CN110738648A - camera shell paint spraying detection system and method based on multilayer convolutional neural network - Google Patents
camera shell paint spraying detection system and method based on multilayer convolutional neural network Download PDFInfo
- Publication number
- CN110738648A CN110738648A CN201910968138.3A CN201910968138A CN110738648A CN 110738648 A CN110738648 A CN 110738648A CN 201910968138 A CN201910968138 A CN 201910968138A CN 110738648 A CN110738648 A CN 110738648A
- Authority
- CN
- China
- Prior art keywords
- layer
- module
- model
- training
- neural network
- 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
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared 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]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a camera shell paint spraying detection system based on a multilayer convolutional neural network, which belongs to the field of neural network image processing and comprises a data collection module, an image judgment module, a notification display module, a preprocessing module, a model construction module and a model training module, wherein the preprocessing module processes images of products on a production line collected by the data collection module into a data set, the model training module inputs the data set into a constructed image recognition model, after a mature model is obtained through gradient descent training, the image judgment module inputs the data set into the mature image recognition model, and after image judgment, results are displayed and output through a display module.
Description
Technical Field
The invention discloses camera shell paint spraying detection system and method based on a multilayer convolutional neural network, and relates to the technical field of neural network image processing.
Background
In the traditional production line, the appearance of a product is mostly manually checked to check whether the product is qualified or not in the flow processing operation, the unqualified product is manually treated and then checked again, a large amount of labor and time are consumed in the process of manual inspection, the production efficiency of the production line is seriously inconsistent, if the process of the manual inspection of the product is qualified or not, the manual inspection is replaced by a detection model based on a convolutional neural network, the production efficiency is improved, the production quality is improved, and the intelligent inspection of the product of the production line is further realized.
Disclosure of Invention
The invention provides camera shell paint spraying detection systems based on multilayer convolutional neural network aiming at the problems of the prior art, and adopts the technical scheme that camera shell paint spraying detection systems based on multilayer convolutional neural network comprise a data collection module, a model construction module, a model training module, an image judgment module and a display notification module;
a data collection module: collecting image data of paint spraying of a shell of a collecting camera;
th layer is composed of convolution layer, LEAKY _ RELU activation function and pooling layer, the second layer is composed of convolution layer, LEAKY _ RELU activation function and pooling layer, the third layer is composed of same convolution layer and LEAKY _ RELU activation function, the fourth layer is composed of same convolution layer and LEAKY _ RELU activation function, the fifth layer is composed of same convolution layer and LEAKY _ RELU activation function, the sixth layer is composed of full-link layer composed of neurons, the seventh layer is composed of full-link layer composed of neurons, and the eighth layer is composed of full-link layer composed of neurons;
a model training module: transmitting the image data collected by the data collection module into an image recognition model constructed by a model construction module for recognition training to obtain a mature model of the image recognition model;
an image determination module: transmitting the image data into a mature model for image judgment;
a display notification module: and displaying and outputting the judgment result.
The data collection module processes the collected image data into a data set through the preprocessing module and transmits the data set to the image recognition model.
The preprocessing module performs preprocessing of noise reduction, binning and graying on the image data.
And the model training module performs model training according to a leaving method.
The model training module uses a leave-out method to divide the data set into a training set and a test set, the training set is used for training, and a mature model is selected according to the effect on the test set by using gradient descent training.
camera shell paint spraying detection method based on multilayer convolutional neural network, which is characterized by comprising the following steps:
s1, collecting image data of camera shell paint spraying;
s2 an image recognition model is constructed by a convolutional neural network, wherein the layer consists of a convolutional layer, an LEAKY _ RELU activation function and a pooling layer, the second layer consists of a convolutional layer, an LEAKY _ RELU activation function and a pooling layer, the third layer consists of a same convolutional layer and an LEAKY _ RELU activation function, the fourth layer consists of a same convolutional layer and an LEAKY _ RELU activation function, the fifth layer consists of a same convolutional layer and an LEAKY _ RELU activation function, the sixth layer consists of a full-link layer consisting of neurons, the seventh layer consists of a full-link layer consisting of neurons, and the eighth layer consists of a full-link layer consisting of neurons;
s3, inputting the image data collected in S1 into the image recognition model in S2 for recognition training to obtain a mature model;
s4, inputting the image data collected in S1 into the mature model in S3 for operation to obtain an image judgment result;
s5 displays and outputs the determination result.
