CN112529893A - Hub surface flaw online detection method and system based on deep neural network - Google Patents
Hub surface flaw online detection method and system based on deep neural network Download PDFInfo
- Publication number
- CN112529893A CN112529893A CN202011523385.1A CN202011523385A CN112529893A CN 112529893 A CN112529893 A CN 112529893A CN 202011523385 A CN202011523385 A CN 202011523385A CN 112529893 A CN112529893 A CN 112529893A
- Authority
- CN
- China
- Prior art keywords
- hub
- neural network
- image
- deep neural
- hub surface
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 50
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 49
- 230000007547 defect Effects 0.000 claims abstract description 60
- 238000000034 method Methods 0.000 claims abstract description 27
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 238000012549 training Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 8
- 238000005286 illumination Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 5
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000005316 response function Methods 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 238000000605 extraction Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 6
- 238000007689 inspection Methods 0.000 description 6
- 238000005266 casting Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000005484 gravity Effects 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 238000003860 storage Methods 0.000 description 4
- 229910000838 Al alloy Inorganic materials 0.000 description 3
- 239000004576 sand Substances 0.000 description 3
- 229910052782 aluminium Inorganic materials 0.000 description 2
- XAGFODPZIPBFFR-UHFFFAOYSA-N aluminium Chemical compound [Al] XAGFODPZIPBFFR-UHFFFAOYSA-N 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000005520 cutting process Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 229910000831 Steel Inorganic materials 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000005242 forging Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 201000001371 inclusion conjunctivitis Diseases 0.000 description 1
- 238000005495 investment casting Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 229910052751 metal Inorganic materials 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005498 polishing Methods 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 206010044325 trachoma Diseases 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Images
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field of hub flaw detection, and particularly relates to a hub surface flaw online detection method and system based on a deep neural network, wherein the method comprises the following steps: acquiring a surface image of the hub by block shooting, wherein each shooting is performed in consideration of adjacent block areas of the hub; preprocessing the wheel hub surface image data by using a filter, and extracting image characteristics of the filtered data; and classifying and identifying the image characteristics by using the trained and optimized flaw grade prediction model to obtain the grade information of the current hub defects. According to the method, the mask image of the flaws on the surface of the hub and the corresponding grade information of the flaws are obtained through the strong feature extraction capability of the deep neural network, so that the labor intensity and the workload are reduced, the efficiency is improved, and the method has a good application value.
Description
Technical Field
The invention belongs to the technical field of hub flaw detection, and particularly relates to a hub surface flaw online detection method and system based on a deep neural network.
Background
With the continuous improvement of the economic level of China, the automobile holding amount is larger and larger. As is known, a hub is one of the essential components of an automobile, and is a rotating part of a wheel core in which the inner profile of a tire is connected by a pillar, i.e., a metal part supporting the center of the tire and mounted on a shaft. Also called rim, steel ring, wheel and tyre bell. The mainstream material of the hub in the current market is aluminum alloy, and the manufacturing methods of the aluminum alloy hub comprise three methods: gravity casting, forging and low-pressure precision casting. The most adopted method is a gravity casting method, wherein the gravity casting method is to pour an aluminum alloy solution into a mold by utilizing gravity, and finish production by processing and polishing through a lathe after forming. The manufacturing process is simple, a precise casting process is not needed, the cost is low, the production efficiency is high, but air bubbles (sand holes) are easily generated, the density is uneven, and the surface smoothness is not enough. In addition, because the hardness of the aluminum product is reduced, collision and scratches are easy to occur in the casting and transportation processes, and the condition that the surface of the hub is polluted by oil stains often occurs. The defects of trachoma, bump and the like in the existing detection mainly depend on manual selection. However, the manual quality inspection is inefficient and limited by the physiological characteristics of human, and the accuracy of quality inspection is very low. Particularly, under the condition that the quality inspection of the hub is strict, the hub is small and slightly collided, and can not be clearly observed by naked eyes, so that the inspection omission is caused. In addition, the resolution of a single camera is limited, and global detail features cannot be presented at a single time.
