CN112525919A - Wood board defect detection system and method based on deep learning - Google Patents
Wood board defect detection system and method based on deep learning Download PDFInfo
- Publication number
- CN112525919A CN112525919A CN202011517629.5A CN202011517629A CN112525919A CN 112525919 A CN112525919 A CN 112525919A CN 202011517629 A CN202011517629 A CN 202011517629A CN 112525919 A CN112525919 A CN 112525919A
- Authority
- CN
- China
- Prior art keywords
- board
- defect
- image
- wood board
- defect detection
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 132
- 238000001514 detection method Methods 0.000 title claims abstract description 97
- 239000002023 wood Substances 0.000 title claims abstract description 95
- 238000013135 deep learning Methods 0.000 title claims abstract description 24
- 238000000034 method Methods 0.000 title claims abstract description 18
- 238000012549 training Methods 0.000 claims description 23
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 239000011248 coating agent Substances 0.000 claims description 7
- 238000000576 coating method Methods 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 7
- 238000012795 verification Methods 0.000 claims description 6
- 230000005540 biological transmission Effects 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 description 5
- 230000000007 visual effect Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 239000002699 waste material Substances 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 210000000887 face Anatomy 0.000 description 1
- 239000003292 glue Substances 0.000 description 1
- 210000003128 head Anatomy 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/89—Investigating the presence of flaws or contamination in moving material, e.g. running paper or textiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- 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
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4038—Image mosaicing, e.g. composing plane images from plane sub-images
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2200/00—Indexing scheme for image data processing or generation, in general
- G06T2200/32—Indexing scheme for image data processing or generation, in general involving image mosaicing
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biophysics (AREA)
- Signal Processing (AREA)
- Biomedical Technology (AREA)
- Geometry (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Textile Engineering (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention provides a board defect detection system and method based on deep learning, belonging to the technical field of board defect detection, wherein the system comprises: an industrial personal computer; the display is connected with the industrial personal computer; the conveying belt is connected with the industrial personal computer; the bottom of the dark box is provided with an opening with the same size as the projection area; the camera bellows is arranged above the conveyor belt; the LED light source is arranged on the side wall in the dark box, and the irradiation direction is upward; one end of the light source controller is connected with the industrial personal computer, and the other end of the light source controller is connected with the LED light source; the camera is arranged at the top end inside the camera bellows, and the shooting direction is downward; one end of the image acquisition card is connected with the industrial personal computer, and the other end of the image acquisition card is connected with the camera; and the laser sensor is arranged at the upper end of the side edge of the conveyor belt, is connected with the industrial personal computer and is used for detecting whether a wood board passes through. The invention has the advantages that: greatly improve the precision and the efficiency of the wood board defect detection of complex textures.
Description
Technical Field
The invention relates to the technical field of board defect detection, in particular to a board defect detection system and method based on deep learning.
Background
In daily life, wood products are seen everywhere, as the demand of the wood products is continuously increased, the consumption of various kinds of wood is correspondingly and continuously increased, people cannot meet the volume pursuit, strict requirements are also put forward to the quality, wherein the surface defects of the wood boards directly influence the product grade, the waste products can be directly caused seriously, and the great resource waste and the production cost are caused. The defect types of the wood board mainly comprise joint cracks, hole joints, water cracks, glue line cracks and the like.
For the detection of template defects, there are traditionally manual visual detection methods, machine visual detection methods, and computer visual detection methods.
The manual visual detection method is not required to be provided with a special clamp and a special testing device, is simple and convenient to apply, and is long-term applied to the defect detection of the surface of the wood board, but has the following defects: 1. the detection speed is low, the production efficiency is low, and the labor cost is high; 2. the detection precision is low, and fine defects cannot be found by naked eyes; 3. the detection standards are not uniform and are easily influenced by subjective factors of detection personnel; 4. over-judgment and under-judgment are easy to occur.
