel of gray value and the vertex neighborhood, if the gray value difference of the gray value and neighborhood each point above limits
Value, it is believed that the point is bright spot or dim spot in wheel hub, if partial dot gray value differences value is less than limit value in the gray value and neighborhood
It but is more than limit value with the gray value difference of multiple points continuous in neighborhood, then it is assumed that the point is the angle point of wheel hub；These are had
The point of key message is set as characteristic point, is stored in the first data capsule；
It saves the corresponding hub type of Current standards wheel hub image, hub radius, spoke and counts to the second data capsule；
Model data library is established according to the data of the first data capsule and the second data capsule.
Preferably, using Flann algorithm by the characteristic of each model in inspected feature data and model data library into
Row matching, step is:
(2-1-1) is with Flann algorithm to the characteristic of one of model in characteristic to be checked and model data library
It is matched, obtains matching double points matrix image；Characteristic point data therein is described with binary stream in a computer；Matching
The step of are as follows: the angle point of each characteristic point to be detected such as wheel hub, the bright spot of wheel hub, dim spot of wheel hub etc. are traversed, in model
The point and secondary close point that type is identical, Euclidean distance is nearest are found in all characteristic points of a model in database,
Saving the two distances is respectively nearest neighbor distance and time nearest neighbor distance, and final calculation result is saved as matching double points matrix diagram
Picture；
(2-1-2) sets a first threshold, to each pair of match point, compares nearest neighbor distance and time nearest neighbor distance, if most
Nearest neighbor distance is less than first threshold divided by the ratio that secondary nearest neighbor distance obtains, it is believed that this is higher to match point similarity, and should
Matching double points are as characteristic to being stored in the first screened container.
Preferably, using RANSAC algorithm to the characteristic in the first screened container to carry out postsearch screening, step is:
4 pairs of not conllinear characteristics pair are randomly choosed in current signature data pair, calculate its mapping matrix model, and
Under this mapping matrix model, if in characteristic point in inspected feature data and model data library in character pair data
The Euclidean distance of characteristic point meets second threshold condition, using the characteristic point in inspected feature data as quadratic character point；It repeats
Above-mentioned mapping matrix model calculating process, finds out the largest number of mapping matrix models of quadratic character point, and by the model
Corresponding quadratic character point is stored in the second screened container.
Preferably, judge whether the maximum value is higher than after successful match number maximum value in obtaining the second screened container
Third threshold value, if it does, then determining otherwise the model of the current wheel hub to be measured of the corresponding model of number maximum value carries out down
Face rotates identification step:
(3-1) repeats to identify again, if with characteristic in model data library if control wheel hub image rotates mass dryness fraction around center
According to successful match, then model is exported, otherwise, record is in time identification, with every kind of model successful match in model data library
With points, (3-2) is entered step；
(3-2) carries out rotating and repeating identification n-1 times again；
(3-3) in n times identification process, the matching points of present image and every kind of model have n, corresponding each model into
Row adduction, therefrom chooses maximum value；If the maximum value is more than the 4th threshold value, the type of the current wheel hub to be checked of corresponding model is determined
Number, otherwise determine that hub type is unidentified.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1, the present invention can preferably overcome illumination, hub positions, wheel hub overlap etc. to interfere, and extract hub positions and feature,
Realize wheel-type matching.
2, the present invention continues to carry out postsearch screening using RANSAC algorithm after match for the first time using Flann algorithm,
The image finished for screening, it is also proposed that feature extraction consistency is ensured by multiple rotary image, improves characteristic matching effect
Fruit.
Detailed description of the invention
Fig. 1 is the flow chart of the present embodiment method.
Fig. 2 is the flow chart that model data library is established in the present embodiment method.
Fig. 3 is the flow chart that identification step is rotated in the present embodiment method.
Specific embodiment
Present invention will now be described in further detail with reference to the embodiments and the accompanying drawings, but embodiments of the present invention are unlimited
In this.
The present embodiment uses a kind of hub type identifying system, carries out for acquiring wheel hub image, and to wheel hub image
Identification, specifically may include host computer, camera module.Wheel hub is placed on filming apparatus, and camera module is fixed on shooting position
The top set, with upper mechatronics.Collected wheel hub image is transmitted to host computer by camera module.Host computer load is originally
The software program for inventing the hub type recognition methods establishes model data library by the software program and carries out hub type
Number identification.