Disclosure of Invention
Therefore, the invention provides the hub surface flaw online detection method and system based on the deep neural network, the mask image of the hub surface flaw and the corresponding grade information of the flaw are obtained through the strong feature extraction capability of the deep neural network, the labor intensity and the workload are reduced, and the efficiency is improved.
According to the design scheme provided by the invention, the hub surface flaw online detection method based on the deep neural network comprises the following contents:
acquiring a surface image of the hub by block shooting, wherein each shooting is performed in consideration of adjacent block areas of the hub;
preprocessing the wheel hub surface image data by using a filter, and extracting image characteristics of the filtered data;
and classifying and identifying the image characteristics by using the trained and optimized flaw grade prediction model to obtain the grade information of the current hub defects.
As the hub surface flaw online detection method based on the deep neural network, the RGB camera for shooting is fixed by the mechanical arm of the detection table, and the RGB camera is driven by the mechanical arm to the corresponding shooting point to acquire the hub surface image.
As the hub surface flaw online detection method based on the deep neural network, a fan-shaped light source for shooting images with uniform illumination is further arranged at the front end of the mechanical arm; and in the acquisition of the same hub surface image, each shooting is carried out under the same image resolution.
As the hub surface flaw online detection method based on the deep neural network, further, the filter adopts a high-pass filter for suppressing low-frequency information, and the filter adopted in the preprocessing process is represented as:
wherein the content of the first and second substances,
as an impulse response function, D0The cutoff frequency is shown, and mu and nu respectively represent horizontal and vertical coordinates on the spectrogram.
As the hub surface flaw online detection method based on the deep neural network, disclosed by the invention, further, the image characteristics after filtering are extracted by using the trained and optimized deep neural network; the trained and optimized deep neural network comprises: the device comprises a defect encoder for extracting color image characteristics of input data, a defect decoder for decoding the color image characteristics, and a unit module for acquiring a hub defect mask.
As the hub surface flaw on-line detection method based on the deep neural network, the deep neural network is trained and optimized by adopting a gradient descent method, and the maximum iteration times or network loss limit is set as an iteration termination condition.
As the hub surface flaw online detection method based on the deep neural network, further, the loss function of the deep neural network is expressed as:
wherein i is an index corresponding to the training set,
to predict value, yiFor the true value, N is the number of samples used for training optimization.
As the hub surface flaw online detection method based on the deep neural network, further, the flaw grade prediction model comprises the following steps: the hub defect detection system comprises a defect grade information encoder used for extracting defect information characteristics, and a full-connection layer used for acquiring current hub defect grade information through full-connection operation.
As the hub surface flaw online detection method based on the deep neural network, further, a flaw grade prediction model adopts a cross loss function to carry out training optimization, wherein the cross loss function is expressed as:
wherein i is the index corresponding to the training set, yiTo predict value, tiIs the target value.
Further, the present invention also provides an online wheel hub surface flaw detection system based on a deep neural network, comprising: an image acquisition module, an image processing module and an image identification module, wherein,
the image acquisition module is used for acquiring the surface image of the hub by block shooting, and the adjacent block areas of the hub are taken into consideration for each shooting;
the image processing module is used for preprocessing the wheel hub surface image data by using a filter and extracting the image characteristics of the filtered data;
and the image identification module is used for carrying out classification identification on the image characteristics by utilizing the trained and optimized flaw grade prediction model to obtain the grade information of the current hub defects.
The invention has the beneficial effects that:
aiming at the problems of low efficiency and accuracy of the existing manual quality inspection and the like, the method utilizes the strong learning representation capability of the deep neural network as a prediction model for detecting the defects of the hub, has good effect and strong robustness, has loose requirements on environmental illumination and is greatly superior to the traditional machine learning scheme; further, on the premise of a common camera, the mechanical arm is adopted to drive the light source and the camera to act, the image acquisition mode is optimized by taking a picture at a fixed point position, meanwhile, the RGB information of the surface of the hub acquired by the color camera is utilized, the grade information of the defect area is taken into consideration, the grade information of the surface defect of the hub is obtained through prediction through the powerful characteristic extraction capability of the deep neural network, the detection accuracy is high, and the method has a good application prospect.
