CN113344923A - Chemical fiber spindle surface defect detection method and device, electronic equipment and storage medium - Google Patents
Chemical fiber spindle surface defect detection method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN113344923A CN113344923A CN202110893697.XA CN202110893697A CN113344923A CN 113344923 A CN113344923 A CN 113344923A CN 202110893697 A CN202110893697 A CN 202110893697A CN 113344923 A CN113344923 A CN 113344923A
- Authority
- CN
- China
- Prior art keywords
- tripwire
- point
- point sets
- sets
- surface area
- 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.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 53
- 239000000835 fiber Substances 0.000 title claims abstract description 29
- 239000000126 substance Substances 0.000 title claims abstract description 28
- 238000003860 storage Methods 0.000 title claims abstract description 10
- 238000001514 detection method Methods 0.000 title claims description 18
- 230000011218 segmentation Effects 0.000 claims abstract description 41
- 238000000034 method Methods 0.000 claims abstract description 38
- 238000004458 analytical method Methods 0.000 claims abstract description 16
- 238000012216 screening Methods 0.000 claims abstract description 13
- 238000010586 diagram Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 14
- 238000007781 pre-processing Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 4
- 238000004519 manufacturing process Methods 0.000 description 10
- 206010017577 Gait disturbance Diseases 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000000137 annealing Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241000533950 Leucojum Species 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000009987 spinning Methods 0.000 description 1
- 239000011800 void material Substances 0.000 description 1
- 238000004804 winding Methods 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- 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
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20076—Probabilistic image processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30124—Fabrics; Textile; Paper
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a device for detecting surface defects of chemical fiber spindles, electronic equipment and a storage medium. The method comprises the following steps: acquiring a surface area of a filament ingot in an input picture, and segmenting the surface area of the filament ingot to obtain a segmentation result picture; carrying out contour analysis on the segmentation result graph to obtain all candidate tracks of tripwire in the input picture; extracting point sets of all the mutually communicated candidate tracks as point sets of a section of tripwire; and (3) screening and integrating the results of the point set according to the length and the position of the point set, comprising the following steps: performing line segment fitting on each point set, and recording the slope, the central point and the end point of the fitted line segment; and traversing all the point sets, merging the similar point sets, and integrating the point sets into the same tripwire. The embodiment of the invention improves the wire tripping number and the accuracy of track statistics.
Description
Technical Field
The invention relates to the technical field of spinning, in particular to a method and a device for detecting surface defects of chemical fiber spindles, electronic equipment and a storage medium.
Background
The chemical fiber industry is an industry with international competitive advantage in China, and the chemical fiber yield accounts for more than two thirds of the world; a chemical fiber spindle is an important product in the chemical fiber industry, and as shown in fig. 1A, fig. 1A is a perspective view of a typical chemical fiber spindle in the prior art. In the chemical fiber industry, a single spindle is usually taken as a minimum management unit, in the normal chemical fiber spindle production process, the surface of the spindle is smooth and uniform (as shown in a normal spindle top view in fig. 1B), but due to the influence of factors such as a production process or later transportation, the defects such as wire stumbling and the like appear on the surface of the spindle (as shown in a first spindle top view example with the wire stumbling defect in fig. 1C and a second spindle top view example with the wire stumbling defect in fig. 1D), the wire stumbling directly affects the product quality, and the production yield of a factory production line is reduced.
As can be seen from the top views of the filament ingots with the stumbled filament defects in fig. 1C and 1D, the stumbled filament defects appear as criss-cross filament strips with different normal textures on the surface of the filament ingots, the whole filament ingots look disordered, and some serious stumbled filament defects present a net-shaped structure on the end surface. The cause of snagging can be manifold, such as source wire path irregularities, paper tube bounce during winding, and equipment failure on the production line. The later stage directly influences the unwinding efficiency of downstream manufacturers, and even the defects of broken filaments and broken threads are introduced in the processing process. Therefore, the detection of the stumble yarn defect has great significance in the production field of chemical fiber yarn spindles.
In the prior art, the tripwire defect is generally detected by adopting the following two modes:
the first method is as follows: and (5) manually observing and selecting. The method is time-consuming, labor-intensive, low in efficiency and easy to cause missing detection and false detection due to the influence of subjective factors of people; and new defects are probably caused by artificial contact and transportation of the intermediate link.
