CN110544249A - Convolutional neural network quality identification method for arbitrary-angle case assembly visual inspection - Google Patents
Convolutional neural network quality identification method for arbitrary-angle case assembly visual inspection Download PDFInfo
- Publication number
- CN110544249A CN110544249A CN201910840509.XA CN201910840509A CN110544249A CN 110544249 A CN110544249 A CN 110544249A CN 201910840509 A CN201910840509 A CN 201910840509A CN 110544249 A CN110544249 A CN 110544249A
- Authority
- CN
- China
- Prior art keywords
- neural network
- convolutional neural
- angle
- case assembly
- assembly
- 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
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 17
- 238000011179 visual inspection Methods 0.000 title claims abstract description 15
- 238000001514 detection method Methods 0.000 claims abstract description 19
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims abstract description 6
- 230000015556 catabolic process Effects 0.000 claims abstract description 6
- 238000006731 degradation reaction Methods 0.000 claims abstract description 6
- 238000009826 distribution Methods 0.000 claims abstract description 4
- 239000000284 extract Substances 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 238000007689 inspection Methods 0.000 claims description 4
- 238000011176 pooling Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 2
- 238000012850 discrimination method Methods 0.000 claims 3
- 230000000007 visual effect Effects 0.000 abstract description 9
- 230000009286 beneficial effect Effects 0.000 abstract description 6
- 238000012512 characterization method Methods 0.000 abstract description 5
- 238000007635 classification algorithm Methods 0.000 abstract description 5
- 238000000605 extraction Methods 0.000 abstract description 5
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
Classifications
-
- 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
- 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
- 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
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
Abstract
The invention provides a convolutional neural network quality identification method for any-angle case assembly visual inspection, which comprises the following steps: constructing a distance solving model of the assembly parts of the case; building a regularization layer based on prior distribution, and introducing a residual block to solve the problem of model degradation; setting a multi-section learning rate to obtain a convolutional neural network model with an optimal gradient descent value; and (4) combining the case assembly characteristic information extracted by the convolutional neural network, and finishing the case assembly quality identification at any angle through a case assembly detection model. The invention utilizes the advantages of the convolutional neural network feature extraction and characterization to solve the problem of insufficient pertinence of the classical image classification algorithm to complex images and specific classification modes, sets a multi-section learning rate to improve the convergence rate and the classification accuracy of the convolutional neural network, and is beneficial to application in visual detection of case assembly quality identification at any angle.
Description
Technical Field
The invention relates to the field of assembly quality identification, in particular to a convolutional neural network quality identification method for assembly visual inspection of a case with any angle.
background
The visual detection technology is widely applied due to high accuracy, non-contact and good applicability. As the assembly parts of the chassis have the characteristics of multiple parts, various types and the like, the traditional manual detection method is difficult to finish the detection task with high efficiency and high quality. The quality of the assembly quality of the case directly influences the use of products, and the quality of the case is very necessary to be detected and analyzed. The features extracted by the visual detection method based on the traditional image classification are all visual features of the bottom layer of the image, the classification precision is low, and the classic image classification method cannot achieve a good effect in a complex scene.
in recent years, with the development of visual inspection technology and the great improvement of computing power, the deep network has been applied and developed on the visual inspection task, and the convolutional neural network is applied to the quality inspection of the manufactured products and achieves remarkable effect. The convolutional neural network extracts the characteristics of the image layer by layer through multilayer convolutional operation, obtains higher-order statistical data, and then realizes multi-classification of the image through a classifier to complete a visual detection task. It can be seen that the application effect of the manual quality detection and the classical image classification method in visual detection of the quality of manufactured products is poor, and meanwhile, the technology for manufacturing the quality of the products based on the convolutional neural network is a future trend in the field. If the convolutional neural network feature extraction and characterization advantages can be utilized, the problem that the classical image classification algorithm is insufficient in pertinence to complex images and specific classification modes is solved, and the model is solved based on the distance of the assembly parts of the case with any angle, so that the method is beneficial to application in visual detection of assembly quality identification of the case with any angle.
Disclosure of Invention
in order to solve the problems and the defects, the invention provides a convolutional neural network quality identification method for any-angle case assembly visual inspection, which utilizes the advantages of convolutional neural network feature extraction and characterization to solve the problem of insufficient pertinence of a classical image classification algorithm to complex images and specific classification modes, sets a multi-section learning rate to improve the convergence rate and the classification accuracy of a convolutional neural network, and is beneficial to application in visual inspection for any-angle case assembly quality identification.
the purpose of the invention is realized by the following technical scheme:
a convolutional neural network quality identification method for any-angle case assembly visual inspection comprises the following steps:
A, constructing a case assembly part distance solving model with any angle;
B, constructing a regularization layer based on prior distribution, and introducing a residual block to solve the problem of model degradation;
C, setting a multi-section learning rate to obtain a convolutional neural network model with an optimal gradient descent value;
and D, combining the case assembly characteristic information extracted by the convolutional neural network, and completing the case assembly quality identification at any angle through a case assembly detection model.