Description of the drawings:
FIG. 1 is a schematic flow chart of an online detection method for hub surface flaws in the embodiment;
FIG. 2 is a schematic diagram of the working principle of the online detection network model in the embodiment;
FIG. 3 is a schematic view of a hub to be inspected in the example;
FIG. 4 is a defect indication of the hub in the embodiment;
FIG. 5 is an enlarged schematic view of the hub of the embodiment showing the bump and scratch;
FIG. 6 is a schematic diagram of the detection platform in the embodiment.
The specific implementation mode is as follows:
in order to make the objects, technical solutions and advantages of the present invention clearer and more obvious, the present invention is further described in detail below with reference to the accompanying drawings and technical solutions.
The embodiment of the invention, as shown in fig. 1, provides an online detection method for hub surface flaws based on a deep neural network, which comprises the following steps:
s101, acquiring a hub surface image through block shooting, wherein each shooting is performed in consideration of adjacent block areas of the hub;
s102, preprocessing the wheel hub surface image data by using a filter, and extracting image characteristics of the filtered data;
s103, classifying and identifying the image characteristics by using the trained and optimized flaw grade prediction model, and acquiring the grade information of the current hub defects.
The method aims at the problems that the efficiency and the accuracy rate of the existing manual quality inspection are low and the like, the powerful learning representation capability of the deep neural network is used as a prediction model for detecting the defects of the hub, the effect is good, the robustness is high, the detection error of the defects of the hub is reduced, the detection accuracy is improved, the requirement on environment illumination is loose, and the method is greatly superior to the traditional machine learning scheme.
As the hub surface flaw online detection method based on the deep neural network in the embodiment of the invention, further, an RGB camera for shooting is fixed by using a mechanical arm of the detection table, and the RGB camera is driven by the mechanical arm to a corresponding shooting point to acquire a hub surface image.
Because the defect regions such as the sand holes and the scratches on the surface of the hub are small, if a traditional shooting mode is adopted, the surface image of the whole hub is obtained, and the imaging effect of the image cannot show the detailed characteristics of the defects such as the sand holes and the scratches. Therefore, the embodiment of the scheme can adopt a mode of taking pictures for multiple times, each time of taking pictures can take part of the hub area, and the whole hub is processed in a blocking mode, so that the proportion of the defect area in the whole image is increased under the same image resolution. In addition, in order to guarantee the shooting consistency, the RGB camera can be further driven to a fixed point position through the mechanical arm, and the shot is triggered, so that RGB images on the surface of the hub are obtained.
As the hub surface flaw online detection method based on the deep neural network in the embodiment of the invention, further, a fan-shaped light source for shooting images with uniform illumination is arranged at the front end of the mechanical arm; and in the acquisition of the same hub surface image, each shooting is carried out under the same image resolution.
Referring to the detection platform structure shown in fig. 6, the hub can be fixed on the detection platform, and the camera is fixed at the front end of the mechanical arm. Among them, the robot arm may be a six-axis freely rotatable robot arm. In addition, in order to more clearly show the defect details, a customized fan-shaped light source is fixed at the front end of the mechanical arm, wherein the fan-shaped angle is 120 degrees. It should be noted that the plane of the camera needs to be adjusted to be parallel to the plane where the fan-shaped light source is located, so that the illumination of the acquired image is relatively uniform. The mechanical arm can be set to five fixed points so as to cover the complete hub surface. The image acquisition of two adjacent areas needs to have a part of overlapping area, so as to prevent the situation that a complete defect is not shot in one picture under certain conditions.