The second method comprises the following steps: the method comprises the steps of using a traditional image processing technology to carry out binarization on an area to be detected, carrying out background separation to obtain profile information of a tripwire defect area, and obtaining the approximate tripwire number by using an area growing method. The production field is complicated and changeable, the traditional binarization processing method hardly considers the light change on the field, the anti-interference performance is poor, and the extracted contour information is inaccurate. And the stumbled silk on the inner side of the silk ingots of part of special batches (snowflake silk) is short and shallow, and when the silk ingot image is divided by fixed threshold binarization, the stumbled silk with smaller pixel value in the image is lost, so that the stumbled silk is missed. In the prior art, the length of a connected domain is taken as the length of a tripwire, and the number of connected domains is taken as the number of tripwires. When a plurality of stumbled wires are interlaced or interlaced to be in a net shape on the surface of the wire spindle, the extraction of a connected domain is inaccurate, and a statistical error of the number of stumbled wires may exist.
Disclosure of Invention
The invention provides a method and a device for detecting surface defects of chemical fiber spindles, electronic equipment and a storage medium, and improves the accuracy of detecting the number and the track of stumbled fibers.
The invention provides a method for detecting the surface defects of chemical fiber spindles, which comprises the following steps:
acquiring a surface area of a filament ingot in an input picture, and segmenting the surface area of the filament ingot to obtain a segmentation result picture;
carrying out contour analysis on the segmentation result graph to obtain all candidate tracks of tripwire in the input picture;
extracting point sets of all the mutually communicated candidate tracks as point sets of a section of tripwire;
and (3) screening and integrating the results of the point set according to the length and the position of the point set, comprising the following steps:
performing line segment fitting on each point set, and recording the slope, the central point and the end point of the fitted line segment;
and traversing all the point sets, merging the similar point sets, and integrating the point sets into the same tripwire.
Optionally, the merging the approximate point sets, integrating into the same tripwire, includes:
when the slope similarity of the two point sets meets a first set slope threshold, the end points of the two point sets are close to meet a set distance threshold, and the slope similarity of the central point connecting line and the slope of a fitting line segment of the two point sets to be merged meets a second set slope threshold, merging the two point sets;
the method further comprises the following steps:
the number of the merged point sets is used as the number of tripwires in the input picture, and the number of the points in each point set is used as the length of the tripwire.
Optionally, before the step of traversing all the point sets, the method further includes: all point sets are sorted in descending order by the number of contained points.
Optionally, performing contour analysis on the segmentation result graph to obtain all candidate trajectories of tripwire in the input picture, including:
thinning the tripwire mask drawing to obtain a tripwire framework drawing formed by each tripwire central line;
extracting intersection points and inflection points in the tripwire framework diagram;
the intersection points in the tripwire framework diagram are disconnected, and each section of tripwire in the mask diagram is ensured to be independent and not intersected with each other;
and traversing all pixels, and taking eight neighborhoods or four neighborhoods of each point as a search range to obtain a candidate track of tripwires in the input picture.
Optionally, the method further includes:
setting the traversed point as 0;
and when all the points are judged to be 0, finishing traversal and obtaining all the candidate tracks of the tripwire in the input picture.
Optionally, the breaking the intersection point in the tripwire skeleton diagram to ensure that each piece of tripwire in the mask diagram is independent from each other and does not intersect with each other includes: and setting the pixel value of the area with the set radius as 0 by taking the found intersection point and inflection point coordinates as centers.
Optionally, the segmenting process is performed on the surface area of the ingot to obtain a segmentation result map, including:
preprocessing the surface area of the silk ingot and then constructing a training data set;
constructing a semantic segmentation network;
training the semantic segmentation network by using the constructed training data set;
and inputting the image to be detected into a trained semantic segmentation network for reasoning to obtain a segmentation result graph with tripwire defects.
Optionally, the preprocessing the surface region of the filament ingot and then constructing a training data set includes:
carrying out overlapped sliding window image taking on the surface area of the silk ingot to obtain an image data block;
and calculating the mean value and the variance of all the image data blocks, and carrying out standardization processing on each image data block to obtain a training data set.