And the image data collected in the S1 is preprocessed into a data set and then transmitted into the image recognition model.
The data set is pre-processed for noise reduction, binning and graying.
The S3 image recognition model is model trained by the leave-out method.
The S3 leave-out method divides the data set into a training set and a test set, the test set is used for training, and a maturity model is selected according to the effect on the test set by using gradient descent training
camera shell paint spraying detection systems based on multilayer convolutional neural network, the preprocessing module processes the camera paint spraying pictures collected by the data collection module into a data set, the image construction module utilizes the convolutional neural network to construct an image recognition model, the image training module inputs the data set into the constructed image recognition model, after a mature model is obtained by gradient descent training, the mature model is judged by the image judgment module, and the result is displayed and output by the display module;
the convolution layer of the image recognition model of the five-layer convolution neural network and the three-layer full-connection network has the characteristics of weight sharing, feature extraction and local connection, can play the roles of extracting image features, reducing parameter quantity and facilitating the training of the image recognition model, and simultaneously uses LEAKY _ RELU as an activation function, compared with the traditional RELU activation function, the training process can be accelerated, the gradient disappearance can be avoided, the pooling layer has a down-sampling function, the function of extracting the features of the image data for classification can be realized, and the full-connection layer can perform fusion classification on high-dimensional feature maps obtained by convolution to obtain a final classification result; compared with the traditional network models such as VGG16, the number of convolutional layers is less, so that the parameters are less, and fitting is easier during training.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, is briefly introduced in the drawings required in the description of the embodiments or the prior art, it is obvious that the drawings in the following description are embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without inventive exercise.
FIG. 1 is a schematic diagram of the system of the present invention; FIG. 2 is a schematic representation of the steps of the process of the present invention.
Detailed Description
The present invention is further illustrated in conjunction with the accompanying drawings and the specific embodiments so that those skilled in the art can better understand the present invention and can practice it, but the embodiments are not to be construed as limiting the present invention.
Example :
camera shell paint spraying detection system based on multilayer convolution neural network, including data collection module, preprocessing module, model construction module, model training module, image judgment module and notification display module;
the data collection module is used for acquiring images of the spray painting condition of ten thousand camera shells on the production line from the USB infrared camera by using opencv, screening and then transmitting the qualified images into the preprocessing module;
the preprocessing module is used for carrying out noise reduction, classification and graying processing on the pictures to form a data set;
the model construction module constructs an image recognition model by using a tenserflow construction multilayer convolutional neural network:
layer consists of convolution layers with convolution kernel 11x11x96, step size 4, LEAKY _ RELU activation function and 3x3 pooling layers with step size 2,
the second layer consists of convolution layer with convolution kernel of 5x5x256 and step size of 2, LEAKY _ RELU activation function and 3x3 pooling layer with step size of 2,
the third layer consists of a same convolution layer with a convolution kernel of 3x3x384 and a LEAKY _ RELU activation function,
the fourth layer consists of a same convolution layer with a convolution kernel of 3x3x384 and a LEAKY _ RELU activation function,
the fifth layer is composed of a same convolution layer with a convolution kernel of 3x3x256 and a LEAKY _ RELU activation function,
the sixth layer consists of a fully connected layer of 2048 neurons,
the seventh layer consists of a fully connected layer of 512 neurons,
the eighth layer consists of a fully connected layer of 2 neurons;
a model training module: the data set is transmitted into an image recognition model constructed by a model construction module for recognition training to obtain a mature model of the image recognition model;
an image determination module: transmitting the data set into a mature model for image judgment;
a display notification module: displaying and outputting the judgment result;
the convolutional layer of the image recognition model of the five-layer convolutional neural network and the three-layer fully-connected network has the three characteristics of weight sharing, feature extraction and local connection, the functions of extracting features from a data set, reducing the number of parameters and facilitating training are achieved, the LEAKY _ RELU activation function can accelerate the training process and avoid gradient disappearance compared with the traditional RELU activation function, the pooling layer has a down-sampling function and achieves the purpose of classifying the image extraction features in the data set, and the fully-connected layer can perform fusion classification on high-dimensional feature maps obtained by convolution to obtain a final classification result.