As an online detection method for hub surface flaws based on a deep neural network in the embodiment of the present invention, further, a high-pass filter for suppressing low-frequency information is adopted as the filter, and the filter adopted in the preprocessing process is represented as:
wherein the content of the first and second substances,
as an impulse response function, D0The representation represents the cut-off frequency, and μ and ν represent the horizontal and vertical coordinates on the spectrogram, respectively.
After the five RGB images are acquired, preprocessing operation is firstly carried out on the RGB images, in order to highlight collision and scratch information, the RGB images are high-frequency information under the common condition, low-frequency information can be eliminated through the filtered images, the high-frequency information is reserved, and partial interference factors are eliminated.
As the hub surface flaw online detection method based on the deep neural network in the embodiment of the invention, further, the image features after filtering are extracted by using the trained and optimized deep neural network; the trained and optimized deep neural network comprises: defect encoder for extracting color image features of input data, defect decoder for decoding color image featuresAnd a unit module for acquiring the wheel hub defect mask. Further, a gradient descent method is adopted to train and optimize the deep neural network, and the maximum iteration times or network loss limit is set as an iteration termination condition. Further, the deep neural network loss function is expressed as:
wherein i is an index corresponding to the training set,
to predict value, yiFor the true value, N is the number of samples used for training optimization.
Referring to fig. 2, firstly, a defect mask prediction branch processes RGB images, and a defect Encoder1 is used to extract the features of color images, so as to obtain FeatureMap 1. And further decoding the feature information FeatureMap1 by using a Decoder1 to finally obtain the flaw mask information Segmentation. The network optimization training can be stopped and the parameter model can be saved by setting when the network is iterated 50000 times or the network loss is less than 0.00001. The scratches and the bumps on the surface of the hub are more, the area of the area is smaller, and the pixel level segmentation may cause more false detection. Therefore, in the embodiment of the invention, after the surface defect of the hub is obtained, a defect area small graph is obtained by a post-processing method, and the severity of the scratch and the bump is predicted by adopting the area small graph.
As an online detection method for hub surface flaws based on a deep neural network in an embodiment of the present invention, further, a flaw level prediction model includes: the hub defect detection system comprises a defect grade information encoder used for extracting defect information characteristics, and a full-connection layer used for acquiring current hub defect grade information through full-connection operation. Further, the flaw level prediction model is trained and optimized by adopting a cross loss function, wherein the cross loss function is expressed as:
wherein i is the index corresponding to the training set, yiTo predict value, tiIs the target value.
Referring to fig. 2, a defect level information Encoder encorder 2 is used for extracting defect information features to obtain FeatureMap2, and finally, a full connection layer FC is used for flattening FeatureMap2 to finally obtain level information of the current hub defect.
Further, based on the above method, an embodiment of the present invention further provides an online wheel hub surface flaw detection system based on a deep neural network, including: an image acquisition module, an image processing module and an image identification module, wherein,
the image acquisition module is used for acquiring the surface image of the hub by block shooting, and the adjacent block areas of the hub are taken into consideration for each shooting;
the image processing module is used for preprocessing the wheel hub surface image data by using a filter and extracting the image characteristics of the filtered data;
and the image identification module is used for carrying out classification identification on the image characteristics by utilizing the trained and optimized flaw grade prediction model to obtain the grade information of the current hub defects.