In another aspect, the present invention provides a device for detecting surface defects of chemical fiber spindles, comprising:
the surface area processing module is used for acquiring the surface area of the filament ingot in the input picture, and carrying out segmentation processing on the surface area of the filament ingot to obtain a segmentation result picture;
the contour analysis module is used for carrying out contour analysis on the segmentation result graph to obtain all candidate tracks of tripwire in the input picture;
the point set extraction module is used for extracting point sets of all the mutually communicated candidate tracks as a point set of a section of tripwire;
the result screening and integrating module is used for carrying out result screening and integrating on the point set according to the length and the position of the point set, and comprises: performing line segment fitting on each point set, and recording the slope, the central point and the end point of the fitted line segment; and traversing all the point sets, merging the similar point sets, and integrating the point sets into the same tripwire.
Another aspect of the present invention also provides an electronic device, including: the detection device comprises a processor and a memory, wherein the memory stores computer executable instructions, and the computer executable instructions are executed by the processor to realize the chemical fiber spindle surface defect detection method.
In another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method for detecting surface defects of chemical fiber ingots.
The embodiment of the invention detects the tripwire defect by judging whether the slope, the central point and the end point of the tripwire candidate track point set meet the set conditions, can accurately extract the tripwire defect area, particularly eliminates the condition of mistakenly merging two tripwire when the two tripwire are parallel by adopting the method of judging the slope of the central point connecting line, improves the detection accuracy rate, and can quickly count the number and the track of the tripwire. The embodiment of the invention can improve the accuracy of detecting the number and the track of the tripwires and can reach the accuracy of more than 99 percent; the accuracy of the tripwire number detection is important because tripwire production enterprises can grade the wire ingots according to the finally detected tripwire number and price according to the grade. The invention can analyze and process the stumble wire defect on line, adjust the process and the equipment running state of the production line, reduce the stumble wire defect and improve the yield of the production line. According to the embodiment of the invention, a simple-Unet semantic segmentation network is constructed through a deep learning technology, the tripwire defect characteristics are directly modeled and learned, certain robustness is provided for ambient light change and thread spindle surface texture change, and more accurate tripwire defect area contour information can be extracted; the invention can realize the accurate positioning of the tiny stumble wire defect. The embodiment of the invention carries out contour analysis, and solves the problem of inaccurate counting of the number of the tripped wires when the tripped wires on the surface of the wire ingot are mutually staggered by dividing the outline of the tripped wires according to the intersection points and the inflection points. The embodiment of the invention overcomes the defects of manual detection of the wire stumble on the surface of the chemical fiber spindle, such as intensive labor, low efficiency, high missing detection, poor robustness, inaccurate counting of the wire stumble quantity and track and the like in the prior art.
Drawings
FIG. 1A is a perspective view of a typical chemical fiber spindle of the prior art;
FIG. 1B is a top view of a normal wire ingot;
FIG. 1C is a top view of an exemplary first filament ingot with a snag filament defect;
FIG. 1D is an illustration of a second top view of a filament ingot with a snagged filament defect;
FIG. 2 is a flowchart of a method for detecting surface defects of a chemical fiber spindle according to a first embodiment of the present invention;
FIG. 3 is a diagram of a simple-Unet network architecture constructed in accordance with an embodiment of the present invention;
FIG. 4 is a wire-engaging zone profile view of an embodiment of the present invention;
FIG. 5 is a skeleton diagram of the tripwire region obtained by thinning the tripwire region mask diagram in the embodiment of the present invention;
FIG. 6A is a schematic view of the same tripwire with different sections determined by the slope of the center point connection line;
FIG. 6B is a schematic diagram of two parallel trip wires using a slope determination of a center point connection line;
FIG. 7 is a diagram showing the results of detecting tripwire defects according to the embodiment of the present invention.
Detailed Description
In order to make the objects, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 2, there is shown a flowchart of a method for detecting surface defects of a chemical fiber spindle according to a first embodiment of the present invention. The method comprises the following steps:
step 101: acquiring a surface area of a filament ingot in an input picture;
step 102: preprocessing the surface area of the silk ingot and then constructing a training data set;
step 103: constructing a simple-Unet semantic segmentation network;
104, training the segmentation network by using a training data set;
step 105: inputting the image to be detected into a trained network for reasoning to obtain a segmentation result graph with tripwire defects;
step 106: carrying out contour analysis on the segmentation result graph to obtain the number and the track of the candidate tripwire;
step 107: screening and integrating results according to a preset rule;
step 108: and outputting the number and the track of the tripwires.