The machine vision detection method is a typical high and new detection technology integrating light collection, machinery, electricity, gas and the like, the core is a machine vision technology, a machine vision detection system can improve the flexibility and the automation degree of production, and although the defects of the manual visual detection method are overcome to a certain degree, the defects exist in the wood board with complex surface textures as follows: 1. the detection scene is single, and the requirements on illumination background and the like are strict; 2. the detection parameters are many, manual experience is relied on, and the debugging period is long; 3. the detection effect on the surface defects of complex textures is poor.
The computer vision detection method achieves the purpose of understanding the image through high-level semantic features of the abstract image, is relatively complex in application scene, can be used for identifying objects with irregular shapes, low regularity and multiple types, and is difficult to use objective quantity as identification basis for manufacturing wood boards with complex surface textures and high production value.
Therefore, how to provide a board defect detection system and method based on deep learning to improve the precision and efficiency of board defect detection of complex textures becomes a problem to be solved urgently.
Disclosure of Invention
The invention aims to provide a wood board defect detection system and method based on deep learning, and the precision and efficiency of wood board defect detection of complex textures are improved.
In a first aspect, the present invention provides a plank defect detection system based on deep learning, including:
an industrial personal computer;
the display is connected with the industrial personal computer;
the conveying belt is connected with the industrial personal computer;
the bottom of the dark box is provided with an opening with the same size as the projection area; the camera bellows is arranged above the conveyor belt;
the LED light source is arranged on the side wall in the dark box, and the irradiation direction is upward;
one end of the light source controller is connected with the industrial personal computer, and the other end of the light source controller is connected with the LED light source;
the camera is arranged at the top end inside the camera bellows, and the shooting direction is downward;
one end of the image acquisition card is connected with the industrial personal computer, and the other end of the image acquisition card is connected with the camera;
and the laser sensor is arranged at the upper end of the side edge of the conveyor belt, is connected with the industrial personal computer and is used for detecting whether a wood board passes through.
Further, the display is a touch display.
Further, the inner surface of the dark box is provided with a white coating.
Further, the LED light source is a strip-shaped LED light source.
Further, the camera is a line scan industrial camera.
In a second aspect, the invention provides a plank defect detection method based on deep learning, which comprises the following steps:
s10, acquiring the wood board images of the wood boards transmitted on the conveyor belt by the industrial personal computer through the camera;
s20, preprocessing the collected wood board image to obtain a data set;
step S30, establishing a board defect detection model based on the deep neural network, and training the board defect detection model by using the data set;
and S40, acquiring an image of the board to be detected, and inputting the image of the board to be detected into the trained board defect detection model for board defect detection.
Further, the step S10 specifically includes:
s11, setting the brightness of the LED light source through the light source controller by the industrial personal computer, and carrying out distortion calibration on the camera through the image acquisition card;
and S12, starting the conveyor belt by the industrial personal computer, sequentially placing the wood boards on the conveyor belt for transmission, and starting the LED light source and the camera to collect wood board images when the industrial personal computer detects that the wood boards pass through the conveyor belt through the laser sensor.
Further, the step S20 specifically includes:
step S21, marking the defect position and defect type of the collected wood board image; marking the defect position by using a defect frame;
step S22, setting a cutting size and a proportion threshold value, creating a training data set, and cutting each wood board image based on the cutting size to obtain wood board sub-images;
step S23, calculating a first IoU value of the clipped defect image in the original defect image in each wood board sub-image based on the defect frame, judging whether the first IoU value is larger than a proportional threshold value, and if so, adding the corresponding wood board sub-image into a data set; and if not, discarding the corresponding wood board sub-image.
Further, the step S30 specifically includes:
step S31, establishing a board defect detection model based on the deep neural network;
s32, clustering the defect frames into 9 anchor frames by using a clustering algorithm, randomly selecting 9 defect frames as the initial length and width of each anchor frame, aligning the defect frames with the upper left corners of the anchor frames to calculate a second IoU value, classifying the defect frames by using the second IoU value, respectively weighting the lengths and the widths of the defect frames of each category to calculate an average value, and further updating the sizes of the 9 anchor frames;
step S33, increasing the sample size of the data set by using a mosaic data enhancement method;
step S34, dividing the data set into a training set and a verification set according to a preset proportion, and carrying out defect image frame selection training on wood board subimages in the training set by the wood board defect detection model through an anchor frame; updating the network weight of the wood board defect detection model by using a gradient descent method in the training process;
and step S35, verifying the trained wood board defect detection model by using the verification set.