To verify the validity of the scheme, the following further explanation is made by combining specific experiments:
referring to fig. 2 to 6, the hub is fixed on the detection platform, the detection platform is composed of an aluminum frame structure, and the camera is fixed at the front end of the mechanical arm. Wherein, the mechanical arm adopts six-axis mechanical arm capable of freely rotating. An industrial ccd and 500-ten-thousand-pixel color camera can be adopted, a fan-shaped customized light source is adopted, the maximum brightness is 300 nits, and the maximum brightness can be adjusted through a light source controller. In order to show the defect details more clearly, a customized fan-shaped light source is fixed at the front end of the mechanical arm, wherein the fan-shaped angle is 120 degrees. It should be noted that the plane of the camera needs to be adjusted to be parallel to the plane where the fan-shaped light source is located, so that the illumination of the acquired image is relatively uniform. The mechanical arm is provided with five fixed points, so that the complete hub surface is covered. Five fixed point positions need to be adjusted according to the shape of the hub. The defect region mask information may be represented by an index 0 and an index 1, respectively, where the index 0 represents background information and the index 1 represents defect information. And finding the defect outline by adopting an outline searching method according to the mask image, further determining the maximum external rectangle of the defect area, and cutting the defect small image from the original image by adopting an image cutting method. And predicting the severity of the scratches and the bumps by using the cut small images. Specifically, defect information characteristics are extracted by a defect level information Encoder Encoder2 to obtain FeatureMap2, and finally, FeatureMap2 is flattened by a full connection layer FC to finally obtain the level information of the current hub defect. 10. Aiming at the training process of the flaw grade prediction module, the network inputs the characteristic information of the color image and outputs the predicted flaw grade information. The invention divides the defect grade into 3 types according to the removing standard, and the index 0 represents no flaw; the index 1 represents that the product has defects, but the defects are slight, and the product is qualified without being removed; index 2 represents severe defects, which need to be removed and the product is unqualified. The implementer can adjust the rejection standard in time according to the requirement.
Unless specifically stated otherwise, the relative steps, numerical expressions, and values of the components and steps set forth in these embodiments do not limit the scope of the present invention.
Based on the foregoing method or system, an embodiment of the present invention further provides a network device, including: one or more processors; a storage device for storing one or more programs which, when executed by the one or more processors, cause the one or more processors to implement the system or perform the method described above.
Based on the above system, the embodiment of the present invention further provides a computer readable medium, on which a computer program is stored, wherein the program, when executed by a processor, implements the above system.
The device provided by the embodiment of the present invention has the same implementation principle and technical effect as the system embodiment, and for the sake of brief description, reference may be made to the corresponding content in the system embodiment for the part where the device embodiment is not mentioned.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the system and the apparatus described above may refer to the corresponding processes in the foregoing system embodiments, and are not described herein again.
In all examples shown and described herein, any particular value should be construed as merely exemplary, and not as a limitation, and thus other examples of example embodiments may have different values.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the system according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present invention, which are used for illustrating the technical solutions of the present invention and not for limiting the same, and the protection scope of the present invention is not limited thereto, although the present invention is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the embodiments of the present invention, and they should be construed as being included therein. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A hub surface flaw online detection method based on a deep neural network is characterized by comprising the following steps:
acquiring a surface image of the hub by block shooting, wherein each shooting is performed in consideration of adjacent block areas of the hub;
preprocessing the wheel hub surface image data by using a filter, and extracting image characteristics of the filtered data;
and classifying and identifying the image characteristics by using the trained and optimized flaw grade prediction model to obtain the grade information of the current hub defects.
2. The in-line detection method for the hub surface flaws based on the deep neural network as claimed in claim 1, wherein a mechanical arm of the detection table is used to fix an RGB camera for shooting, and the mechanical arm drives the RGB camera to a corresponding shooting point to acquire a hub surface image.
3. The hub surface flaw online detection method based on the deep neural network as claimed in claim 2, wherein a fan-shaped light source for shooting an image with uniform illumination is arranged at the front end of the mechanical arm; and in the acquisition of the same hub surface image, each shooting is carried out under the same image resolution.
4. The method for detecting the hub surface flaws on the basis of the deep neural network as claimed in claim 1, wherein the filter adopts a high-pass filter for suppressing low-frequency information, and the filter adopted in the preprocessing process is represented as:
wherein the content of the first and second substances,
as an impulse response function, D0The cutoff frequency is shown, and mu and nu respectively represent horizontal and vertical coordinates on the spectrogram.
5. The hub surface flaw online detection method based on the deep neural network is characterized in that the trained and optimized deep neural network is used for extracting the filtered image features; the trained and optimized deep neural network comprises: the device comprises a defect encoder for extracting color image characteristics of input data, a defect decoder for decoding the color image characteristics, and a unit module for acquiring a hub defect mask.