In a preferred embodiment of the present application, the obtaining of the surface area of the ingot in the input picture in step 101 includes:
(B1) knowing that the outer side of the surface area of the filament ingot is circular, and obtaining the circle center and the radius of the surface area of the filament ingot by a rounding tool;
(B2) and converting the circle center and the radius to obtain the position information of a circumscribed rectangle of the surface area of the filament ingot, wherein the rectangular area is the surface area of the filament ingot.
In a preferred embodiment of the present application, the preprocessing of the surface area of the ingot in step 102 is performed to construct a training data set, which includes:
(C1) and (4) drawing the surface area of the silk ingot by overlapping sliding windows. Specifically, cutting image data of the surface area of the ingot by using a sliding window with the window size of 512 multiplied by 3 and the step length of 480 multiplied by 3 to obtain a plurality of image data blocks with the size of 512 multiplied by 3; since the map of the surface area of the ingot is large and inconvenient to handle, it is divided into small maps for handling.
(C2) And calculating the mean value and the variance of all the image data blocks, and carrying out standardization processing on each image data block to obtain a training data set.
In a preferred embodiment of the present application, a simple-Unet semantic segmentation network is constructed and trained in step 103, and the steps of the technical scheme are as follows:
(D1) and (3) constructing an encoder for extracting high-level semantic information in the input picture, wherein the encoder mainly comprises 10 convolutions, 4 maximum pooling layers and 1 cavity convolution layer, as shown in the left side of the figure 3.
(D2) And constructing a decoder for mapping the characteristics of the low-resolution encoder to the characteristics with the same resolution as the input picture so as to carry out pixel-level classification. Mainly comprising 12 convolutional layers, 4 deconvolution layers, and 3 characteristic cascade layers, as shown on the right side of fig. 3.
(D3) And a loss function, wherein the model adopts a weighted cross entropy loss function of a common softmax classifier, and is optimized based on a gradient descent and cosine annealing learning rate updating strategy. Assuming the output of the network
The label information of the specimen is
The output result of each dimension passing through the softmax classifier is
The output of the softmax classifier is
Model final loss function
Is composed of
Wherein,
，
for the weighting factors added on the basis of the original cross entropy loss function,
the number of the pixels is the total number of the pixels,
as a class of labels
The number of pixels. Thereby making the model more concerned with a smaller number of samples to alleviate the class imbalance problem in the image.
(D4) And training the constructed semantic segmentation network by using a training data set. Specifically, model optimization is performed by using random gradient descent (SGD), the initial learning rate is set to be 0.01, a cosine annealing learning rate updating strategy is adopted, the network convergence is accelerated, and the model is iterated for 10000 generations in total.
The embodiment of the invention adopts a deep learning semantic segmentation network, selects a deconvolution module with learning capability to perform upsampling, effectively avoids the problem that the conventional upsampling boundary characteristics are seriously lost, uses residual connection to introduce detail information in shallow layer characteristics, and simultaneously uses the void convolution to further enlarge the receptive field, so that the obtained segmentation result is more accurate in tripwire edge, the model is easier to converge, and the characteristics of tiny tripwire defects can be effectively extracted.
In a preferred embodiment of the present application, in step 105, a picture to be detected is input into a pre-trained semantic segmentation network for inference, and the steps of the technical scheme are as follows:
(E1) processing the picture to be detected in the same steps as the steps (C1) and (C2) to obtain sub-pictures to be detected, and recording the coordinate information of each sub-picture;
(E2) carrying out standardization operation on the subgraph to be detected obtained in the step (E1) by using the mean value and the variance obtained in the step (C2) to obtain test data;
(E3) inputting test data into a trained network for forward derivation to obtain a mask diagram of contour information of a tripwire area corresponding to each sub-graph;
(E4) and splicing the mask images of the sub-images according to the coordinate information of the sub-images to obtain a tripwire mask image corresponding to the surface area of the silk ingot, as shown in fig. 4. The upper half part of the figure is a photographed original picture, and the lower half part is a tripwire mask picture obtained by the embodiment of the invention.
In a preferred embodiment of the present application, the step 106 of performing a contour analysis on the segmentation result map includes the following steps:
(F1) thinning the tripwire mask diagram by using a K3M algorithm to obtain a framework formed by the central line of each tripwire area; the refined results are shown in fig. 5: the upper half part in FIG. 5 is a photographed original picture, and the lower half part in FIG. 5 is a skeleton diagram obtained by thinning a tripwire mask diagram;
(F2) extracting intersection points and inflection points in the tripwire skeleton diagram by using a Harris angular point detection algorithm;
(F3) and the intersection points in the tripwire framework diagram are disconnected, so that each section of tripwire in the mask diagram is independent and does not intersect with each other. Specifically, the pixel value of the area having the center of the coordinates of the found intersection and inflection point and having a radius of N pixels is set to 0. The value of N may be determined according to the mask pattern line width.