Further, the step S40 is specifically:
acquiring an image of the wood board to be detected, and cutting the image of the wood board to be detected based on the cutting size to obtain a subimage of the wood board to be detected; inputting each subimage of the board to be detected into a trained board defect detection model for detection, performing frame selection on the defect image of each subimage of the board to be detected by using the anchor frame, and splicing the subimages of the board to be detected after frame selection to obtain a defect frame selection image of the board to be detected.
The invention has the advantages that:
the camera is placed in the camera box with the white coating on the inner surface, the LED light source is arranged on the side wall in the camera box, the irradiation direction is upward, light emitted by the LED light source can be uniformly reflected to the wood board through the inner surface of the camera box, the brightness of the LED light source is set before the wood board image is collected, the camera is subjected to distortion calibration, and the quality of the wood board image is greatly improved; a board defect detection model is created based on the deep neural network, a data set generated by a high-quality board image is used for training the board defect detection model, and finally the trained board defect detection model is used for detecting the board defects, so that the precision of board defect detection with complex textures is greatly improved; whether have the plank process through setting up laser sensor detection, have plank process time automatic start LED light source and camera collection to wait to examine the plank image, plank defect detection model after the direct utilization training is again waited to examine the plank image and is carried out plank defect detection, very big promotion plank defect detection's efficiency, finally realize the contactless detection of plank, discover plank surface defect in advance, in time restore plank surface defect, reduce the waste of plank surface UV coating material, reduce the manual work and detect the cost.
Drawings
The invention will be further described with reference to the following examples with reference to the accompanying drawings.
FIG. 1 is a schematic circuit block diagram of a board defect detection system based on deep learning according to the present invention.
FIG. 2 is a schematic structural diagram of a plank defect detection system based on deep learning according to the present invention.
FIG. 3 is a flowchart of a plank defect detection method based on deep learning according to the present invention.
Description of the labeling:
100-a board defect detection system based on deep learning, 1-an industrial personal computer, 2-a display, 3-a conveyor belt, 4-a dark box, 5-an LED light source, 6-a light source controller, 7-a camera, 8-an image acquisition card, 9-a laser sensor and 41-an opening.
Detailed Description
Referring to fig. 1 to 3, a preferred embodiment of a wood board defect detecting system 100 based on deep learning according to the present invention includes:
an industrial personal computer 1 for controlling the detection system 100;
the display 2 is connected with the industrial personal computer 1 and used for displaying the detection result of the defects of the wood board;
the conveyor belt 3 is connected with the industrial personal computer 1 and is used for conveying wood boards (not shown), and the size specification range of the wood boards is 0.5m x0.5m to 1.5m x 1.5m;
the bottom of the dark box 4 is provided with an opening 41 with the same size as the projection area; the camera bellows 4 is arranged above the conveyor belt 3 and used for providing a good shooting environment for the camera 7;
the LED light source 5 is arranged on the side wall inside the dark box 4, the irradiation direction of the LED light source is upward, and the LED light source is used for supplementing light to the wood board;
a light source controller 6, one end of which is connected with the industrial personal computer 1 and the other end of which is connected with the LED light source 5, and is used for adjusting the brightness of the LED light source 5 and controlling a switch;
the camera 7 is arranged at the top end inside the camera bellows 4, faces downwards in shooting direction and is used for collecting wood board images;
one end of the image acquisition card 8 is connected with the industrial personal computer 1, and the other end of the image acquisition card is connected with the camera 7 and used for controlling the camera 7 to acquire images;
and the laser sensor 9 is arranged at the upper end of the side edge of the conveyor belt 3 and connected with the industrial personal computer 1 for detecting whether a wood board passes through.
The display 2 is a touch display.