Example two:
on the basis of the embodiment , the model training module divides a data set into a training set and a test set according to a leave-out method of 4 and 1, the test set is used for training, a practical gradient descent training selects a mature model according to the effect on the test set, and the model training module uses a dropout technology to avoid overfitting, gradient disappearance and gradient explosion and improve the stability in the system.
Example three:
camera shell paint spraying detection method based on multilayer convolutional neural network, comprising the following steps:
s1, acquiring images of the painting condition of the shell of ten thousand cameras on a production line from the USB infrared camera by using opencv, and performing noise reduction, treatment and graying treatment to form a data set after primary screening;
s2, constructing a multilayer convolutional neural network by tenserflow to construct an image recognition model:
layer consists of convolution layers with convolution kernel 11x11x96, step size 4, LEAKY _ RELU activation function and 3x3 pooling layers with step size 2,
the second layer consists of convolution layer with convolution kernel of 5x5x256 and step size of 2, LEAKY _ RELU activation function and 3x3 pooling layer with step size of 2,
the third layer consists of a same convolution layer with a convolution kernel of 3x3x384 and a LEAKY _ RELU activation function,
the fourth layer consists of a same convolution layer with a convolution kernel of 3x3x384 and a LEAKY _ RELU activation function,
the fifth layer is composed of a same convolution layer with a convolution kernel of 3x3x256 and a LEAKY _ RELU activation function,
the sixth layer consists of a fully connected layer of 2048 neurons,
the seventh layer consists of a fully connected layer of 512 neurons,
the eighth layer consists of a fully connected layer of 2 neurons;
s3, inputting the data set of S1 into the image recognition model in S2 for recognition training to obtain a mature model;
s4, inputting the data set of S1 into the mature model of S3 for operation to obtain an image judgment result;
s5 displaying and outputting the judgment result;
the convolutional layer of the image recognition model of the five-layer convolutional neural network and the three-layer fully-connected network has the three characteristics of weight sharing, feature extraction and local connection, the functions of extracting features from a data set, reducing the number of parameters and facilitating training are achieved, the LEAKY _ RELU activation function can accelerate the training process and avoid gradient disappearance compared with the traditional RELU activation function, the pooling layer has a down-sampling function and achieves the purpose of classifying the image extraction features in the data set, and the fully-connected layer can perform fusion classification on high-dimensional feature maps obtained by convolution to obtain a final classification result.
Example four:
on the basis of the third embodiment, the S3 divides the data set into a training set and a test set according to a leave-out method and a 4:1 method, wherein the test set is used for training, and the practical gradient descent training selects a mature model according to the effect on the test set; s3 avoids overfitting, gradient disappearance and gradient explosion by using a dropout technology, and improves the internal stability of the method.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
- The system is characterized in that the detection system comprises a data collection module, a model construction module, a model training module, an image judgment module and a display notification module;a data collection module: collecting image data of paint spraying of a shell of a collecting camera;th layer is composed of convolution layer, LEAKY _ RELU activation function and pooling layer, the second layer is composed of convolution layer, LEAKY _ RELU activation function and pooling layer, the third layer is composed of same convolution layer and LEAKY _ RELU activation function, the fourth layer is composed of same convolution layer and LEAKY _ RELU activation function, the fifth layer is composed of same convolution layer and LEAKY _ RELU activation function, the sixth layer is composed of full-link layer composed of neurons, the seventh layer is composed of full-link layer composed of neurons, and the eighth layer is composed of full-link layer composed of neurons;a model training module: transmitting the image data collected by the data collection module into an image recognition model constructed by a model construction module for recognition training to obtain a mature model of the image recognition model;an image determination module: transmitting the image data into a mature model for image judgment;a display notification module: and displaying and outputting the judgment result.