Referring to Fig. 1,2, a kind of hub type recognition methods of the present embodiment, main includes establishing model data library and online
Identify that two steps are with reference to the accompanying drawing specifically described step.
S1, model data library is established
What is stored in model data library is the corresponding characteristic of various model standard wheel hub images, therefore, need to be for every
A kind of wheel hub progress feature extraction of current version.Referring to fig. 2, steps are as follows:
S1.1 finds wheel hub outer profile to collected wheel hub image, and removes the picture material other than wheel-hub contour, i.e.,
Wheel hub background.
Firstly, carrying out gaussian filtering denoising, canny operator lookup edge to the wheel hub image of acquisition, edge image is obtained.
Secondly, circle is looked for Hough algorithm on the wheel hub image of acquisition, by the central coordinate of circle and radius of each circle found
It is stored in the first array.
Then, according to the central coordinate of circle of circle each in the first array and radius in wheel hub image upper drawing circle, and one by one with side
Edge image compares, will be with the highest circle of edge image registration as wheel hub outer profile circle.
Finally, removing all image informations outside wheel hub outer profile circle in original image, obtained image is stored as spy
Levy image.
S1.2 carries out feature point extraction to characteristic image with ORB algorithm, is saved in the first data capsule.The extraction of characteristic point
Step is the gray value of the gray value of pixel current in wheel hub image and the pixel of the vertex neighborhood to be obtained, if the point
The gray value difference of gray value and neighborhood each point is all larger, it is believed that the point is bright spot or dim spot in wheel hub, if the gray scale
It is worth smaller with partial dot gray value differences value in neighborhood but larger with the gray value difference of multiple points continuous in neighborhood, then it is assumed that should
Point is angle point of wheel hub, etc.；These points with key message are set as characteristic point.
S1.3 reads model, hub radius, the wheel divergence of the wheel hub of user's input, is saved in the second data capsule.
The data of first data capsule and the second data capsule are saved in model data library by S1.4.
S1.5 executes step S1.1 to S1.4 to all new model wheel hubs, establishes model data library.
S2, online recognition
Referring to Fig. 1, wheel hub image to be checked is obtained, method is processed similarly according to step S1.1, S1.2, carries out feature extraction,
Inspected feature data are obtained, for inspected feature data, it is matched with model data library, are worked as according to matching result judgement
The model of front hub.Steps are as follows:
Inspected feature data are stored in data capsule to be checked by S2.1.
S2.2 reads a kind of hubless feature data of model from model data library, is stored in third data capsule.
S2.3 matches data capsule to be checked with third data capsule with Flann algorithm, characteristic point data therein
It is described in a computer with binary stream；Matched committed step is to traverse each of data capsule to be checked characteristic point
As the angle point of wheel hub, the bright spot of wheel hub, wheel hub dim spot, in all characteristic points of third data capsule find type it is identical,
The nearest point of Euclidean distance and secondary close point save the two distances respectively nearest neighbor distance and time nearest neighbor distance, will be final
Calculated result saves as matching double points matrix image.
Obtain matching double points image, set a threshold value T, to each pair of match point, compare nearest neighbor distance and time neighbour away from
From if the ratio r atio that nearest neighbor distance is obtained divided by secondary nearest neighbor distance makees the match point less than determining threshold value T
Data are characterized to being stored in the first screened container.
S2.4 carries out postsearch screening to the first screened container using RANSAC algorithm, i.e., random in current signature data pair
4 pairs of not conllinear characteristics pair are selected, its mapping matrix model are calculated, and under this mapping matrix model, by spy to be checked
The Euclidean distance of characteristic point in characteristic point and model data library in sign data in character pair data meets second threshold item
The characteristic point of part is as quadratic character point；It repeats the above process repeatedly, finds out the largest number of mapping matrixes of quadratic character point
Model, and the corresponding quadratic character point of the model is stored in the second screened container.
S2.5 repeats step S2.2 to S2.5, until wheel hub to be identified is matched with models all in model data library completion.
S2.6 searches the maximum value of successful match numerical value in the second screened container, judges whether the value is greater than the threshold value of setting
T1, if so, will identify that model ID is set as the model of the corresponding matching wheel hub of the value, otherwise, it fails to match.