The inner surface of the dark box 4 is provided with a white coating (not shown) for uniformly reflecting the light emitted by the LED light source 5 to the wood board.
The LED light source 5 is a strip-shaped LED light source and is used for providing uniform light.
The camera 7 is a line scanning industrial camera, and the resolution range of the collected image is wide: 1000-8000, high: 1000 to 8000.
The invention discloses a better embodiment of a wood board defect detection method based on deep learning, which comprises the following steps:
s10, acquiring the wood board images of the wood boards transmitted on the conveyor belt by the industrial personal computer through the camera;
s20, preprocessing the collected wood board image to obtain a data set;
step S30, establishing a board defect detection model based on the deep neural network, and training the board defect detection model by using the data set;
and S40, acquiring an image of the board to be detected, and inputting the image of the board to be detected into the trained board defect detection model for board defect detection.
The step S10 specifically includes:
s11, setting the brightness of the LED light source through the light source controller by the industrial personal computer to enable imaging to be clear, and carrying out distortion calibration on the camera through the image acquisition card;
and S12, starting the conveyor belt by the industrial personal computer, sequentially placing the wood boards on the conveyor belt for transmission, and starting the LED light source and the camera to collect wood board images when the industrial personal computer detects that the wood boards pass through the conveyor belt through the laser sensor.
The step S20 specifically includes:
step S21, marking the defect position and defect type of the collected wood board image; marking the defect position by using a defect frame;
step S22, setting a cutting size and a proportion threshold value, creating a training data set, and cutting each wood board image based on the cutting size to obtain wood board sub-images;
step S23, calculating a first IoU value of the clipped defect image in the original defect image in each wood board sub-image based on the defect frame, judging whether the first IoU value is larger than a proportional threshold value, and if so, adding the corresponding wood board sub-image into a data set; and if not, discarding the corresponding wood board sub-image. I.e. screening the sub-images of the board.
The step S30 specifically includes:
step S31, establishing a board defect detection model based on the deep neural network;
the board defect detection model is built by using a pytorch, a CSPResblock block is used as a main feature extraction network, an SPP space pyramid pooling and FPN feature pyramid network is used as a neck for extracting and integrating different high-level semantic features, a DIoU loss function is used for position regression of a network head, two-classification cross entropy loss functions are used for confidence coefficient and classification, a final loss function is obtained through weighted summation of the three functions, and training acceleration is performed by using a GPU.
S32, clustering the defect frames into 9 anchor frames by using a clustering algorithm, randomly selecting 9 defect frames as the initial length and width of each anchor frame, aligning the defect frames with the upper left corners of the anchor frames to calculate a second IoU value, classifying the defect frames by using the second IoU value, respectively weighting the lengths and the widths of the defect frames of each category to calculate an average value, and further updating the sizes of the 9 anchor frames;
step S33, increasing the sample size of the data set by using a mosaic data enhancement method;
namely, 4 wood sub-images are randomly spliced to obtain a large image, the large image is subjected to random rotation, translation and other transformations, and then the large image is cut to obtain an image with the same size as the wood sub-images, so that the sample size is increased.
Step S34, dividing the data set into a training set and a verification set according to a preset proportion, and carrying out defect image frame selection training on wood board subimages in the training set by the wood board defect detection model through an anchor frame; in the training process, updating the network weight of the wood board defect detection model by using a gradient descent method, and preventing gradient explosion by using technologies such as gradient cutting, learning rate attenuation and batch normalization;
and step S35, verifying the trained wood board defect detection model by using the verification set.
The step S40 specifically includes:
acquiring an image of the wood board to be detected, and cutting the image of the wood board to be detected based on the cutting size to obtain a subimage of the wood board to be detected; inputting each subimage of the board to be detected into a trained board defect detection model for detection, performing frame selection on the defect image of each subimage of the board to be detected by using the anchor frame, and splicing the subimages of the board to be detected after frame selection to obtain a defect frame selection image of the board to be detected. In order to prevent the GPU video memory from overflowing, the blocksize of the model, namely the number of images, needs to be dynamically set according to the resolution of the to-be-detected wood board image.