6. The method for detecting the hub surface flaws on the basis of the deep neural network as claimed in claim 5, wherein a gradient descent method is adopted to train and optimize the deep neural network, and a maximum iteration number or a network loss limit is set as an iteration termination condition.
7. The hub surface flaw online detection method based on the deep neural network as claimed in claim 5 or 6, wherein the deep neural network loss function is expressed as:
wherein i is an index corresponding to the training set,
to predict value, yiFor the true value, N is the number of samples used for training optimization.
8. The hub surface flaw online detection method based on the deep neural network as claimed in claim 1, wherein the flaw grade prediction model comprises: the hub defect detection system comprises a defect grade information encoder used for extracting defect information characteristics, and a full-connection layer used for acquiring current hub defect grade information through full-connection operation.
9. The hub surface flaw online detection method based on the deep neural network as claimed in claim 1 or 8, wherein the flaw grade prediction model is trained and optimized by using a cross loss function, wherein the cross loss function is expressed as:
wherein i is the index corresponding to the training set, yiTo predict value, tiIs the target value.
10. A hub surface flaw online detection system based on a deep neural network is characterized by comprising: an image acquisition module, an image processing module and an image identification module, wherein,
the image acquisition module is used for acquiring the surface image of the hub by block shooting, and the adjacent block areas of the hub are taken into consideration for each shooting;
the image processing module is used for preprocessing the wheel hub surface image data by using a filter and extracting the image characteristics of the filtered data;
and the image identification module is used for carrying out classification identification on the image characteristics by utilizing the trained and optimized flaw grade prediction model to obtain the grade information of the current hub defects.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011523385.1A CN112529893A (en) | 2020-12-22 | 2020-12-22 | Hub surface flaw online detection method and system based on deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011523385.1A CN112529893A (en) | 2020-12-22 | 2020-12-22 | Hub surface flaw online detection method and system based on deep neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112529893A true CN112529893A (en) | 2021-03-19 |
Family
ID=75002439
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011523385.1A Pending CN112529893A (en) | 2020-12-22 | 2020-12-22 | Hub surface flaw online detection method and system based on deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112529893A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113470018A (en) * | 2021-09-01 | 2021-10-01 | 深圳市信润富联数字科技有限公司 | Hub defect identification method, electronic device, device and readable storage medium |
| CN114332081A (en) * | 2022-03-07 | 2022-04-12 | 泗水县亿佳纺织厂 | Textile surface abnormity determination method based on image processing |
| CN117491391A (en) * | 2023-12-29 | 2024-02-02 | 登景(天津)科技有限公司 | Glass substrate light three-dimensional health detection method and equipment based on chip calculation |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105118044A (en) * | 2015-06-16 | 2015-12-02 | 华南理工大学 | Method for automatically detecting defects of wheel-shaped cast product |
| CN111398292A (en) * | 2020-04-07 | 2020-07-10 | 苏州哈工吉乐优智能装备科技有限公司 | Gabor filtering and CNN-based cloth defect detection method, system and equipment |
| CN111414500A (en) * | 2020-05-08 | 2020-07-14 | 刘如意 | Steel wire rope breakage early warning system based on block chain and BIM |
| CN111724367A (en) * | 2020-06-16 | 2020-09-29 | 哈尔滨全感科技有限公司 | Glass panel degumming identification method based on image method |
| CN111899224A (en) * | 2020-06-30 | 2020-11-06 | 烟台市计量所 | Nuclear power pipeline defect detection system based on deep learning attention mechanism |
| CN111929314A (en) * | 2020-08-26 | 2020-11-13 | 湖北汽车工业学院 | Wheel hub weld visual detection method and detection system |
-
2020
- 2020-12-22 CN CN202011523385.1A patent/CN112529893A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105118044A (en) * | 2015-06-16 | 2015-12-02 | 华南理工大学 | Method for automatically detecting defects of wheel-shaped cast product |
| CN111398292A (en) * | 2020-04-07 | 2020-07-10 | 苏州哈工吉乐优智能装备科技有限公司 | Gabor filtering and CNN-based cloth defect detection method, system and equipment |
| CN111414500A (en) * | 2020-05-08 | 2020-07-14 | 刘如意 | Steel wire rope breakage early warning system based on block chain and BIM |
| CN111724367A (en) * | 2020-06-16 | 2020-09-29 | 哈尔滨全感科技有限公司 | Glass panel degumming identification method based on image method |
| CN111899224A (en) * | 2020-06-30 | 2020-11-06 | 烟台市计量所 | Nuclear power pipeline defect detection system based on deep learning attention mechanism |
| CN111929314A (en) * | 2020-08-26 | 2020-11-13 | 湖北汽车工业学院 | Wheel hub weld visual detection method and detection system |
Non-Patent Citations (2)
| Title |
|---|
| 李念芦等: "《电影制作技术手册》", 31 December 2017 * |
| 黄德双等, 合肥：中国科学技术大学出版社 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113470018A (en) * | 2021-09-01 | 2021-10-01 | 深圳市信润富联数字科技有限公司 | Hub defect identification method, electronic device, device and readable storage medium |
| CN114332081A (en) * | 2022-03-07 | 2022-04-12 | 泗水县亿佳纺织厂 | Textile surface abnormity determination method based on image processing |
| CN117491391A (en) * | 2023-12-29 | 2024-02-02 | 登景(天津)科技有限公司 | Glass substrate light three-dimensional health detection method and equipment based on chip calculation |
| CN117491391B (en) * | 2023-12-29 | 2024-03-15 | 登景(天津)科技有限公司 | Glass substrate light three-dimensional health detection method and equipment based on chip calculation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113450307B (en) | Product edge defect detection method | |
| CN112529893A (en) | Hub surface flaw online detection method and system based on deep neural network | |
| CN114723681B (en) | Concrete crack defect detection method based on machine vision | |
| CN107228860B (en) | Gear defect detection method based on image rotation period characteristics | |
| CN109785245B (en) | Light spot image trimming method | |
| CN107369136B (en) | Visual detection method for surface cracks of polycrystalline diamond compact | |
| CN107330373A (en) | A kind of parking offense monitoring system based on video | |
| CN113920090B (en) | Prefabricated rod appearance defect automatic detection method based on deep learning | |
| CN116777916B (en) | Defect detection method based on metal shell of pump machine | |
| CN111046862B (en) | Character segmentation method, device and computer readable storage medium | |
| CN112330598B (en) | Method, device and storage medium for detecting stiff yarn defects on chemical fiber surface | |
| CN109101976B (en) | Method for detecting surface defects of arc-extinguishing grid plate | |
| CN116777907A (en) | Sheet metal part quality detection method | |
| WO2017120796A1 (en) | Pavement distress detection method and apparatus, and electronic device | |
| WO2021000948A1 (en) | Counterweight weight detection method and system, and acquisition method and system, and crane | |
| CN116109637B (en) | System and method for detecting appearance defects of turbocharger impeller based on vision | |
| CN111489337A (en) | Method and system for removing false defects through automatic optical detection | |
| CN108802051B (en) | System and method for detecting bubble and crease defects of linear circuit of flexible IC substrate | |
| CN111047556A (en) | Strip steel surface defect detection method and device | |
| CN115409785A (en) | Method for detecting defects of small pluggable transceiver module base | |
| CN109558877B (en) | KCF-based offshore target tracking algorithm | |
| CN113298775B (en) | Self-priming pump double-sided metal impeller appearance defect detection method, system and medium | |
| CN113591507A (en) | Robust two-dimensional code DataMatrix positioning method and system | |
| CN112614113A (en) | Strip steel defect detection method based on deep learning | |
| CN115797314B (en) | Method, system, equipment and storage medium for detecting surface defects of parts |
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: 20210319 |
|
| RJ01 | Rejection of invention patent application after publication |