(F4) And traversing all pixels, taking eight neighborhoods or four neighborhoods of each point as a search range, extracting point sets of all mutually communicated tracks as track point sets of a section of tripwire framework, and setting the traversed points as 0. When all the points are 0, finishing traversal to obtain all candidate tracks of tripwires in the input picture;
in a preferred embodiment of the present application, the result screening integration method comprises the following steps:
(G1) performing line segment fitting on each point set, and recording the slope, the central point and the end point of the fitted line segment;
(G2) sorting all point sets in descending order by the number of contained points, descending order since longer segments are more reliable (more likely to be a stumbling wire) and shorter segments are more likely to be a disturbance.
Traversing all the point sets, and when the slopes of the two point sets are similar (satisfy)
) The two end points of the two point sets are close to each other (satisfy the requirement of
，
To set a distance threshold) and the slope of the center point connecting line is close to the slope of the line segment fitted to the two point sets to be merged (satisfies the requirement of merging the two point sets)
) Merging two point sets, namely considering the two point sets to be different parts of the same tripwire;
with particular reference to fig. 6A and 6B. FIG. 6A is a schematic view of the same tripwire with different sections determined by the slope of the center point connection line; the figure shows a first end point 603 of a first section of tripwire and a first end point 604 of a second section of tripwire, and also shows a central point 601 of the first section of tripwire, a central point 602 of the second section of tripwire, a central point 601 of the first section of tripwire and a central point connecting line 605 of the central point 602 of the second section of tripwire, and the two sections of tripwire are judged to be the same tripwire by judging that the slopes of the first section of tripwire, the second section of tripwire and the central point connecting line 605 are similar.
FIG. 6B is a schematic diagram of two parallel trip wires using a slope determination of a center point connection line. The figure shows a first end point 606 of a third section of tripwire and a first end point 607 of a fourth section of tripwire, the figure also shows a central point 608 of the third section of tripwire, a central point 609 of the fourth section of tripwire, a central point 608 of the third section of tripwire and a central point connecting line 610 of the central point 609 of the fourth section of tripwire, and the two sections of tripwire are judged not to be the same tripwire by judging that the slopes of the third section of tripwire, the fourth section of tripwire and the central point connecting line 610 are not similar.
Therefore, the embodiment of the invention eliminates the condition that two tripwire are mistakenly combined when the two tripwire are parallel, and improves the tripwire detection accuracy.
Wherein,k 1is the slope of the first set of points,k 2is the slope of the second set of points,
setting a slope threshold for the first;
(G3) the number of the remaining point sets after merging is used as the number of tripwires in the input picture, and the number of the points in each point set is used as the length of the tripwire; results as shown in fig. 7, the upper half of fig. 7 is a picture of the original tripwire shot, and the lower half of fig. 7 is a picture of the final tripwire result obtained by the defect detection method of the present invention. The different shades of gray in the figure represent different tripwire, and the numbers in the upper left corner represent the number of tripwire found in the figure.
The embodiment of the invention also provides a device for detecting the surface defects of the chemical fiber spindles, which comprises:
the surface area processing module is used for acquiring the surface area of the filament ingot in the input picture, and carrying out segmentation processing on the surface area of the filament ingot to obtain a segmentation result picture;
the contour analysis module is used for carrying out contour analysis on the segmentation result graph to obtain all candidate tracks of tripwire in the input picture;
the point set extraction module is used for extracting point sets of all the mutually communicated candidate tracks as a point set of a section of tripwire;
the result screening and integrating module is used for carrying out result screening and integrating on the point set according to the length and the position of the point set, and comprises: performing line segment fitting on each point set, and recording the slope, the central point and the end point of the fitted line segment; and traversing all the point sets, merging the similar point sets, and integrating the point sets into the same tripwire.
Wherein, in a specific embodiment,
the surface area processing module determines the surface area of the silk ingot through the rounding tool and inputs the surface area of the silk ingot into a subsequent module for processing; preprocessing the surface area of the silk ingot and then constructing a training data set; building a simple-Unet semantic segmentation network and training; inputting the picture to be detected into a pre-trained network, and carrying out forward reasoning to obtain a mask map of the tripwire defect area outline information.