In summary, the invention has the advantages that:
the camera is placed in the camera box with the white coating on the inner surface, the LED light source is arranged on the side wall in the camera box, the irradiation direction is upward, light emitted by the LED light source can be uniformly reflected to the wood board through the inner surface of the camera box, the brightness of the LED light source is set before the wood board image is collected, the camera is subjected to distortion calibration, and the quality of the wood board image is greatly improved; a board defect detection model is created based on the deep neural network, a data set generated by a high-quality board image is used for training the board defect detection model, and finally the trained board defect detection model is used for detecting the board defects, so that the precision of board defect detection with complex textures is greatly improved; whether have the plank process through setting up laser sensor detection, have plank process time automatic start LED light source and camera collection to wait to examine the plank image, plank defect detection model after the direct utilization training is again waited to examine the plank image and is carried out plank defect detection, very big promotion plank defect detection's efficiency, finally realize the contactless detection of plank, discover plank surface defect in advance, in time restore plank surface defect, reduce the waste of plank surface UV coating material, reduce the manual work and detect the cost.
Although specific embodiments of the invention have been described above, it will be understood by those skilled in the art that the specific embodiments described are illustrative only and are not limiting upon the scope of the invention, and that equivalent modifications and variations can be made by those skilled in the art without departing from the spirit of the invention, which is to be limited only by the appended claims.
Claims (10)
1. The utility model provides a plank defect detecting system based on degree of deep learning which characterized in that: the method comprises the following steps:
an industrial personal computer;
the display is connected with the industrial personal computer;
the conveying belt is connected with the industrial personal computer;
the bottom of the dark box is provided with an opening with the same size as the projection area; the camera bellows is arranged above the conveyor belt;
the LED light source is arranged on the side wall in the dark box, and the irradiation direction is upward;
one end of the light source controller is connected with the industrial personal computer, and the other end of the light source controller is connected with the LED light source;
the camera is arranged at the top end inside the camera bellows, and the shooting direction is downward;
one end of the image acquisition card is connected with the industrial personal computer, and the other end of the image acquisition card is connected with the camera;
and the laser sensor is arranged at the upper end of the side edge of the conveyor belt, is connected with the industrial personal computer and is used for detecting whether a wood board passes through.
2. The board defect detection system based on deep learning of claim 1, wherein: the display is a touch display.
3. The board defect detection system based on deep learning of claim 1, wherein: the inner surface of the dark box is provided with a white coating.
4. The board defect detection system based on deep learning of claim 1, wherein: the LED light source is a strip-shaped LED light source.
5. The board defect detection system based on deep learning of claim 1, wherein: the camera is a line scan industrial camera.
6. A plank defect detection method based on deep learning is characterized in that: the method entails using a detection system according to any one of claims 1 to 5, comprising the steps of:
s10, acquiring the wood board images of the wood boards transmitted on the conveyor belt by the industrial personal computer through the camera;
s20, preprocessing the collected wood board image to obtain a data set;
step S30, establishing a board defect detection model based on the deep neural network, and training the board defect detection model by using the data set;
and S40, acquiring an image of the board to be detected, and inputting the image of the board to be detected into the trained board defect detection model for board defect detection.
7. The wood board defect detection method based on deep learning as claimed in claim 6, wherein: the step S10 specifically includes:
s11, setting the brightness of the LED light source through the light source controller by the industrial personal computer, and carrying out distortion calibration on the camera through the image acquisition card;
and S12, starting the conveyor belt by the industrial personal computer, sequentially placing the wood boards on the conveyor belt for transmission, and starting the LED light source and the camera to collect wood board images when the industrial personal computer detects that the wood boards pass through the conveyor belt through the laser sensor.
8. The wood board defect detection method based on deep learning as claimed in claim 6, wherein: the step S20 specifically includes:
step S21, marking the defect position and defect type of the collected wood board image; marking the defect position by using a defect frame;
step S22, setting a cutting size and a proportion threshold value, creating a training data set, and cutting each wood board image based on the cutting size to obtain wood board sub-images;
step S23, calculating a first IoU value of the clipped defect image in the original defect image in each wood board sub-image based on the defect frame, judging whether the first IoU value is larger than a proportional threshold value, and if so, adding the corresponding wood board sub-image into a data set; and if not, discarding the corresponding wood board sub-image.