- 2. The camera housing painting detection system based on the multilayer convolutional neural network as claimed in claim 1, further comprising a preprocessing module, wherein the data collection module processes the acquired image data into a data set through the preprocessing module, and transmits the data set to the image recognition model.
- 3. The multi-layer convolutional neural network-based camera housing painting detection system of claim 2, wherein the preprocessing module performs preprocessing of noise reduction, classification and graying on the image data.
- 4. The system of , wherein the model training module performs model training according to the leave-out method.
- 5. The multi-layer convolutional neural network-based camera housing painting detection system of claim 5, wherein the model training module uses a leave-out method to divide the data set into a training set and a test set, the training set is used for training, and a gradient descent training is used to select a mature model according to the effect on the test set.
- 6, camera shell paint spraying detection method based on multilayer convolution neural network, which is characterized by comprising the following steps:s1, collecting image data of camera shell paint spraying;s2 an image recognition model is constructed by a convolutional neural network, wherein the layer consists of a convolutional layer, an LEAKY _ RELU activation function and a pooling layer, the second layer consists of a convolutional layer, an LEAKY _ RELU activation function and a pooling layer, the third layer consists of a same convolutional layer and an LEAKY _ RELU activation function, the fourth layer consists of a same convolutional layer and an LEAKY _ RELU activation function, the fifth layer consists of a same convolutional layer and an LEAKY _ RELU activation function, the sixth layer consists of a full-link layer consisting of neurons, the seventh layer consists of a full-link layer consisting of neurons, and the eighth layer consists of a full-link layer consisting of neurons;s3, inputting the image data collected in S1 into the image recognition model in S2 for recognition training to obtain a mature model;s4, inputting the image data collected in S1 into the mature model in S3 for operation to obtain an image judgment result;s5 displays and outputs the determination result.
- 7. The method for detecting paint spraying on camera casings based on the multilayer convolutional neural network as claimed in claim 6, wherein the image data collected in S1 is preprocessed into a data set and then transmitted into an image recognition model.
- 8. The method for detecting paint spraying of camera housings based on multilayer convolutional neural network as claimed in claim 7, wherein the data set is preprocessed by denoising, classifying and graying.
- 9. The method for detecting paint spraying on camera housings based on multilayer convolutional neural network of any of claims 6-8, wherein the S3 image recognition model is model-trained by the leave-out method.
- 10. The method for detecting paint spraying on camera housings based on multilayer convolutional neural network of claim 9, wherein the S3 leave method divides the data set into a training set and a test set, the test set is used for training, and the gradient descent training is used to select a mature model according to the effect on the test set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910968138.3A CN110738648A (en) | 2019-10-12 | 2019-10-12 | camera shell paint spraying detection system and method based on multilayer convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910968138.3A CN110738648A (en) | 2019-10-12 | 2019-10-12 | camera shell paint spraying detection system and method based on multilayer convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110738648A true CN110738648A (en) | 2020-01-31 |
Family
ID=69268748
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910968138.3A Pending CN110738648A (en) | 2019-10-12 | 2019-10-12 | camera shell paint spraying detection system and method based on multilayer convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110738648A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017166155A1 (en) * | 2016-03-31 | 2017-10-05 | 富士通株式会社 | Method and device for training neural network model, and electronic device |
| CN107563999A (en) * | 2017-09-05 | 2018-01-09 | 华中科技大学 | A kind of chip defect recognition methods based on convolutional neural networks |