And the contour analysis module is used for determining the tripwire defect when the silk thread is changed into a straight line by taking the center of the silk spindle as the center of a circle and taking the normal silk thread as an arc shape according to the defect characteristic of the tripwire. Therefore, after the mask map is thinned, intersection points and inflection points existing in the picture are searched, intersecting straight lines in the mask map are divided into line segments which are independent and not intersected, pixel traversal is carried out, and a point set of each line segment is obtained.
And the result screening and integrating module merges the candidate point sets according to preset rules such as the length and the position of the tripwire, merges the point sets according to the slope, the end point distance and the center distance of the line segment fitted by the point sets, finally obtains the candidate point set of the tripwire in the input picture and obtains the final tripwire number and track.
The embodiment of the invention provides a tripwire track point set screening and integrating method; the method detects the tripwire defect by judging whether the slope, the central point and the end point of the tripwire candidate track point set meet the set conditions, can accurately extract the tripwire defect area, particularly eliminates the condition of mistakenly merging two tripwire when the two tripwire are parallel by adopting the method of judging the slope of the central point connecting line, and improves the detection accuracy. The technical scheme in the embodiment of the application constructs a simple-Unet semantic segmentation network and provides a contour analysis method; according to the embodiment of the invention, a simple-Unet semantic segmentation network is constructed through a deep learning technology, the tripwire defect characteristics are directly modeled and learned, certain robustness is provided for ambient light change and thread spindle surface texture change, and more accurate tripwire defect area contour information can be extracted; the invention can realize the accurate positioning of the tiny stumble wire defect.
In another aspect, an embodiment of the present invention further provides an electronic device, including: the detection device comprises a processor and a memory, wherein the memory stores computer executable instructions, and the computer executable instructions are executed by the processor to realize the chemical fiber spindle surface defect detection method.
In another aspect, the present invention further provides a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the method for detecting surface defects of chemical fiber ingots is implemented.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.
Claims (11)
1. A method for detecting surface defects of chemical fiber spindles is characterized by comprising the following steps:
acquiring a surface area of a filament ingot in an input picture, and segmenting the surface area of the filament ingot to obtain a segmentation result picture;
carrying out contour analysis on the segmentation result graph to obtain all candidate tracks of tripwire in the input picture;
extracting point sets of all the mutually communicated candidate tracks as point sets of a section of tripwire;
and (3) screening and integrating the results of the point set according to the length and the position of the point set, comprising the following steps:
performing line segment fitting on each point set, and recording the slope, the central point and the end point of the fitted line segment;
and traversing all the point sets, merging the similar point sets, and integrating the point sets into the same tripwire.
2. The method of claim 1, wherein merging the sets of points that approximate the phase into a single tripwire comprises:
when the slope similarity of the two point sets meets a first set slope threshold, the end points of the two point sets are close to meet a set distance threshold, and the slope similarity of the central point connecting line and the slope of a fitting line segment of the two point sets to be merged meets a second set slope threshold, merging the two point sets;
the method further comprises the following steps:
the number of the merged point sets is used as the number of tripwires in the input picture, and the number of the points in each point set is used as the length of the tripwire.
3. The method of claim 1, wherein said step of traversing all sets of points is preceded by the step of: all point sets are sorted in descending order by the number of contained points.
4. The method of claim 1, wherein performing contour analysis on the segmentation result map to obtain all candidate trajectories of tripwire in the input picture comprises:
thinning the tripwire mask drawing to obtain a tripwire framework drawing formed by each tripwire central line;
extracting intersection points and inflection points in the tripwire framework diagram;
the intersection points in the tripwire framework diagram are disconnected, and each section of tripwire in the mask diagram is ensured to be independent and not intersected with each other;
and traversing all pixels, and taking eight neighborhoods or four neighborhoods of each point as a search range to obtain a candidate track of tripwires in the input picture.
5. The method of claim 4, further comprising:
setting the traversed point as 0;
and when all the points are judged to be 0, finishing traversal and obtaining all the candidate tracks of the tripwire in the input picture.
6. The method of claim 4, wherein said breaking the intersection points in the tripwire skeleton map to ensure that each piece of tripwire in the mask map is independent of each other and does not intersect each other, comprises: and setting the pixel value of the area with the set radius as 0 by taking the found intersection point and inflection point coordinates as centers.