9. The wood board defect detection method based on deep learning as claimed in claim 8, wherein: the step S30 specifically includes:
step S31, establishing a board defect detection model based on the deep neural network;
s32, clustering the defect frames into 9 anchor frames by using a clustering algorithm, randomly selecting 9 defect frames as the initial length and width of each anchor frame, aligning the defect frames with the upper left corners of the anchor frames to calculate a second IoU value, classifying the defect frames by using the second IoU value, respectively weighting the lengths and the widths of the defect frames of each category to calculate an average value, and further updating the sizes of the 9 anchor frames;
step S33, increasing the sample size of the data set by using a mosaic data enhancement method;
step S34, dividing the data set into a training set and a verification set according to a preset proportion, and carrying out defect image frame selection training on wood board subimages in the training set by the wood board defect detection model through an anchor frame; updating the network weight of the wood board defect detection model by using a gradient descent method in the training process;
and step S35, verifying the trained wood board defect detection model by using the verification set.
10. The wood board defect detection method based on deep learning as claimed in claim 9, wherein: the step S40 specifically includes:
acquiring an image of the wood board to be detected, and cutting the image of the wood board to be detected based on the cutting size to obtain a subimage of the wood board to be detected; inputting each subimage of the board to be detected into a trained board defect detection model for detection, performing frame selection on the defect image of each subimage of the board to be detected by using the anchor frame, and splicing the subimages of the board to be detected after frame selection to obtain a defect frame selection image of the board to be detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011517629.5A CN112525919A (en) | 2020-12-21 | 2020-12-21 | Wood board defect detection system and method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011517629.5A CN112525919A (en) | 2020-12-21 | 2020-12-21 | Wood board defect detection system and method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112525919A true CN112525919A (en) | 2021-03-19 |
Family
ID=75002005
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011517629.5A Pending CN112525919A (en) | 2020-12-21 | 2020-12-21 | Wood board defect detection system and method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112525919A (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103056111A (en) * | 2012-12-21 | 2013-04-24 | 浙江大学 | Prawns quality detecting and classifying device based on machine vision technology |
| CN205317678U (en) * | 2015-12-08 | 2016-06-15 | 广东锐视智能检测有限公司 | Metal product printing defect on -line measuring system based on machine vision |
| CN106442549A (en) * | 2016-12-03 | 2017-02-22 | 河池学院 | Gear defect detection system |
| CN109711474A (en) * | 2018-12-24 | 2019-05-03 | 中山大学 | A kind of aluminium material surface defects detection algorithm based on deep learning |
| CN109784476A (en) * | 2019-01-12 | 2019-05-21 | 福州大学 | A method of improving DSOD network |
| CN110310259A (en) * | 2019-06-19 | 2019-10-08 | 江南大学 | It is a kind of that flaw detection method is tied based on the wood for improving YOLOv3 algorithm |
| CN110517224A (en) * | 2019-07-12 | 2019-11-29 | 上海大学 | A kind of photovoltaic panel defect inspection method based on deep neural network |
| US20200175352A1 (en) * | 2017-03-14 | 2020-06-04 | University Of Manitoba | Structure defect detection using machine learning algorithms |
| CN210788240U (en) * | 2019-09-19 | 2020-06-19 | 西安工程大学 | Tablet packaging shell defect detection machine based on machine vision |
| CN111351860A (en) * | 2019-11-29 | 2020-06-30 | 浙江农林大学 | Wood internal defect detection method based on Faster R-CNN |
| CN112016614A (en) * | 2020-08-27 | 2020-12-01 | 北京理工大学 | Construction method of optical image target detection model, target detection method and device |
-
2020
- 2020-12-21 CN CN202011517629.5A patent/CN112525919A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103056111A (en) * | 2012-12-21 | 2013-04-24 | 浙江大学 | Prawns quality detecting and classifying device based on machine vision technology |
| CN205317678U (en) * | 2015-12-08 | 2016-06-15 | 广东锐视智能检测有限公司 | Metal product printing defect on -line measuring system based on machine vision |
| CN106442549A (en) * | 2016-12-03 | 2017-02-22 | 河池学院 | Gear defect detection system |
| US20200175352A1 (en) * | 2017-03-14 | 2020-06-04 | University Of Manitoba | Structure defect detection using machine learning algorithms |
| CN109711474A (en) * | 2018-12-24 | 2019-05-03 | 中山大学 | A kind of aluminium material surface defects detection algorithm based on deep learning |
| CN109784476A (en) * | 2019-01-12 | 2019-05-21 | 福州大学 | A method of improving DSOD network |
| CN110310259A (en) * | 2019-06-19 | 2019-10-08 | 江南大学 | It is a kind of that flaw detection method is tied based on the wood for improving YOLOv3 algorithm |
| CN110517224A (en) * | 2019-07-12 | 2019-11-29 | 上海大学 | A kind of photovoltaic panel defect inspection method based on deep neural network |
| CN210788240U (en) * | 2019-09-19 | 2020-06-19 | 西安工程大学 | Tablet packaging shell defect detection machine based on machine vision |
| CN111351860A (en) * | 2019-11-29 | 2020-06-30 | 浙江农林大学 | Wood internal defect detection method based on Faster R-CNN |
| CN112016614A (en) * | 2020-08-27 | 2020-12-01 | 北京理工大学 | Construction method of optical image target detection model, target detection method and device |
Non-Patent Citations (2)
| Title |
|---|
| TING HE等: "A Fully Convolutional Neural Network for Wood Defect Location and Identification" * |
| 岳慧慧等: "基于改进 YOLOv3 的木结缺陷检测方法研究" * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109724984B (en) | Defect detection and identification device and method based on deep learning algorithm | |
| CN110618134A (en) | Steel plate surface quality defect detection and rating system and method | |
| CN109916910B (en) | Photovoltaic glass edge defect detection system and corresponding method | |
| CN106409711A (en) | Solar silicon wafer defect detecting system and method | |
| CN107944504B (en) | Board recognition and machine learning method and device for board recognition and electronic equipment | |
| CN106770332A (en) | A kind of electronic die blank defects detection implementation method based on machine vision | |
| CN107966454A (en) | A kind of end plug defect detecting device and detection method based on FPGA | |
| CN107084991A (en) | The detection of quartz pushrod bubble and quality grading method based on machine vision | |
| CN113155851A (en) | Copper-clad plate surface defect visual online detection method and device based on deep learning | |
| CN104483320A (en) | Digitized defect detection device and detection method of industrial denitration catalyst | |
| CN110554052A (en) | artificial board surface defect detection method and system | |
| CN106780464A (en) | A kind of fabric defect detection method based on improvement Threshold segmentation | |
| CN112116582A (en) | Cigarette detection and identification method under stock or display scene | |
| CN113869300A (en) | Workpiece surface defect and character recognition method and system based on multi-vision fusion | |
| CN114638797A (en) | Method and device for detecting copper surface defects based on linear array camera | |
| CN114119591A (en) | Display screen picture quality detection method | |
| CN210775265U (en) | Artificial board surface defect detection system | |
| CN111239142A (en) | Paste appearance defect detection device and method | |
| CN107423944A (en) | A kind of portable one-stop detection service system and method | |
| CN114881987B (en) | Hot-pressing light guide plate defect visual detection method based on improvement YOLOv5 | |
| CN114359235A (en) | Wood surface defect detection method based on improved YOLOv5l network | |
| CN106093051A (en) | Paper roll tangent plane burr detection method based on machine vision and device | |
| CN109115775A (en) | A kind of betel nut level detection method based on machine vision | |
| CN114972246A (en) | Die-cutting product surface defect detection method based on deep learning | |
| CN113947598A (en) | Plastic lunch box defect detection method, device and system based on image processing |
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 |