| CN109359608A (en) * | 2018-10-25 | 2019-02-19 | 电子科技大学 | A kind of face identification method based on deep learning model |
| CN109543632A (en) * | 2018-11-28 | 2019-03-29 | 太原理工大学 | A kind of deep layer network pedestrian detection method based on the guidance of shallow-layer Fusion Features |
| CN110100774A (en) * | 2019-05-08 | 2019-08-09 | 安徽大学 | River crab male and female recognition methods based on convolutional neural networks |
| CN110248191A (en) * | 2019-07-15 | 2019-09-17 | 山东浪潮人工智能研究院有限公司 | A kind of video-frequency compression method based on deep layer convolutional neural networks |
-
2019
- 2019-10-12 CN CN201910968138.3A patent/CN110738648A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017166155A1 (en) * | 2016-03-31 | 2017-10-05 | 富士通株式会社 | Method and device for training neural network model, and electronic device |
| CN107563999A (en) * | 2017-09-05 | 2018-01-09 | 华中科技大学 | A kind of chip defect recognition methods based on convolutional neural networks |
| CN109359608A (en) * | 2018-10-25 | 2019-02-19 | 电子科技大学 | A kind of face identification method based on deep learning model |
| CN109543632A (en) * | 2018-11-28 | 2019-03-29 | 太原理工大学 | A kind of deep layer network pedestrian detection method based on the guidance of shallow-layer Fusion Features |
| CN110100774A (en) * | 2019-05-08 | 2019-08-09 | 安徽大学 | River crab male and female recognition methods based on convolutional neural networks |
| CN110248191A (en) * | 2019-07-15 | 2019-09-17 | 山东浪潮人工智能研究院有限公司 | A kind of video-frequency compression method based on deep layer convolutional neural networks |
Non-Patent Citations (1)
| Title |
|---|
| 陈龙现等: "基于CNN的木材内部CT图像缺陷辨识", 《林业科学》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107463949B (en) | Video action classification processing method and device | |
| CN107679522B (en) | Multi-stream LSTM-based action identification method | |
| CN110059741B (en) | Image recognition method based on semantic capsule fusion network | |
| CN107784654A (en) | Image partition method, device and full convolutional network system | |
| CN108615010A (en) | Facial expression recognizing method based on the fusion of parallel convolutional neural networks characteristic pattern | |
| CN108875821A (en) | The training method and device of disaggregated model, mobile terminal, readable storage medium storing program for executing | |
| CN107944442A (en) | Based on the object test equipment and method for improving convolutional neural networks | |
| CN104063686B (en) | Crop leaf diseases image interactive diagnostic system and method | |
| CN109035142A (en) | A kind of satellite image ultra-resolution method fighting network integration Aerial Images priori | |
| CN113627504B (en) | Multi-mode multi-scale feature fusion target detection method based on generation of countermeasure network | |
| CN108960288B (en) | Three-dimensional model classification method and system based on convolutional neural network | |
| CN110288597A (en) | Wireless capsule endoscope saliency detection method based on attention mechanism | |
| CN108665005A (en) | A method of it is improved based on CNN image recognition performances using DCGAN | |
| CN108875906B (en) | A kind of multiple dimensioned convolutional neural networks learning method gradually to add up | |
| CN109377487B (en) | Fruit surface defect detection method based on deep learning segmentation | |
| CN109801232A (en) | A kind of single image to the fog method based on deep learning | |
| CN111401116B (en) | Bimodal emotion recognition method based on enhanced convolution and space-time LSTM network | |
| CN111582095A (en) | Light-weight rapid detection method for abnormal behaviors of pedestrians | |
| CN104408470A (en) | Gender detection method based on average face preliminary learning | |
| CN111597933A (en) | Face recognition method and device | |
| CN109060838A (en) | A kind of product surface scratch detection method based on machine vision | |
| CN116893175A (en) | Multi-station multi-defect unified connector measurement system based on deep learning | |
| CN112101094A (en) | Suicide risk assessment method based on body language | |
| Sun et al. | Multi‐vision attention networks for on‐line red jujube grading | |
| CN107729885A (en) | A kind of face Enhancement Method based on the study of multiple residual error |
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: 20200131 |