In order to further overcome illumination, hub positions, wheel hub overlap etc. to interfere, Feature Points Matching identification is caused to miss
Difference, the present embodiment propose after it fails to match for the first time, then are identified by multiple rotary that further progress judges.Referring to
Fig. 3, step are:
Wheel hub image is rotated 22.5 ° around center by S3.1, second of invocation step S1, S2 hub type recognizer, when
Host computer exports matched model when successful match, otherwise, executes following step；
Wheel hub image is rotated 45 ° around center by S3.2, third time invocation step S1, S2 hub type recognizer, when
Host computer exports matched model when with success, otherwise, executes following step；
Wheel hub image is rotated 67.5 ° around center by S3.3, third time invocation step S1, S2 hub type recognizer, when
Host computer exports matched model when successful match, otherwise, executes following step；
S3.4 inquires preceding four every kind of models matching points, by the matched matching points difference phase of preceding four every kind of models
Add, obtain the summation of four every kind of matching points, be stored in the second array, traverse the second array, choose maximum value therein,
If this maximum value is greater than a pre-set threshold value, host computer exports the corresponding hub type of this maximum value, otherwise upper
Machine shows " hub type is unidentified ".
It can implement the technology that the present invention describes by various means.For example, these technologies may be implemented in hardware, consolidate
In part, software or combinations thereof.For hardware embodiments, processing module may be implemented in one or more specific integrated circuits
(ASIC), digital signal processor (DSP), programmable logic device (PLD), field-programmable logic gate array (FPGA), place
Manage device, controller, microcontroller, electronic device, other electronic units for being designed to execute function described in the invention or
In a combination thereof.
It, can be with the module of execution functions described herein (for example, process, step for firmware and/or Software implementations
Suddenly, process etc.) implement the technology.Firmware and/or software code are storable in memory and are executed by processor.Storage
Device may be implemented in processor or outside processor.
Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above method embodiment can pass through
The relevant hardware of program instruction is completed, and program above-mentioned can store in a computer-readable storage medium, the program
When being executed, step including the steps of the foregoing method embodiments is executed；And storage medium above-mentioned includes: ROM, RAM, magnetic disk or light
The various media that can store program code such as disk.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (7)
1. a kind of hub type recognition methods, which is characterized in that comprising steps of
(1) it establishes model data library: its standard wheel hub image zooming-out feature being deposited feature for the wheel hub of each model
In model data library；
(2) online recognition: obtaining wheel hub image to be checked, after feature extraction, obtains inspected feature data, by inspected feature number
According to successively being matched with the characteristic of each model in model data library, step is:
Inspected feature data are matched with the characteristic of current versions in model data library using Flann algorithm, general
With successful characteristic to being stored in the first screened container；
This feature number is further judged to postsearch screening is carried out to the characteristic in the first screened container using RANSAC algorithm
According to whether matching, if it does, being deposited into the second screened container；
Using the corresponding model of successful match number maximum value in the second screened container as the model of current wheel hub to be measured.
2. hub type recognition methods according to claim 1, which is characterized in that first before wheel hub image zooming-out feature
Wheel hub outer profile is first extracted, method is:
(1-1) is filtered denoising and edge extracting to the wheel hub image of acquisition, obtains edge image；
(1-2) looks for circle with Hough algorithm on the wheel hub image of acquisition, and the central coordinate of circle of each circle found and radius are stored in
In first array；
(1-3) according to the central coordinate of circle and radius of circle each in the first array in wheel hub image upper drawing circle, and one by one with edge graph
It, will be with the highest circle of edge image registration as wheel hub outer profile circle as comparing；
(1-4) removes all image informations outside wheel hub outer profile circle in original image, is stored as characteristic image.
3. hub type recognition methods according to claim 2, which is characterized in that the filtering uses gaussian filtering side
Method, edge extracting use canny operator.
4. hub type recognition methods according to claim 2, which is characterized in that when establishing model data library, using ORB
Algorithm carries out feature point extraction to characteristic image；Characteristic point Criterion of Selecting therein are as follows: obtain pixel current in wheel hub image
The gray value of the pixel of the gray value and vertex neighborhood of point, if the gray value difference of the gray value and neighborhood each point is all super
Cross limit value, it is believed that the point is bright spot or dim spot in wheel hub, if the gray value is less than with partial dot gray value differences value in neighborhood
Limit value but with the gray value difference of multiple points continuous in neighborhood be more than limit value, then it is assumed that the point is the angle point of wheel hub；By these
Point with key message is set as characteristic point, is stored in the first data capsule；
It saves the corresponding hub type of Current standards wheel hub image, hub radius, spoke and counts to the second data capsule；
Model data library is established according to the data of the first data capsule and the second data capsule.
5. hub type recognition methods according to claim 1, which is characterized in that use Flann algorithm by inspected feature
Data are matched with the characteristic of each model in model data library, and step is:
(2-1-1) is carried out with characteristic of the Flann algorithm to characteristic to be checked and one of model in model data library
Matching obtains matching double points matrix image；Characteristic point data therein is described with binary stream in a computer；Matched step
Suddenly are as follows: traverse each characteristic point to be detected, find type in all characteristic points of a model in model data library
Point and secondary close point identical, Euclidean distance is nearest saves the two distances respectively nearest neighbor distance and time nearest neighbor distance,
Final calculation result is saved as into matching double points matrix image；
(2-1-2) sets a first threshold, to each pair of match point, compares nearest neighbor distance and time nearest neighbor distance, if arest neighbors
The ratio that distance is obtained divided by secondary nearest neighbor distance is less than first threshold, using the matching double points as characteristic to being stored in first
Screened container.
6. hub type recognition methods according to claim 1, which is characterized in that screened using RANSAC algorithm to first
To postsearch screening is carried out, step is characteristic in container:
4 pairs of not conllinear characteristics pair are randomly choosed in current signature data pair, calculate its mapping matrix model, and at this
Under a mapping matrix model, by the characteristic point in inspected feature data and the characteristic point in character pair data in model data library
Euclidean distance meet the characteristic point of second threshold condition as quadratic character point；It repeats the above process repeatedly, finds out secondary spy
The largest number of mapping matrix models of point are levied, and the corresponding quadratic character point of the model is stored in the second screened container.
7. hub type recognition methods according to claim 1, which is characterized in that matched in obtaining the second screened container
After success number maximum value, judge whether the maximum value is higher than third threshold value, if it does, then determining that number maximum value is corresponding
Otherwise the model of the current wheel hub to be measured of model carries out rotating identification step below:
(3-1) repeats to identify again, if with characteristic in model data library if control wheel hub image rotates mass dryness fraction around center
With success, then model is exported, otherwise, record is in time identification, the match point with every kind of model successful match in model data library
Number, enters step (3-2)；
(3-2) carries out rotating and repeating identification n-1 times again；
(3-3) in n times identification process, the matching points of present image and every kind of model have n, and corresponding each model is added
With therefrom choose maximum value；If the maximum value is more than the 4th threshold value, the model of the current wheel hub to be checked of corresponding model is determined,
Otherwise determine that hub type is unidentified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811352681.2A CN109614977A (en) | 2018-11-14 | 2018-11-14 | A kind of hub type recognition methods |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811352681.2A CN109614977A (en) | 2018-11-14 | 2018-11-14 | A kind of hub type recognition methods |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109614977A true CN109614977A (en) | 2019-04-12 |
Family
ID=66003307
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811352681.2A Pending CN109614977A (en) | 2018-11-14 | 2018-11-14 | A kind of hub type recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109614977A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110530863A (en) * | 2019-08-27 | 2019-12-03 | 南京末梢信息技术有限公司 | A kind of automotive hub mixes package detection device and method |
| CN111238370A (en) * | 2020-02-20 | 2020-06-05 | 中国科学院声学研究所东海研究站 | Intelligent detection method and device for KIT board |
| CN111507404A (en) * | 2020-04-17 | 2020-08-07 | 无锡雪浪数制科技有限公司 | Hub model identification method based on deep vision |
| CN112132783A (en) * | 2020-08-21 | 2020-12-25 | 成都飞机工业（集团）有限责任公司 | Part identification method based on digital image processing technology |
| CN116205923A (en) * | 2023-05-06 | 2023-06-02 | 威海锐鑫丰金属科技有限公司 | Nondestructive testing method for internal defects of automobile hub based on X-RAY |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279763A (en) * | 2013-05-25 | 2013-09-04 | 中北大学 | Method for automatically identifying wheel hub types based on structural features |
| CN103425969A (en) * | 2013-08-07 | 2013-12-04 | 华南理工大学 | Detection system and detection method for identifying type of wheel hub |
| CN105548185A (en) * | 2016-01-08 | 2016-05-04 | 浙江科技学院 | Automobile wheel hub screw hole recognition method based on machine vision and shielding method and system |
| CN106372644A (en) * | 2016-08-22 | 2017-02-01 | 保定市立中车轮制造有限公司 | Image identification method used for wheel hub sorting system |
| CN106643566A (en) * | 2016-09-30 | 2017-05-10 | 华南理工大学 | Method of automatically measuring initial angle of spokes of wheel hub |
-
2018
- 2018-11-14 CN CN201811352681.2A patent/CN109614977A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279763A (en) * | 2013-05-25 | 2013-09-04 | 中北大学 | Method for automatically identifying wheel hub types based on structural features |
| CN103425969A (en) * | 2013-08-07 | 2013-12-04 | 华南理工大学 | Detection system and detection method for identifying type of wheel hub |
| CN105548185A (en) * | 2016-01-08 | 2016-05-04 | 浙江科技学院 | Automobile wheel hub screw hole recognition method based on machine vision and shielding method and system |
| CN106372644A (en) * | 2016-08-22 | 2017-02-01 | 保定市立中车轮制造有限公司 | Image identification method used for wheel hub sorting system |
| CN106643566A (en) * | 2016-09-30 | 2017-05-10 | 华南理工大学 | Method of automatically measuring initial angle of spokes of wheel hub |
Non-Patent Citations (3)
| Title |
|---|
| 任克强等: "基于改进SURF算子的彩色图像配准算法", 《电子测量与仪器学报》 * |
| 杨光等: "轮型代码自动识别系统图像处理算法改进研究 ", 《核电子学与探测技术》 * |
| 谈绍熙等: "一种在铸件缺陷识别中的区域分形分割方法 ", 《中国图象图形学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110530863A (en) * | 2019-08-27 | 2019-12-03 | 南京末梢信息技术有限公司 | A kind of automotive hub mixes package detection device and method |
| CN110530863B (en) * | 2019-08-27 | 2022-04-05 | 南京末梢信息技术有限公司 | Automobile hub mixed package detection device and method |
| CN111238370A (en) * | 2020-02-20 | 2020-06-05 | 中国科学院声学研究所东海研究站 | Intelligent detection method and device for KIT board |
| CN111507404A (en) * | 2020-04-17 | 2020-08-07 | 无锡雪浪数制科技有限公司 | Hub model identification method based on deep vision |
| CN112132783A (en) * | 2020-08-21 | 2020-12-25 | 成都飞机工业（集团）有限责任公司 | Part identification method based on digital image processing technology |
| CN116205923A (en) * | 2023-05-06 | 2023-06-02 | 威海锐鑫丰金属科技有限公司 | Nondestructive testing method for internal defects of automobile hub based on X-RAY |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109614977A (en) | A kind of hub type recognition methods | |
| CN111325713B (en) | Neural network-based wood defect detection method, system and storage medium | |
| CN111815630B (en) | Defect detection method and device for LCD screen | |
| CN109816644B (en) | Bearing defect automatic detection system based on multi-angle light source image | |
| CN108765465B (en) | Unsupervised SAR image change detection method | |
| CN109636824B (en) | Multi-target counting method based on image recognition technology | |
| Türkyılmaz et al. | License plate recognition system using artificial neural networks | |
| JP5706647B2 (en) | Information processing apparatus and processing method thereof | |
| CN101256632B (en) | Information processing apparatus and method | |
| CN105184225B (en) | A kind of multinational banknote image recognition methods and device | |
| CN109858438B (en) | Lane line detection method based on model fitting | |
| CN107844737B (en) | Iris image detection method and device | |
| CN109102004A (en) | Cotton-plant pest-insects method for identifying and classifying and device | |
| CN111798409A (en) | Deep learning-based PCB defect data generation method | |
| CN109285147B (en) | Image processing method and device for breast molybdenum target calcification detection and server | |
| CN116109637B (en) | System and method for detecting appearance defects of turbocharger impeller based on vision | |
| CN114022479A (en) | Battery tab appearance defect detection method | |
| Warif et al. | CMF-iteMS: An automatic threshold selection for detection of copy-move forgery | |
| CN111814825B (en) | Apple detection grading method and system based on genetic algorithm optimization support vector machine | |
| CN115880248A (en) | Surface scratch defect identification method and visual detection equipment | |
| CN114359591A (en) | Self-adaptive image matching algorithm with edge features fused | |
| CN111161295A (en) | Background stripping method for dish image | |
| CN111814852A (en) | Image detection method, image detection device, electronic equipment and computer-readable storage medium | |
| CN114004813A (en) | Identification method and device applied to clinical target area of cervical cancer radiotherapy | |
| CN112396580A (en) | Circular part 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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190412 |
|
| RJ01 | Rejection of invention patent application after publication |