7. The method of claim 1, wherein the segmenting the surface region of the ingot to obtain a segmentation result map comprises:
preprocessing the surface area of the silk ingot and then constructing a training data set;
constructing a semantic segmentation network;
training the semantic segmentation network by using the constructed training data set;
and inputting the image to be detected into a trained semantic segmentation network for reasoning to obtain a segmentation result graph with tripwire defects.
8. The method of claim 7, wherein preprocessing the surface region of the ingot to construct a training data set comprises:
carrying out overlapped sliding window image taking on the surface area of the silk ingot to obtain an image data block;
and calculating the mean value and the variance of all the image data blocks, and carrying out standardization processing on each image data block to obtain a training data set.
9. A chemical fiber spindle surface defect detection device is characterized in that the device comprises:
the surface area processing module is used for acquiring the surface area of the filament ingot in the input picture, and carrying out segmentation processing on the surface area of the filament ingot to obtain a segmentation result picture;
the contour analysis module is used for carrying out contour analysis on the segmentation result graph to obtain all candidate tracks of tripwire in the input picture;
the point set extraction module is used for extracting point sets of all the mutually communicated candidate tracks as a point set of a section of tripwire;
the result screening and integrating module is used for carrying out result screening and integrating on the point set according to the length and the position of the point set, and comprises: performing line segment fitting on each point set, and recording the slope, the central point and the end point of the fitted line segment; and traversing all the point sets, merging the similar point sets, and integrating the point sets into the same tripwire.
10. An electronic device, comprising: a processor and a memory, wherein the memory stores computer executable instructions which, when executed by the processor, implement the method of detecting defects on the surface of a chemical fiber spindle according to any one of claims 1 to 8.
11. A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method for detecting defects on the surface of a chemical fiber spindle according to any one of claims 1 to 8 is implemented.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110893697.XA CN113344923B (en) | 2021-08-05 | 2021-08-05 | Chemical fiber spindle surface defect detection method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110893697.XA CN113344923B (en) | 2021-08-05 | 2021-08-05 | Chemical fiber spindle surface defect detection method and device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113344923A true CN113344923A (en) | 2021-09-03 |
| CN113344923B CN113344923B (en) | 2021-11-23 |
Family
ID=77480787
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110893697.XA Active CN113344923B (en) | 2021-08-05 | 2021-08-05 | Chemical fiber spindle surface defect detection method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113344923B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114299065A (en) * | 2022-03-03 | 2022-04-08 | 科大智能物联技术股份有限公司 | Method for detecting and grading defective appearance forming defects of silk ingots, storage medium and equipment |
| WO2023070919A1 (en) * | 2021-10-25 | 2023-05-04 | 无锡闻泰信息技术有限公司 | Product movement track calculation method and apparatus, device, and storage medium |
| CN116563276A (en) * | 2023-07-05 | 2023-08-08 | 菲特（天津）检测技术有限公司 | Chemical fiber filament online defect detection method and detection system |
| CN116912260A (en) * | 2023-09-15 | 2023-10-20 | 沂水友邦养殖服务有限公司 | Broiler chicken breeding health state detection method based on artificial intelligence |
| CN118411700A (en) * | 2024-06-27 | 2024-07-30 | 浙江恒逸石化有限公司 | Detection method, detection device, detection equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109255787A (en) * | 2018-10-15 | 2019-01-22 | 杭州慧知连科技有限公司 | Silk ingot scratch detection system and method based on deep learning and image processing techniques |
| CN111815564A (en) * | 2020-06-09 | 2020-10-23 | 浙江华睿科技有限公司 | Method and device for detecting silk ingots and silk ingot sorting system |
| CN112330595A (en) * | 2020-10-13 | 2021-02-05 | 浙江华睿科技有限公司 | Tripwire detection method and device, electronic equipment and storage medium |
| CN112330598A (en) * | 2020-10-14 | 2021-02-05 | 浙江华睿科技有限公司 | Method and device for detecting stiff silk defects on chemical fiber surface and storage medium |
-
2021