The invention has the beneficial effects that:
The method solves the problem that the classical image classification algorithm is insufficient in pertinence to complex images and specific classification modes by utilizing the characteristic extraction and characterization advantages of the convolutional neural network, improves the convergence rate and classification accuracy of the convolutional neural network by setting a multi-section learning rate, and is beneficial to application in visual detection of case assembly quality identification at any angle.
Drawings
FIG. 1 is a flow chart of a convolutional neural network quality identification method for any-angle case assembly visual inspection.
Detailed Description
the present invention will be described in further detail with reference to the following examples and accompanying drawings.
The invention relates to a convolutional neural network quality identification method for any-angle case assembly visual inspection, which comprises the following steps of:
step 10, constructing a distance solving model of the assembly parts of the case with any angle:
When any angle is solved, the actual position of the industrial camera shooting case is mapped to a shooting plane, the rotation angle is defined as theta, namely, the value of any angle is as follows:
In the formula, j is a j-th case, and the sum represents the coordinates of the upper left corner and the lower right corner of the case in the image during the front shooting and the actual shooting respectively.
Obtaining the real assembly space distance between each part and the case based on the size invariance, wherein the formula of the real assembly space distance between each part and the case is as follows:
in the formula, i is a type I part, dj is a type j chassis size, and c is a part positioning point.
step 20, building a regularization layer based on prior distribution, introducing a residual block, and solving the degradation problem:
introducing a regularization layer into a depth convolution neural network optimization target function H to obtain better feature selection; for the two features v1, v2, the regularizing objective function is:
In the formula, the value is a constant, n is the number of samples, and ω is a hyper-parameter, and is used for controlling the degree of regularization.
Introducing a residual block into a network layer to solve the degradation problem caused by model depth; for a certain network layer input H (x, vi), the layer output is:
Y＝H(x,v)+x
step 30, setting a multi-section learning rate to obtain a convolutional neural network model with a gradient descent optimal value:
setting a multi-stage learning rate, and selecting different learning rates after a certain number of iterations to obtain a gradient descent optimal value; the convolutional neural network composed of convolutional layers, pooling layers, full-link layers and the like extracts the features of the parts, and the extracted information of each layer is input into a classifier to complete the classification of the parts. The convolutional neural network of the present embodiment is composed of 50 layers of residual modules stacked: the 1 st convolution layer adopts 7 multiplied by 7 convolution stacking and is connected with the 3 multiplied by 3 pooling layer; the middle consecutive stack is connected by a 1 × 1 convolution with a 3 × 3 convolution and then with the residual module of the 1 × 1 convolution; and finally, inputting the data into an average pooling layer, inputting the data into a full-connection layer, and outputting 1000 classes. The learning rate before the 120k iteration count is set to 0.001, and the learning rate after the 120k iteration count is set to 0.00001.
step 40, combining the case assembly characteristic information extracted by the convolutional neural network, and completing case assembly quality identification at any angle through a case assembly detection model:
the convolutional neural network model determines the network structure in step 20, sets the hyper-parameters in step 30, and extracts the case assembly characteristic information. Step 10 completes the assembly quality identification of the case with any angle according to the steps 20 and 30. The assembly detection model of the case with any angle is as follows:
In the formula, ri is an assembly quality identification result of the i-type parts, and rj is an assembly quality identification result of the j-type chassis. In the embodiment, the convolutional neural network identification information is input into the case assembly detection model, and the final detection result is the case assembly quality identification result at any angle.
the method utilizes the characteristic extraction and characterization advantages of the convolutional neural network, solves the problem that the classical image classification algorithm is insufficient in pertinence to complex images and specific classification modes, sets a multi-section learning rate to improve the convergence rate and classification accuracy of the convolutional neural network, and is beneficial to application in visual detection of case assembly quality identification at any angle.
Although the embodiments of the present invention have been described above, the above descriptions are only for the convenience of understanding the present invention, and are not intended to limit the present invention. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (7)
1. A convolutional neural network quality identification method for any-angle case assembly visual inspection is characterized by comprising the following steps:
A, constructing a case assembly part distance solving model with any angle;
B, constructing a regularization layer based on prior distribution, and introducing a residual block to solve the problem of model degradation;
c, setting a multi-section learning rate to obtain a convolutional neural network model with an optimal gradient descent value;
and D, combining the case assembly characteristic information extracted by the convolutional neural network, and completing the case assembly quality identification at any angle through a case assembly detection model.
2. The convolutional neural network quality discrimination method for any-angle cabinet assembly vision inspection as claimed in claim 1, wherein in the step a, the actual position of the industrial camera shooting cabinet is mapped to the shooting plane when solving any angle, the rotation angle is defined as θ, that is, the value of any angle is:
In the formula, j is a j-th case, and the sum represents the coordinates of the upper left corner and the lower right corner of the case in the image during the front shooting and the actual shooting respectively.
3. the convolutional neural network quality discrimination method for any-angle chassis mount vision inspection as claimed in claim 1, wherein in step a, the distance between each component and the real chassis mount space is obtained based on the dimensional invariance, and the distance between each component and the real chassis mount space is calculated according to the following formula:
In the formula, i is a type I part, dj is a type j chassis size, and c is a part positioning point.
4. The method for identifying the quality of the convolutional neural network for any-angle chassis assembly visual inspection as claimed in claim 1, wherein in the step B, a regularization layer is introduced to a deep convolutional neural network optimization objective function H to obtain a more optimal feature selection; for the two features v1, v2, the regularizing objective function is:
in the formula, the value is a constant, n is the number of samples, and ω is a hyper-parameter, and is used for controlling the degree of regularization.
5. the method for identifying the quality of the convolutional neural network for any-angle chassis assembly visual inspection as claimed in claim 1, wherein in the step B, a residual block is introduced into the network layer to solve the degradation problem caused by model depth; for a certain network layer input H (x, vi), the layer output is:
Y＝H(x,v)+x。
6. the convolutional neural network quality discrimination method for any-angle cabinet assembly vision inspection as claimed in claim 1, wherein in the step C, a multi-stage learning rate is set, different learning rates are selected after a certain number of iterations, and a gradient descent optimal value is obtained; and a convolutional neural network consisting of a convolutional layer, a pooling layer and a full-connection layer is used for extracting the feature of the part, and the extracted information of each layer is input into a classifier to complete the classification of the part.
7. the method for identifying the quality of the convolutional neural network for any-angle case assembly visual inspection as claimed in claim 1, wherein in the step D, the convolutional neural network model extracts case assembly characteristic information by determining a network structure and setting hyper-parameters; the case assembly detection model in the case assembly quality identification at any angle is as follows:
in the formula, ri is an assembly quality identification result of the i-type parts, and rj is an assembly quality identification result of the j-type chassis.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910840509.XA CN110544249A (en) | 2019-09-06 | 2019-09-06 | Convolutional neural network quality identification method for arbitrary-angle case assembly visual inspection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910840509.XA CN110544249A (en) | 2019-09-06 | 2019-09-06 | Convolutional neural network quality identification method for arbitrary-angle case assembly visual inspection |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110544249A true CN110544249A (en) | 2019-12-06 |
Family
ID=68712632
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910840509.XA Pending CN110544249A (en) | 2019-09-06 | 2019-09-06 | Convolutional neural network quality identification method for arbitrary-angle case assembly visual inspection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110544249A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110889461A (en) * | 2019-12-09 | 2020-03-17 | 华南理工大学 | Image multi-feature extraction and classification method for chassis assembly quality detection |
| CN110991540A (en) * | 2019-12-09 | 2020-04-10 | 华南理工大学 | Lightweight image classification method for quick detection of case assembly quality |
| CN113049600A (en) * | 2021-03-31 | 2021-06-29 | 上汽通用五菱汽车股份有限公司 | Engine part wrong and neglected loading detection method and system based on visual detection |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106874889A (en) * | 2017-03-14 | 2017-06-20 | 西安电子科技大学 | Multiple features fusion SAR target discrimination methods based on convolutional neural networks |
| CN108009525A (en) * | 2017-12-25 | 2018-05-08 | 北京航空航天大学 | A kind of specific objective recognition methods over the ground of the unmanned plane based on convolutional neural networks |
| CN109376792A (en) * | 2018-11-07 | 2019-02-22 | 河北工业大学 | Photovoltaic cell open defect classification method based on multichannel residual error neural network |
| CN109637522A (en) * | 2018-12-26 | 2019-04-16 | 杭州电子科技大学 | A kind of speech-emotion recognition method extracting deep space attention characteristics based on sound spectrograph |
| CN110097521A (en) * | 2019-05-08 | 2019-08-06 | 华南理工大学 | A kind of convolutional neural networks image recovery method towards reflecting metal vision-based detection |
| CN110110719A (en) * | 2019-03-27 | 2019-08-09 | 浙江工业大学 | A kind of object detection method based on attention layer region convolutional neural networks |
-
2019
- 2019-09-06 CN CN201910840509.XA patent/CN110544249A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106874889A (en) * | 2017-03-14 | 2017-06-20 | 西安电子科技大学 | Multiple features fusion SAR target discrimination methods based on convolutional neural networks |
| CN108009525A (en) * | 2017-12-25 | 2018-05-08 | 北京航空航天大学 | A kind of specific objective recognition methods over the ground of the unmanned plane based on convolutional neural networks |
| CN109376792A (en) * | 2018-11-07 | 2019-02-22 | 河北工业大学 | Photovoltaic cell open defect classification method based on multichannel residual error neural network |
| CN109637522A (en) * | 2018-12-26 | 2019-04-16 | 杭州电子科技大学 | A kind of speech-emotion recognition method extracting deep space attention characteristics based on sound spectrograph |
| CN110110719A (en) * | 2019-03-27 | 2019-08-09 | 浙江工业大学 | A kind of object detection method based on attention layer region convolutional neural networks |
| CN110097521A (en) * | 2019-05-08 | 2019-08-06 | 华南理工大学 | A kind of convolutional neural networks image recovery method towards reflecting metal vision-based detection |
Non-Patent Citations (1)
| Title |
|---|
| GUIXIONG LIU ET,AL.: "Chassis Assembly Detection and Identification Based on Deep Learning Component Instance Segmentation", 《SYMMETRY-BASEL》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110889461A (en) * | 2019-12-09 | 2020-03-17 | 华南理工大学 | Image multi-feature extraction and classification method for chassis assembly quality detection |
| CN110991540A (en) * | 2019-12-09 | 2020-04-10 | 华南理工大学 | Lightweight image classification method for quick detection of case assembly quality |
| CN110889461B (en) * | 2019-12-09 | 2023-04-28 | 华南理工大学 | Image multi-feature extraction and classification method for chassis assembly quality detection |
| CN110991540B (en) * | 2019-12-09 | 2023-04-28 | 华南理工大学 | Lightweight image classification method for quick detection of chassis assembly quality |
| CN113049600A (en) * | 2021-03-31 | 2021-06-29 | 上汽通用五菱汽车股份有限公司 | Engine part wrong and neglected loading detection method and system based on visual detection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109410321B (en) | Three-dimensional reconstruction method based on convolutional neural network | |
| CN110363122B (en) | Cross-domain target detection method based on multi-layer feature alignment | |
| Tian et al. | HyNet: Learning local descriptor with hybrid similarity measure and triplet loss | |
| CN104599275B (en) | The RGB-D scene understanding methods of imparametrization based on probability graph model | |
| CN111489358A (en) | Three-dimensional point cloud semantic segmentation method based on deep learning | |
| CN108334881B (en) | License plate recognition method based on deep learning | |
| CN109684922B (en) | Multi-model finished dish identification method based on convolutional neural network | |
| CN111062915A (en) | Real-time steel pipe defect detection method based on improved YOLOv3 model | |
| CN108090472B (en) | Pedestrian re-identification method and system based on multi-channel consistency characteristics | |
| CN110544249A (en) | Convolutional neural network quality identification method for arbitrary-angle case assembly visual inspection | |
| CN107808129A (en) | A kind of facial multi-characteristic points localization method based on single convolutional neural networks | |
| CN109766873B (en) | Pedestrian re-identification method based on hybrid deformable convolution | |
| CN111160249A (en) | Multi-class target detection method of optical remote sensing image based on cross-scale feature fusion | |
| CN104268602A (en) | Shielded workpiece identifying method and device based on binary system feature matching | |
| CN110135460B (en) | Image information enhancement method based on VLAD convolution module | |
| CN111242026B (en) | Remote sensing image target detection method based on spatial hierarchy perception module and metric learning | |
| CN109685045A (en) | A kind of Moving Targets Based on Video Streams tracking and system | |
| CN108154158B (en) | Building image segmentation method for augmented reality application | |
| CN107704867A (en) | Based on the image characteristic point error hiding elimination method for weighing the factor in a kind of vision positioning | |
| CN109920050B (en) | Single-view three-dimensional flame reconstruction method based on deep learning and thin plate spline | |
| CN111768415A (en) | Image instance segmentation method without quantization pooling | |
| CN108615401B (en) | Deep learning-based indoor non-uniform light parking space condition identification method | |
| CN109919246A (en) | Pedestrian's recognition methods again based on self-adaptive features cluster and multiple risks fusion | |
| CN110827304A (en) | Traditional Chinese medicine tongue image positioning method and system based on deep convolutional network and level set method | |
| Bappy et al. | Real estate image classification |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191206 |