- 2021-08-05 CN CN202110893697.XA patent/CN113344923B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109255787A (en) * | 2018-10-15 | 2019-01-22 | 杭州慧知连科技有限公司 | Silk ingot scratch detection system and method based on deep learning and image processing techniques |
| CN111815564A (en) * | 2020-06-09 | 2020-10-23 | 浙江华睿科技有限公司 | Method and device for detecting silk ingots and silk ingot sorting system |
| CN112330595A (en) * | 2020-10-13 | 2021-02-05 | 浙江华睿科技有限公司 | Tripwire detection method and device, electronic equipment and storage medium |
| CN112330598A (en) * | 2020-10-14 | 2021-02-05 | 浙江华睿科技有限公司 | Method and device for detecting stiff silk defects on chemical fiber surface and storage medium |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023070919A1 (en) * | 2021-10-25 | 2023-05-04 | 无锡闻泰信息技术有限公司 | Product movement track calculation method and apparatus, device, and storage medium |
| CN114299065A (en) * | 2022-03-03 | 2022-04-08 | 科大智能物联技术股份有限公司 | Method for detecting and grading defective appearance forming defects of silk ingots, storage medium and equipment |
| CN114299065B (en) * | 2022-03-03 | 2022-06-17 | 科大智能物联技术股份有限公司 | Method for detecting and grading defective appearance forming defects of silk ingots, storage medium and equipment |
| CN116563276A (en) * | 2023-07-05 | 2023-08-08 | 菲特（天津）检测技术有限公司 | Chemical fiber filament online defect detection method and detection system |
| CN116563276B (en) * | 2023-07-05 | 2023-09-01 | 菲特（天津）检测技术有限公司 | Chemical fiber filament online defect detection method and detection system |
| CN116912260A (en) * | 2023-09-15 | 2023-10-20 | 沂水友邦养殖服务有限公司 | Broiler chicken breeding health state detection method based on artificial intelligence |
| CN116912260B (en) * | 2023-09-15 | 2023-11-28 | 沂水友邦养殖服务有限公司 | Broiler chicken breeding health state detection method based on artificial intelligence |
| CN118411700A (en) * | 2024-06-27 | 2024-07-30 | 浙江恒逸石化有限公司 | Detection method, detection device, detection equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113344923B (en) | 2021-11-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113344923B (en) | Chemical fiber spindle surface defect detection method and device, electronic equipment and storage medium | |
| CN110084095B (en) | Lane line detection method, lane line detection apparatus, and computer storage medium | |
| WO2018107939A1 (en) | Edge completeness-based optimal identification method for image segmentation | |
| CN111815564B (en) | Method and device for detecting silk ingots and silk ingot sorting system | |
| CN111191628B (en) | Remote sensing image earthquake damage building identification method based on decision tree and feature optimization | |
| CN109034245A (en) | A kind of object detection method merged using characteristic pattern | |
| CN109615604A (en) | Accessory appearance flaw detection method based on image reconstruction convolutional neural networks | |
| CN114923912A (en) | Machine vision-based method for detecting outer surface defects of nylon spinning cakes | |
| CN112883797A (en) | Tobacco shred sundry detection method based on Yolo V3 model | |
| CN118429341B (en) | Defect detection method for gradual change texture fabric | |
| CN111652846A (en) | Semiconductor defect identification method based on characteristic pyramid convolution neural network | |
| WO2020164042A1 (en) | Region merging image segmentation algorithm based on boundary extraction | |
| Lee | A Study on Fruit Quality Identification Using YOLO V2 Algorithm | |
| CN114092462A (en) | Composite defect detection method based on semantic segmentation and target detection fusion model | |
| Jeffrey Kuo et al. | Self-organizing map network for automatically recognizing color texture fabric nature | |
| CN114078106A (en) | Defect detection method based on improved Faster R-CNN | |
| CN114022442B (en) | Unsupervised learning-based fabric defect detection algorithm | |
| Shabankareh et al. | Improvement of edge-tracking methods using genetic algorithm and neural network | |
| CN114267052A (en) | Pedestrian detection method based on yolov5 | |
| CN114972359B (en) | Mesh fabric defect rapid detection method based on symmetry analysis | |
| CN115937077A (en) | Workpiece surface micro-defect detection method based on improved SSD algorithm | |
| JP2005346222A (en) | Sweat gland pore removing device, sweat gland pore removing method and sweat gland pore removing program | |
| CN113963049A (en) | Method for measuring pixel size and actual size and electronic equipment | |
| Haverkamp et al. | Complementary methods for extracting road centerlines from IKONOS imagery | |
| CN108229486B (en) | Fiber identification and segmentation method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |