CN111627053B - Method and system for detecting cleaning cleanliness of universal spinneret orifice - Google Patents
Method and system for detecting cleaning cleanliness of universal spinneret orifice Download PDFInfo
- Publication number
- CN111627053B CN111627053B CN202010412672.9A CN202010412672A CN111627053B CN 111627053 B CN111627053 B CN 111627053B CN 202010412672 A CN202010412672 A CN 202010412672A CN 111627053 B CN111627053 B CN 111627053B
- Authority
- CN
- China
- Prior art keywords
- curve
- detection
- contour
- hole
- dirt
- 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.)
- Active
Links
- 230000003749 cleanliness Effects 0.000 title claims abstract description 82
- 238000000034 method Methods 0.000 title claims abstract description 73
- 238000004140 cleaning Methods 0.000 title claims abstract description 49
- 238000001514 detection method Methods 0.000 claims abstract description 194
- 238000012937 correction Methods 0.000 claims abstract description 15
- 238000013507 mapping Methods 0.000 claims abstract description 14
- 230000011218 segmentation Effects 0.000 claims description 46
- 238000009826 distribution Methods 0.000 claims description 20
- 239000002689 soil Substances 0.000 claims description 15
- 238000000605 extraction Methods 0.000 claims description 13
- 238000005457 optimization Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 10
- 238000004458 analytical method Methods 0.000 claims description 7
- 238000003754 machining Methods 0.000 claims description 6
- 238000011160 research Methods 0.000 claims description 6
- 235000004035 Cryptotaenia japonica Nutrition 0.000 claims description 5
- 102000007641 Trefoil Factors Human genes 0.000 claims description 5
- 235000015724 Trifolium pratense Nutrition 0.000 claims description 5
- 238000007664 blowing Methods 0.000 claims description 4
- 230000004927 fusion Effects 0.000 claims description 4
- 239000000203 mixture Substances 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 230000007704 transition Effects 0.000 claims description 4
- 238000011049 filling Methods 0.000 claims description 3
- 238000007670 refining Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 230000001105 regulatory effect Effects 0.000 claims description 2
- 229910000831 Steel Inorganic materials 0.000 claims 1
- 239000010959 steel Substances 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 25
- 238000004364 calculation method Methods 0.000 description 7
- 230000002159 abnormal effect Effects 0.000 description 5
- 230000004807 localization Effects 0.000 description 5
- 239000000126 substance Substances 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 239000000835 fiber Substances 0.000 description 4
- 238000002156 mixing Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 239000011148 porous material Substances 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000007380 fibre production Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000007789 sealing Methods 0.000 description 2
- 238000009987 spinning Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000004506 ultrasonic cleaning Methods 0.000 description 2
- 238000005299 abrasion Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000003556 assay Methods 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000009833 condensation Methods 0.000 description 1
- 238000002425 crystallisation Methods 0.000 description 1
- 230000008025 crystallization Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012850 discrimination method Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000007667 floating Methods 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
- 239000003550 marker Substances 0.000 description 1
- 238000000386 microscopy Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 229920000728 polyester Polymers 0.000 description 1
- 229920000642 polymer Polymers 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000005507 spraying Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 239000004753 textile Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000009966 trimming Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- 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
- G01N21/94—Investigating contamination, e.g. dust
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/181—Segmentation; Edge detection involving edge growing; involving edge linking
-
- 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
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Geometry (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Spinning Methods And Devices For Manufacturing Artificial Fibers (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method and a system for detecting cleaning cleanliness of a universal spinneret orifice, comprising the following steps: extracting an edge contour closed curve and a curvature curve of the edge contour closed curve of the standard clean hole high gradient threshold value, and constructing a parameterized segmented closed curve standard detection template containing control points; in the dirt detection stage, mapping the parameterized segmented closed curve standard detection template onto a contour to be detected through non-rigid registration; constructing a dirt detection curve, wherein the dirt detection curve is formed by weighting the parameterized segmented closed curve standard detection template, the distance curve of the contour to be detected based on nearest neighbor searching and the mixed curvature curve; and performing floor benchmark correction on the dirt detection curve to form a global threshold value so as to divide and locate dirt and quantify the cleanliness index.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method and a system for detecting cleaning cleanliness of a universal spinneret orifice.
Background
In chemical fiber production quality monitoring, whether the spinneret holes are cleaned or not is an important factor affecting the quality of the formed fibers. At present, most of computer vision detection methods are adopted, and characteristic indexes such as the area, the perimeter and the like of the outline area of the hole are analyzed and extracted through rear projection light source imaging to judge whether dirt remains in the hole. The method has obvious effect on conventional dirt, has high misjudgment rate on fine dirt, and particularly has basically ineffective judgment on indexes such as area, perimeter and the like when the deformation degree of the hole is large.
Specifically, the chemical fiber is a silk thread formed by spraying molten polyester polymer through a spinneret orifice with a specific shape on a spinneret plate under high temperature and high pressure conditions to form re-condensation crystallization, the shape and the size of the spinneret orifice determine the chemical fiber performance, and the spinneret plate can be reused after being subjected to ultrasonic cleaning and drying after a certain service period. The quality monitoring of the spinneret orifices is an important process for guaranteeing the quality of the spinneret, wherein dirt detection is the most important link, and abnormal spinning quality parameters can be caused after the unclean spinneret plate is on line. In actual production, detection personnel can carry out dirt analysis, hole positioning, automatic hole blowing and manual cleaning through full-automatic computer vision detection equipment after the spinneret plate is cleaned and dried. In recent years, computer vision technology is widely applied in the textile industry field, but the related research of the micropore defect detection is less, and the technology is mainly concentrated on the application level of general vision detection such as accurate focusing, size detection, control positioning and the like. In practical application, the general cleanliness detection method is to collect hole images through an image collecting device under the illumination of a vertical rear projection light source, if dirt exists on the periphery of a hole outline, the hole permeability is reduced, the edge outline of a hole permeable area is changed, and therefore whether the dirt remains in the hole can be judged by analyzing and extracting characteristic indexes such as the area, the perimeter and the like of the hole outline area. However, the conventional area and perimeter index discrimination method has obvious effect on detecting dirt with obvious visual difference, and has high misjudgment rate on most fine dirt, because the area and perimeter index fluctuation caused by the dispersion of the machining precision of the hole per se exceeds the fine dirt detection sensitivity range.
Disclosure of Invention
The invention aims to provide a general spinneret orifice cleaning cleanliness detection method and system, which are used for solving the problem of high misjudgment rate of fine dirt of the existing spinneret orifice.
In order to solve the technical problems, the invention provides a method and a system for detecting the cleaning cleanliness of a universal spinneret orifice, wherein the method for detecting the cleaning cleanliness of the universal spinneret orifice comprises the following steps:
extracting an edge contour closed curve and a curvature curve of the edge contour closed curve of the standard clean hole high gradient threshold value, and constructing a parameterized segmented closed curve standard detection template containing control points;
in the dirt detection stage, mapping the parameterized segmented closed curve standard detection template onto a contour to be detected through non-rigid registration;
constructing a dirt detection curve, wherein the dirt detection curve is formed by weighting the parameterized segmented closed curve standard detection template, the distance curve of the contour to be detected based on nearest neighbor searching and the mixed curvature curve;
and performing floor benchmark correction on the dirt detection curve to form a global threshold value so as to divide and locate dirt and quantify the cleanliness index.
Optionally, in the method for detecting the cleaning cleanliness of a general spinneret orifice, the method for detecting the cleaning cleanliness of a general spinneret orifice further includes:
Extracting a standard hole profile and a curvature curve thereof through standard clean hole image edge analysis, and carrying out segmentation and fitting on a straight line segment and an arc segment of the hole profile based on a double-threshold segmentation algorithm of the curvature curve;
dividing and fitting according to the straight line segment and the circular arc segment of the standard hole profile, and reconstructing a parameterized segmented closed curve standard detection template based on control points;
extracting an edge contour closed curve of a high gradient threshold value of a hole to be detected, calculating the length of the edge contour closed curve of the high gradient threshold value of the hole to be detected and the area surrounded by the edge contour closed curve, detecting significant dirt, if the detection result is unqualified, detecting the next hole to be detected, otherwise, detecting non-significant dirt;
extracting an edge contour closed curve of a low gradient threshold value of a hole to be detected, mapping the edge contour closed curve of a high gradient threshold value of the standard clean hole to the edge contour closed curve of the high gradient threshold value of the hole to be detected through non-rigid registration, constructing a nearest neighbor search-based distance curve between the edge contour closed curve of the high gradient threshold value of the standard clean hole and the edge contour curve of the low gradient threshold value of the hole to be detected, and constructing a dirt detection curve by taking a mixed weighting curve as a weight;
Positioning and quantifying dirt by using global threshold segmentation through floor reference calibration processing of the dirt detection curve, constructing a cleanliness index for dirt discrimination, and judging whether hole dirt detection is qualified or not;
round holes, flat holes, trefoil holes and cross holes are used as research objects, and result analysis and conclusion are given.
Optionally, in the method for detecting cleaning cleanliness of a universal spinneret orifice, extracting an edge profile closed curve of a standard cleaning orifice high gradient threshold includes:
extracting a hole edge profile curve by adopting a Canny edge operator, respectively detecting strong and weak edges by adopting a double-gradient threshold method, and only retaining the weak edges connected with the strong edges;
setting a first gradient threshold sigma thr1 And a second gradient threshold sigma thr2 The strong contour edge and the strong and weak combined contour edge are respectively extracted, the first gradient threshold value is smaller than the second gradient threshold value, and the first gradient threshold value and the second gradient threshold value respectively correspond to Canny detection double-ladder threshold [ sigma ] thr1 ,3*σ thr1 ][ sigma ] thr2 ,3*σ thr2 ];
The extracted image edges are a first edge and a second edge respectively, wherein the first edge is used for constructing parameterized segmented closed curve standard detection and non-rigid registration, and significant dirt detection, and the second edge is used for non-significant dirt detection.
Optionally, in the method for detecting cleaning cleanliness of a universal spinneret orifice, reconstructing a parameterized piecewise closed curve standard detection template based on control points includes:
refining based on 8 neighborhood and searching edge crossing points, and breaking the edge crossing points;
searching curve end points, tracking the contour track, counting the length, and if the length is smaller than a set threshold value, reversely tracking along the end points and erasing the track;
searching the longest curve segment, marking the starting point and the end point position of the longest curve segment, searching the starting point of the nearest adjacent curve segment from the end point position, marking the searched curve segment, and connecting the end point-starting point pair by using a straight line;
and continuing to search unmarked nearest neighbor curve segments from the last adjacent curve segment end point, continuing to connect and iterate in a circulating way until the searched start point is the start point of the initial longest line segment, and ending the iteration.
Optionally, in the method for detecting the cleaning cleanliness of a universal spinneret orifice, reconstructing an edge profile closed curve of the high gradient threshold of the orifice to be detected from the first edge as follows:
which is a kind ofFor the coordinates m of the contour points 1 Counting the number of the contour points; />
Gauge L 1 The indexes of the circumferential P and the filling surface S of the hole pattern are counted, the indexes of qualified and clean N holes (N > 1024) in the same hole pattern are counted, and the average mu is calculated p 、μ s Sum standard sigma p 、σ s Index, taking the optimal threshold value as tau p =μ p ±α*σ p 、τ s =μ s ±α*σ s I.e. |P-. Mu. p |≤α*σ p And |S-. Mu. s |≤α*σ s And if the cleanliness is qualified, entering the subsequent non-obvious dirt detection, otherwise, directly carrying out the next hole detection alpha=3.5.
Optionally, in the method for detecting the cleaning cleanliness of a general spinneret orifice, the method for detecting the cleaning cleanliness of a general spinneret orifice further includes:
the nonlinear registration of the contour to be detected is carried out by adopting parameterized straight line segment and circular arc segment piecewise curve models based on control point positioning, and the nonlinear registration specifically comprises the following steps: calculating a closed contour curvature curve, setting a proper double threshold value according to curvature distribution characteristics, dividing the contour curve into a plurality of straight lines and large arc sections, automatically positioning each sectional control point position, decomposing the closed curve into straight line sections and arc sections which are connected end to end, fitting each divided line section by adopting a least square method according to corresponding straight line or arc parameters, calculating the intersection point of each adjacent line section after fitting as a new control point, and reconstructing the whole new parameterized sectional closed curve as a standard matching template.
Optionally, in the method for detecting the cleaning cleanliness of a general spinneret orifice, the method for detecting the cleaning cleanliness of a general spinneret orifice further includes:
Extracting the edge contour sealing curve L of the high gradient threshold value of the hole to be detected 1 Ith trackingAnd a radius range (r) 2 ) Inner furthest contour +.>And->Fitting a circle by adopting a least square algorithm, and calculating c by taking the radius of the fitted circle as the curvature i Gauge L 1 The overall profile curvature distribution of (C = { C) i ,i∈[1,m 1 ]};
Setting a curvature dividing threshold value c thr1 And c thr2 Segment length threshold g thr ;c i >c thr1 Marked as straight line segment, c i ≤c thr1 And c i >c thr2 Marking as a large arc section, otherwise marking as a small arc section;
the control points of each divided line segment are respectively the starting position and the ending position, the starting position and the ending position of two adjacent divided line segments share the same position, and meanwhile, the arc segment is added with a midpoint control point which is positioned at the center of the arc and used for positioning the direction of the arc compared with the straight line segment;
after curvature segmentation, carrying out length statistics on segmented line segments to obtain a length set G= { G i ,i∈[1,f g ]Gram of (g) i Representing the length f of each segment g The number of the segments is divided;
searching for minimum segmentation line segment length g kmin For example g kmin <g thr Discarding kmin segment, wherein the common control points of two adjacent segment segments are respectively replaced by the midpoints of the current segment;
carrying out length statistics, minimum length judgment and fusion again, and carrying out loop iteration until all the segmentation line segments meet the conditions; if only one arc segment is obtained after the curvature is divided, the hole pattern is a conventional circle, and the hole pattern is divided into two connected arc segments; here the curvature segmentation parameter takes c thr1 =400,c thr2 =120,g thr =40;
The parameterized piecewise closed curve standard detection template is represented as follows:
F={f i =(f i,x ,f i,y ),i∈[1,f n ]},P={p i =(s,e),i∈[1,f l ],s∈[1,f n ],e∈[1,f n ],s≠e},
Q={q i =(s,c,e),i∈[1,fc],s∈[1,f n ],e∈[1,f n ],s≠e,s≠c,c≠e},
wherein F is that all characteristic point sets P are straight line segment index sets Q are arc segment index sets,fitting a curvature radius set s for the arc segment to obtain a starting characteristic point number e as an ending characteristic point number, and c as an arc midpoint characteristic point number f n For the total number f of the whole feature points l For the total number of straight line segments f c Is the total number of arc line segments.
Optionally, in the method for detecting cleaning cleanliness of a universal spinneret orifice, the non-rigid registration includes:
the method comprises the steps of adopting a nonlinear least square parameter optimization method, regulating the position of a line segment control F of a parameterized piecewise closed curve standard detection template, carrying out non-rigid registration on the parameterized piecewise closed curve standard detection template to an edge contour closed curve of a high gradient threshold value of a hole to be detected, which is reconstructed by a first edge, so that the distance mean square error of each point of the contour curve to be detected projected to a template reconstruction piecewise closed curve { F, P, Q } is minimum;
is provided withFor the projection distance function of the k point of the contour to be detected on the straight line segment of the parameterized segmented closed curve standard detection template i,Detecting a projection distance function of a template j arc segment for a profile k point to be detected on the parameterized piecewise closed curve standard; the nonlinear least square optimization objective function based on the parameterized piecewise closed curve standard detection template control F is as follows:
Setting F for optimized template control point τ The optimization method adopts a ceres-software algorithm library.
Optionally, in the method for detecting the cleaning cleanliness of the universal spinneret orifice, the edge profile closed curve of the low gradient threshold value of the hole to be detected extracted from the second edge is:
which is a kind ofFor the coordinates m of the contour points 2 Counting the number of the contour points;
from F τ The reconstructed closed curve is:
which is a kind ofFor the coordinates m of the contour points 3 The number of points of the reconstructed curve;
calculating a reconstruction curve L 3 The distance maximum value of all search points (j, i) at the point i in (i) is taken as the distance curve value of the point as follows:
the curvature weight of the straight line segment is 1, the arc segment weighting function adopts a gradient interpolation method, and the straight line segment and the arc segment weighting function are as follows:
its D l Is L 3 Middle section straight line area point set D c Is L 3 Center c section arc area point set
The method comprises the steps of carrying out a first treatment on the surface of the The blend curvature weighting function is as follows:
which is a convolution operator, gaussian g (i) is used to smooth the coefficient jumps in the curve weighted curvature transition region, μ 1 =0,σ 1 =3.0;
The single control point regional gaussian weighting function is as follows:
f i control point L for f 3 A position index; mixing
The gaussian weighted function is as follows:
sigma of which 2 As a Gaussian function standard deviation sigma 2 =2.5;
The hybrid curvature weighting function multiplied by the single control point region gaussian weighting function is as follows:
W t (i)=W t1 (i)·W t2 (i) (8)
Dirt detection curve L weighted by distance curve 5 The following are provided:
L 5 ={W(i)=W t (i)·D(i),i∈[1,m 3 ]} (9)。
optionally, in the method for detecting the cleaning cleanliness of a general spinneret orifice, the method for detecting the cleaning cleanliness of a general spinneret orifice further includes:
eliminating the non-uniformity of the floor reference by adopting tophat conversion, and setting the global dirt threshold to be 2.5;
let i dirt fingers P i Is { W i ,H i ,S i },i∈[1,n]N is the number of local dirt, and the overall cleanliness index of the spinneret orifice is calculated as follows:
h is of thr =15,w thr =4;
In the actual detection, a cleanliness threshold P meeting the process requirements is set thr ,P>P thr And when the hole is judged to be unqualified, automatic hole blowing is firstly carried out, then the hole is re-detected, and if the hole is still unqualified, manual cleaning is carried out.
The invention also provides a general spinneret orifice cleaning cleanliness detection system, which comprises a closed curve extraction module, a template mapping module, a dirt detection curve module and a reference correction module, wherein:
the closed curve extraction module is configured to extract an edge contour closed curve of a standard clean hole high gradient threshold value and a curvature curve thereof, and construct a parameterized segmented closed curve standard detection template containing control points;
in a dirt detection stage, the template mapping module is configured to map the parameterized piecewise closed curve standard detection template onto a contour to be detected through non-rigid registration;
The dirt detection curve module is configured to construct a dirt detection curve, wherein the dirt detection curve is formed by weighting the parameterized segmented closed curve standard detection template, the distance curve of the profile to be detected based on nearest neighbor search and the mixed curvature curve;
the reference correction module is configured to form a global threshold by floor reference correction of the soil detection curve to segment and locate soil and quantify a cleanliness index.
In the method and the system for detecting the cleaning cleanliness of the universal spinneret holes, tiny dirt characteristic research is taken as a starting point, a clean hole contour curve and a curvature curve are extracted from edge contour curve distribution characteristics of the holes, a parameterized piecewise curve template comprising control points is constructed, the template contour is mapped onto the contour to be detected through nonlinear registration, a distance curve based on nearest neighbor search between the registration template contour and the contour curve to be detected is constructed, a conventional circle and abnormal hole normalized dirt detection curve is constructed by taking a mixed weighting curve as a weight, dirt is segmented and positioned through curve singularity detection, and a dirt discrimination cleanliness index is constructed, so that the problem of detecting the cleanliness of the universal spinneret holes is solved.
Drawings
FIG. 1 is a graph of edge profile extraction and comparison to global optimum threshold segmentation for a spinneret image in accordance with one embodiment of the present invention;
FIG. 2 is a graph of a reconstruction of the edge profile of a spinneret orifice in accordance with another embodiment of the present invention;
FIG. 3 is a histogram of conventional round-passing hole area and perimeter index values and a normal too-distribution fit thereof according to another embodiment of the present invention;
FIG. 4 is a schematic diagram of a hole standard template generation in accordance with another embodiment of the present invention;
FIG. 5 is a schematic diagram of standard template nonlinear matching in accordance with another embodiment of the present invention;
FIG. 6 is a schematic diagram of distance curve calculation according to another embodiment of the present invention;
FIG. 7 is a schematic view of a fouling detection curve according to another embodiment of the present invention;
FIG. 8 is a graph illustrating a scale detection curve segmentation according to another embodiment of the present invention;
FIG. 9 is a schematic diagram of a conventional round hole fouling detection result according to another embodiment of the present invention;
FIG. 10 is a graph showing the results of flat hole fouling detection according to another embodiment of the present invention;
FIG. 11 is a schematic illustration of a three-lobe hole soil test result according to another embodiment of the present invention;
fig. 12 is a schematic diagram of cross hole fouling detection results according to another embodiment of the present invention.
Detailed Description
The method and system for detecting the cleaning cleanliness of the universal spinneret orifice provided by the invention are further described in detail below with reference to the accompanying drawings and specific embodiments. Advantages and features of the invention will become more apparent from the following description and from the claims. It should be noted that the drawings are in a very simplified form and are all to a non-precise scale, merely for convenience and clarity in aiding in the description of embodiments of the invention.
The invention provides a general spinneret orifice cleaning cleanliness detection method and system, which aim to solve the problem of high misjudgment rate of fine dirt of the existing spinneret orifice.
The core idea of the invention is to provide a general spinneret orifice cleaning cleanliness detection method to improve detection accuracy.
In order to achieve the above-mentioned idea, the present invention provides a method and a system for detecting the cleaning cleanliness of a general spinneret orifice, wherein the method for detecting the cleaning cleanliness of the general spinneret orifice comprises: extracting an edge contour closed curve and a curvature curve of the edge contour closed curve of the standard clean hole high gradient threshold value, and constructing a parameterized segmented closed curve standard detection template containing control points; in the dirt detection stage, mapping the parameterized segmented closed curve standard detection template onto a contour to be detected through non-rigid registration; constructing a dirt detection curve, wherein the dirt detection curve is formed by weighting the parameterized segmented closed curve standard detection template, the distance curve of the contour to be detected based on nearest neighbor searching and the mixed curvature curve; and performing floor benchmark correction on the dirt detection curve to form a global threshold value so as to divide and locate dirt and quantify the cleanliness index.
Example 1
The embodiment provides a method and a system for detecting cleaning cleanliness of a universal spinneret orifice, wherein the method for detecting the cleaning cleanliness of the universal spinneret orifice comprises the following steps: extracting an edge contour closed curve and a curvature curve of the edge contour closed curve of the standard clean hole high gradient threshold value, and constructing a parameterized segmented closed curve standard detection template containing control points; in the dirt detection stage, mapping the parameterized segmented closed curve standard detection template onto a contour to be detected through non-rigid registration; constructing a dirt detection curve, wherein the dirt detection curve is formed by weighting the parameterized segmented closed curve standard detection template, the distance curve of the contour to be detected based on nearest neighbor searching and the mixed curvature curve; and performing floor benchmark correction on the dirt detection curve to form a global threshold value so as to divide and locate dirt and quantify the cleanliness index.
Specifically, in the method for detecting the cleaning cleanliness of the universal spinneret orifice, the method for detecting the cleaning cleanliness of the universal spinneret orifice further comprises: extracting a standard hole profile and a curvature curve thereof through standard clean hole image edge analysis, and carrying out segmentation and fitting on a straight line segment and an arc segment of the hole profile based on a double-threshold segmentation algorithm of the curvature curve; dividing and fitting according to the straight line segment and the circular arc segment of the standard hole profile, and reconstructing a parameterized segmented closed curve standard detection template based on control points; extracting an edge contour closed curve of a high gradient threshold value of a hole to be detected, calculating the length of the edge contour closed curve of the high gradient threshold value of the hole to be detected and the area surrounded by the edge contour closed curve, detecting significant dirt, if the detection result is unqualified, detecting the next hole to be detected, otherwise, detecting non-significant dirt; extracting an edge contour closed curve of a low gradient threshold value of a hole to be detected, mapping the edge contour closed curve of a high gradient threshold value of the standard clean hole to the edge contour closed curve of the high gradient threshold value of the hole to be detected through non-rigid registration, constructing a nearest neighbor search-based distance curve between the edge contour closed curve of the high gradient threshold value of the standard clean hole and the edge contour curve of the low gradient threshold value of the hole to be detected, and constructing a dirt detection curve by taking a mixed weighting curve as a weight; positioning and quantifying dirt by using global threshold segmentation through floor reference calibration processing of the dirt detection curve, constructing a cleanliness index for dirt discrimination, and judging whether hole dirt detection is qualified or not; round holes, flat holes, trefoil holes and cross holes are used as research objects, and result analysis and conclusion are given.
Further, in the general spinneret orifice cleaning cleanliness detection method, extracting the edge profile closed curve of the standard cleaning orifice high gradient threshold value includes: extracting a hole edge profile curve by adopting a Canny edge operator, respectively detecting strong and weak edges by adopting a double-gradient threshold method, and only retaining the weak edges connected with the strong edges; setting a first gradient threshold sigma thr1 And a second gradient threshold sigma thr2 The strong contour edge and the strong and weak combined contour edge are respectively extracted, the first gradient threshold value is smaller than the second gradient threshold value, and the first gradient threshold value and the second gradient threshold value respectively correspond to Canny detection double-ladder threshold value [ sigma ] thr1 ,3*σ thr1 ][ sigma ] thr2 ,3*σ thr2 ]The method comprises the steps of carrying out a first treatment on the surface of the The extracted image edges are a first edge and a second edge respectively, wherein the first edge is used for constructing parameterized segmented closed curve standard detection and non-rigid registration, and significant dirt detection, and the second edge is used for non-significant dirt detection. The contour curve extraction comprises contour edge extraction, reconstruction of a closed curve, significant dirt hole prefiltering and hole contour standard template generation.
FIG. 1 is a graph of edge profile extraction and comparison to global optimum threshold segmentation for the spinneret image of this embodiment. (a) to (c): clean round holes and non-significant dirt and significant dirt images thereof; (d) to (f): clean trilobal wells and non-significant soil and significant soil images. Lines 1 to 4 are original images and optimal threshold segmentation images E respectively 2 E 1 The method comprises the steps of carrying out a first treatment on the surface of the 5 th behavior optimal threshold segmentation contour E 2 Is denoted by o and x, respectively.
Contour edge extraction includes: clean hole image contour edge is smooth, dirt area gray scale in non-clean hole is along with dirt size and exit holeThe distance varies to a certain extent. It is difficult to find a suitable threshold to meet various dirty region segmentation requirements using a global thresholding method, and the up-and-down fluctuation of the threshold can cause local drift of the segmented contour edges. Compared with a global threshold method, the edge detection algorithm can optimally position the edge position of the hole profile according to the first-order gradient maximization and the second-order gradient zero crossing point characteristics. The method uses the Canny edge operator to extract the hole edge profile curve, and is particularly suitable for extracting the profile edge of a dirt area because the method adopts a double-gradient threshold method to detect strong and weak edges respectively and only keeps the weak edges connected with the strong edges. Here two gradient thresholds σ are set thr1 Sum sigma thr2 To extract strong contour edges and strong + weak contour edges (sigma thr2 <σ thr1 ) The corresponding channel detection double ladder threshold values are [ sigma ] respectively thr1 ,3*σ thr1 ][ sigma ] thr2 ,3*σ thr2 ]Extracted image edges E respectively 1 And E is 2 Its E 1 E for subsequent template construction, registration, and significant fouling detection 2 As a subsequent non-significant soil test. Figure 1 shows the contour edge positioning of same-gauge circular holes and same-gauge trefoil holes containing no, non-significant, and significant fouling compared to the optimal global threshold segmentation results. As can be seen from fig. 1 (b), the globally optimal threshold is not suitable for segmentation of non-significant fouling. As can be seen from the partial edge blending signature of line 5 of FIG. 1, the binary edge E of the cleaning aperture 2 The position deviation is smaller, but the dirt area binary edge E in the dirt hole 2 The positional deviation is large, particularly in the region of small dirt. Although the optimal thresholding result is close to the actual contour edge, there is still a pixel level difference between the two, which is clearly not satisfactory for high-precision microscopy applications. The pure global optimum thresholding method is not suitable for hole contour edge extraction.
FIG. 2 is a graph of a reconstruction of the edge profile of a spinneret orifice. Fig. 2 (a) - (f) correspond to the reconstructed edge profile curves of fig. 1 (a) - (f). Reconstructing the closed curve includes: in general E 1 May be discontinuous, or have bifurcation, or have isolated noise, straight between edge pointsThe closed contour curve cannot be extracted by the edge tracking algorithm. To ensure E 1 The mesoporous edge distribution has contour sealing characteristics, edge trimming and connection are required, and the algorithm flow comprises the following steps: (1) Refining based on 8 neighborhood and searching edge crossing points, and breaking the edge crossing points; (2) Searching curve end points, tracking the contour track, counting the length, and if the length is smaller than a set threshold value, reversely tracking along the end points and erasing the track; (3) Searching the longest curve segment, marking the starting point and the end point position of the longest curve segment, searching the starting point of the nearest adjacent curve segment from the end point position, marking the searched curve segment, and connecting the end point-starting point pair by using a straight line; (3) And continuing to search unmarked nearest neighbor curve segments from the last adjacent curve segment end point, continuing to connect and iterate in a circulating way until the searched start point is the start point of the initial longest line segment, and ending the iteration. Fig. 2 (a) to 2 (f) respectively correspond to the contour edges of fig. 1 (a) to 1 (f) to reconstruct closed curves.
FIG. 3 is a histogram of index values of the area and perimeter of a conventional round-passing hole and a normal too-distribution fit thereof. Left: an area index; right: perimeter index. Significant soil pore prefiltering includes: set E 1 Reconstructed closed curveIt->For the coordinates m of the contour points 1 Points for the outline. Gauge L 1 The indexes of the circumferential P and the filling surface S of the hole pattern are counted, the indexes of qualified and clean N holes (N > 1024) in the same hole pattern are counted, and the average mu is calculated p 、μ s Sum standard sigma p 、σ s Index, taking the optimal threshold value as tau p =μ p ±α*σ p 、τ s =μ s ±α*σ s I.e. |P-. Mu. p |≤α*σ p And |S-. Mu. s |≤α*σ s And when the cleanliness is qualified, entering the subsequent non-obvious dirt detection, and otherwise, directly carrying out the next hole detection. In practical applications, a=3.5. FIG. 3 shows a histogram of conventional circle-passing spinneret hole area and circumference detection index and its positive signThe profile fit curve, spinneret specification was pra_0094 (outer diameter/mm) -144 (number of holes) -0.18 (pore diameter/mm) x0.54 (depth of holes/mm). Fitting parameters in the graph are μ s =0.025503,σ s =0.000305,μ p =0.599034,σ p =0.004071。
The hole profile standard template generation includes: as can be seen from fig. 3, machining errors result in some variability in the shape of the orifice profile. If the closed contour curve of fig. 2 is used as a template to directly perform a least squares based rigid match, the local match bias due to the discretization of the local shape of the normal hole contour can easily lead to false identification of subsequent fouling detection. Therefore, the invention adopts parameterized straight line segment and circular arc segment piecewise curve model based on control point positioning to carry out nonlinear registration of the contour to be detected, and the error is reduced to the minimum. The template construction process comprises the following steps: calculating a closed contour curvature curve, setting a proper double threshold value according to curvature distribution characteristics, dividing the contour curve into a plurality of straight lines and large arc sections, automatically positioning each sectional control point position, decomposing the closed curve into straight line sections and arc sections (most of the closed curve is in a straight line and arc structure when the spinneret orifice is designed), performing corresponding straight line or arc parameter fitting on each divided line section by adopting a least square method, calculating the intersection point of each adjacent line section after fitting to serve as a new control point, and finally reconstructing the whole new parameterized sectional closed curve to serve as a standard matching template.
FIG. 4 is a schematic diagram of well standard template generation. (a): a curvature calculation model; (b) to (e): FIG. 2 (b) is a schematic diagram of a closed curve of a reconstructed template after contour curvature distribution curve, segmentation of circular arc segments and least square fitting of the circular arc segments; (e) to (i): fig. 2 (e) is a schematic diagram of profile curvature distribution curve, straight line segment and circular arc segment segmentation, control point positioning, and template closed curve after reconstruction of the result of least squares fitting of straight line segment and circular arc segment.
The hole profile curvature curve extraction includes: draw out contour closed curve L 1 Ith trackingAnd a radius range (r) 2 ) Inner furthest contour +.>And->Fitting a circle by adopting a least square algorithm, and calculating c by taking the radius of the fitted circle as the curvature i Calculate the overall profile curvature distribution c= { C i ,i∈[1,m 1 ]}. Fig. 4 (a) is a curvature calculation schematic diagram, and fig. 4 (b) and fig. 4 (e) are profile closed curve curvature distribution diagrams of fig. 1 (b) and fig. 1 (e), respectively.
The template generation includes: setting a curvature dividing threshold value c thr1 And c thr2 The length threshold of the segmentation line segment is g thr 。c i >c thr1 Marked as straight line segment, c i ≤c thr1 And c i >c thr2 And marking as a large arc section, otherwise marking as a small arc section. The control points of each divided line segment are respectively the starting position and the ending position, the starting position and the ending position of two adjacent divided line segments share the same position, and meanwhile, the arc segment is added with a midpoint control point which is positioned at the center of the arc and used for positioning the direction of the arc compared with the straight line segment. After curvature segmentation, carrying out length statistics on segmented line segments to obtain a length set G= { G i ,i∈[1,f g ]Gram of (g) i Representing the length of each segment, f g To divide the number of line segments. Searching for minimum segmentation line segment length g kmin For example g kmin <g thr And discarding the kmin segmentation line segment, wherein the common control points of the two adjacent segmentation line segments are respectively replaced by the midpoints of the current segmentation line segments. And then carrying out length statistics, minimum length judgment and fusion again, and carrying out loop iteration until all the segmentation line segments meet the conditions. If the curvature is divided into only one arc segment, which indicates that the hole pattern is a regular circle, the hole pattern is divided into two connected arc segments. Here curvature segmentation parameter c thr1 =400,c thr2 =120,g thr =40。
As can be seen from fig. 4, the curvature is larger than the straight line segment (such as the black line segment in fig. 4 (e) to 4 (f)) corresponding to the straight line threshold, the curvature is smaller than the straight line threshold and larger than the large arc segment (such as the green line segment in fig. 4 (b) to 4 (c) and 4 (e) to 4 (f)) corresponding to the arc threshold, and otherwise, the curvature is smaller than the small arc segment (such as the red line segment in fig. 4 (e) to 4 (f)). Fig. 4 (f) is a schematic diagram of distribution before segment fusion. Fig. 4 (g) is a graph showing the control point distribution of fig. 4 (f) after the straight line segments and the circular arc segments are fused. Fig. 4 (h) is a graph of the fitting result of each straight line and circular arc segment in fig. 4 (g) based on the least square parameter. Fig. 4 (i) is a graph of the template closed curve after reconstruction, which is obtained by solving the control points of adjacent intersections of each fitted straight line and the arc segment in fig. 4 (h). The template parameterized model is expressed as follows:
F={f i =(f i,x ,f i,y ),i∈[1,f n ]},P={p i =(s,e),i∈[1,f l ],s∈[1,f n ],e∈[1,f n ],s≠e},Q={q i =(s,c,e),i∈[1,f c ],s∈[1,f n ],e∈[1,f n ],s≠e,s≠c,c≠e},Wherein F is that all feature point sets P are straight line segment index sets Q are arc segment index sets +.>Fitting a curvature radius set s for the arc segment, wherein the initial characteristic point number e is an end characteristic point number c is an arc midpoint characteristic point number f n For the total number f of the whole feature points l For the total number of straight line segments f c Is the total number of arc line segments.
Further, the cleanliness detection comprises non-rigid registration, distance curve construction, dirt detection curve construction, dirt segmentation, recognition and cleanliness index generation. The non-rigid registration is to adopt a nonlinear least square parameter optimization method, adjust the position of the template line segment control F, and non-rigidly register the template reconstruction segmented closed curve to E 1 Reconstructed contour closed curve L to be detected 1 And finally, enabling the mean square error of the distance between each point of the contour curve to be detected and the template reconstruction segmentation closed curve { F, P, Q } to be minimum. Order theProjection distance function of k point of contour to be measured on straight line segment of template i>And (3) a projection distance function of the point k of the profile to be measured on the arc section of the template j. The nonlinear least squares-based optimization objective function of the template control F is as follows:
setting F for optimized template control point τ . FIG. 5 is a graph of the standard template nonlinear registration results of FIGS. 2 (b) and 2 (e) and its reconstruction curves, and the optimization method uses a ceres-solver algorithm library.
Fig. 5 is a standard template nonlinear match. (a) to (b): the template non-linear matching of fig. 2 (b) and its reconstruction curve; (c) to (d): the template non-linear matching of fig. 2 (e) and its reconstruction curve; (e) is a partial enlarged view of (d). Blue-reconstruction curve, red-contour curve to be measured.
Fig. 6 is a schematic diagram of distance curve calculation. (a): calculating a local distance curve; (b): calculating a local distance curve; (c) and (d): distance curves of fig. 1 (b) and 1 (e); (e) Is L 4 Converting into a schematic diagram of a 2-dimensional distance curve; (f) And (g) a smoothed 2d distance curve of (c) and (d). In the diagrams (b), (e), (f) and (g), the blue curve is a template reconstruction curve, and the red curve is a distance curve.
The distance profile includes: set E 2 Contour edge curve extracted (edge extracted by weak gradient edge threshold)It->For the coordinates of the contour points, m 2 Points for the outline. Set F τ Closed curve after reconstructionIt->For the coordinates of the contour points, m 3 Number of reconstructed curve points. FIG. 6 is a schematic diagram showing calculation of a distance curve in which a blue dot line L 3 Red dotted line L 2 Wherein fig. 6 (a) is an enlarged view of the partial scale of fig. 6 (b). When calculating the distance curve, m is determined first 2 ×m 3 And initialized to-999, and then search for the contour curvature L 2 Middle j point to L 3 Nearest i and distance thereofAs shown by the blue line segment in fig. 6 (a). After the search is completed, the maximum value of the i columns is searched in the moment M
Such asTable L 3 No L at point i in 2 The midpoint is searched as the nearest neighbor, and then the i columns are sequentially searched in the front and back columns for the presence +.>And->Points (j 1, i 1) and (j 2, i 2), and L 3 I to L in 2 The shortest distance k between j1 and j2 is filled in->I.e.
As shown by the green line segment in fig. 6 (a). Finally, calculating the reconstruction curve L 3 The maximum value of the distances of all the search point pairs (j, i) at the i point in (a) is defined as the point distance curve value, as shown in red in fig. 6 (b), as follows:
FIG. 6 (b) shows the result of searching the local distance curve of FIG. 6 (a), from which it can be seen that the distance curve L 4 There are a number of local disturbances due to irregular dirt edge distribution, where L 4 And (3) performing expansion treatment by adopting a linear morphological operator of 1 multiplied by 9 to filter out local tooth noise. FIG. 6 shows the distance curve calculation results of FIGS. 1 (b) and 1 (e), wherein FIG. 6 (e) L 4 And converting into a schematic diagram of a 2-dimensional distance curve, wherein the normal direction of the curve points to a closed curve area of the template.
Fig. 7 is a schematic view of a fouling detection curve, and a blue curve is a template reconstruction curve. (a) and (d): the weighting curves of fig. 1 (b) and 1 (e), respectively; (b) and (e): 2-dimensional curve profiles of (a) and (d), respectively; (c) and (f): the fouling detection curves (red) of fig. 6 (c) and 6 (e), respectively.
Constructing a fouling detection curve includes: in the spinneret machining process, the smaller the curvature, the larger the contour machining error, resulting in a reconstruction curve with larger contour errors from the original edge, as shown in fig. 5 (e). In order to suppress this error as much as possible, the distance curve L is here 4 Curvature-based weighting is performed. The curvature weight of the straight line segment is 1, and the arc segment weighting function adopts a gradient interpolation method, so that the smaller the curvature, the smaller the weight. Weighting functions of straight line segment and circular arc segment shown in formula 3 and formula 4 respectively, D thereof l Is L 3 Middle/section straight line area point set, D c Is L 3 And c sections of arc area point sets. Equation 5 is a mixed curvature weighting function, equation is a convolution operator, and the gaussian function g (i) is to smooth the coefficient jump of the curvature transition region after curve weighting, μ 1 =0,σ 1 =3.0. At the same time, certain discontinuity of curvature distribution of the control point area of the adjacent line segment is considered, based onIn the nonlinear registration of the error least square optimal principle, a matching error exists between an optimal control point and an actual curvature transition point, if certain dirt exists in a contour to be detected or a machining error is large, the error is expressed as a singular bulge (shown in fig. 5 (e)) in a certain range of the control point in a distance curve, the bulge is easily misjudged as dirt in subsequent dirt segmentation, and therefore, the singular bulge suppression is carried out on a control point area by adopting Gaussian function weighting. 6 is a single control point region Gaussian weighting function f i Control point L for f 3 Mid-position index, equation 7 is a mixture Gaussian weighting function, equation where σ 2 Standard deviation of Gaussian function, typically σ 2 =2.5. Equation 8 is a mixed weighting function of multiplying equation 5 by equation 7.
Dirt detection curve L weighted by distance curve 5 As shown in equation 9.
L 5 ={W(i)=W t (i)·D(i),i∈[1,m 3 ]} (9)
Fig. 7 is a graph of the fouling detection curve calculation process of fig. 6 (c) and 6 (e), wherein the blue curve in fig. 7 (b) and 7 (e) is a template contour after nonlinear registration, and the distance from each point on the curve to the red curve along the normal of the curve is a mixing weighting coefficient thereof. As can be seen from a comparison of fig. 7 (f) with fig. 6 (g), the non-linear registration error in the fouling detection curve is substantially suppressed after the control point and the low curvature region are subjected to the hybrid weighting. Fig. 7 (c) is substantially no different from fig. 6 (f) because the mixing weighting effect of the round holes is almost negligible.
Fig. 8 is a schematic view of a scale detection curve segmentation. (a) and (b): the curve floor reference estimates and the calibrated curves of fig. 7 (c), respectively; (d) and (e): the curve floor reference estimates and the calibrated curves of fig. 7 (f), respectively; (c) and (f): 2-dimensional fouling detection curves (red) for (b) and (e), respectively; (g): scale threshold segmentation and parameter definition schematic diagrams; (h) to (i): a soil segmentation and localization result map of (c) and (f).
Soil segmentation and identification include: in general, the clean hole profile edge distance curve and its fouling detection curve are theoretically approximately smooth, and the presence of local fouling appears as a singular protrusion on the detection curve, which can be segmented and localized by global thresholding. However, due to the factors such as deformation of the plate and abrasion of the orifice after the spinneret plate is used more frequently, local distortion occurs at the edge of the hole profile, and the irregularities of the floor reference are represented on the dirt detection curve (as shown in fig. 8 (a) and 8 (d)), and the irregularities cause that the global thresholding method is not suitable for the division and positioning of dirt, so that the tophat transformation is adopted to eliminate the non-uniformity of the floor reference. Fig. 8 shows the reference correction and segmentation localization process of the fouling detection curve, with the fouling global threshold set to 2.5. From the reference calibration results of fig. 8 (b) and 8 (e), the reference unevenness of the fouling detection curve is well eliminated. Fig. 8 (g) is a schematic view of scale division, wherein W, H and S parameters represent the width, height and area of the scale, respectively. In the scale division and localization marker diagrams of fig. 8 (H) and 8 (i), scale localization is shown as a black line frame in the figure, and local scale parameters are shown as (W, H, S) in the figure.
The cleanliness index generation includes: let i dirt fingers P i Is { W i ,H i ,S i },i∈[1,n]N is the number of local dirt. In spinning production, some sharp dirt causes yarn floating phenomenon, so that the weight of the sharp dirt must be increased when the cleanliness index is generated. The overall cleanliness index of the spinneret orifice is calculated as follows:
h is as follows thr =15,w thr =4. In actual detection, a cleanliness threshold P meeting the process requirements is set thr ,P>P thr And when the hole is judged to be unqualified, automatic hole blowing is firstly carried out, then the hole is re-detected, and if the hole is still unqualified, manual cleaning is carried out. The pore cleanliness indexes of FIGS. 8 (h) and 8 (i) are shown as corresponding pore center values.
The soil detection results were analyzed as follows:
fig. 9 is a schematic diagram of a conventional round hole fouling detection result. (a) to (b): fig. 1 (a) is a graph of the distance curve and the fouling detection result; (c) to (d): FIG. 1 (c) is a graph of the distance profile and the fouling detection result; (e): and part of conventional round hole dirt detection results are marked as 'v' if the qualified result is marked as 'x' if the qualified result is not marked as 'x'.
Fig. 10 is a schematic diagram of the flat hole fouling detection results. (a) to (d): a flat hole image, a nonlinear registration of a standard template, a dirt detection curve and a dirt detection result diagram; (e): and part of the flat hole dirt detection results are marked as 'v' if the qualified result is marked as 'x' if the qualified result is not marked as 'x'.
FIG. 11 is a schematic illustration of a three-leaf well fouling assay. (a) to (d): the standard template nonlinear registration, scale detection curve and scale detection result graph of fig. 1 (f); (d): and part of the three-leaf hole dirt detection results are marked as 'V' if the qualified result is marked as 'x' if the qualified result is not marked as 'x'.
Fig. 12 is a schematic diagram of cross-hole fouling detection results. (a) to (d): the method comprises the steps of non-linear registration of a cross hole image and a standard template, a dirt detection curve and a dirt detection result diagram; (e): and part of the cross hole dirt detection results are marked as 'v' if the qualified result is marked as 'x' if the qualified result is not marked as 'x'. Fig. 9 is a schematic diagram of the fouling detection result of a conventional round hole. For conventional round holes, the curvature L is closed due to the reconstruction of the template 3 The curvature difference of the two sections of circular arcs is very small, and the curve is smooth, so that the generation of the dirt detection curve is hardly interfered by the matching error of the characteristic points, and the dirt detection curve after the calibration of the floor reference can correctly reflect the dirt singularity distribution characteristics. As can be seen from fig. 9, this detection can meet the soil segmentation and localization requirements of different occlusion levels. FIG. 9 (e) P thr =5, the pass holes are marked with 'v' at the lower right corner of the image, whereas the cleanliness index is marked with 'x' at the upper left corner of the image, red value, and the same applies.
Fig. 10 to 12 are schematic diagrams showing the results of dirt detection of flat holes, trefoil holes and cross holes in a common special-shaped hole. From the figure, the template nonlinear registration process can accurately relocate the control point to the target position of the contour to be measured for the weak and obvious dirt hole. On the dirt detection curve, whether the dirt is located at the edge of the straight line segment or the edge of the circular arc segment, the dirt can be correctly segmented and positioned. If critical errors occur in the process of pre-screening the significant dirt holes, the significant dirt holes flow into the subsequent detection stage, and the control point positioning optimization fails in the nonlinear registration process because the edge distribution of the contour to be detected and the local difference of the contour of the template are too large. However, even so, the distance curve between the reconstructed closed curve contour and the contour to be measured is still enough to reflect the abnormal protrusion of the actual dirt contour, so that the result only affects the quantitative value of the cleanliness index, and the qualified hole discrimination based on the cleanliness index is not affected. Meanwhile, under the normal condition, the through section of the special-shaped hole is larger than the size of the conventional circular through section, and the probability of obviously blocking the hole after ultrasonic cleaning is very low, so that the probability of occurrence of the positioning failure in actual detection is negligible.
The detection of the cleaning cleanliness of the spinneret orifices is an important quality monitoring link of chemical fiber production enterprises. Weak and obvious dirt detection, especially abnormal hole weak and obvious dirt detection, has been a technical problem in full-automatic microscopic instrument software. According to the embodiment, the actual dirt detection results of the conventional round, flat wire, three-leaf hole and cross hole are provided, so that the general spinneret hole cleaning cleanliness detection method provided by the invention can meet the spinneret hole cleanliness detection requirements of chemical fiber enterprises.
In summary, the foregoing embodiments describe in detail different configurations of the method and system for detecting the cleaning cleanliness of a general-purpose spinneret orifice, however, the present invention includes, but is not limited to, the configurations listed in the foregoing embodiments, and any configuration that is changed based on the configurations provided in the foregoing embodiments falls within the scope of protection of the present invention. One skilled in the art can recognize that the above embodiments are illustrative.
< example two >
The embodiment provides a general spinneret orifice washs cleanliness detection system, general spinneret orifice washs cleanliness detection system includes closed curve extraction module, template mapping module, dirt detection curve module and benchmark correction module, wherein: the closed curve extraction module is configured to extract an edge contour closed curve of a standard clean hole high gradient threshold value and a curvature curve thereof, and construct a parameterized segmented closed curve standard detection template containing control points; in a dirt detection stage, the template mapping module is configured to map the parameterized piecewise closed curve standard detection template onto a contour to be detected through non-rigid registration; the dirt detection curve module is configured to construct a dirt detection curve, wherein the dirt detection curve is formed by weighting the parameterized segmented closed curve standard detection template, the distance curve of the profile to be detected based on nearest neighbor search and the mixed curvature curve; the reference correction module is configured to form a global threshold by floor reference correction of the soil detection curve to segment and locate soil and quantify a cleanliness index.
In the method and the system for detecting the cleaning cleanliness of the universal spinneret holes, tiny dirt characteristic research is taken as a starting point, a clean hole contour curve and a curvature curve are extracted from edge contour curve distribution characteristics of the holes, a parameterized piecewise curve template comprising control points is constructed, the template contour is mapped onto the contour to be detected through nonlinear registration, a distance curve based on nearest neighbor search between the registration template contour and the contour curve to be detected is constructed, a conventional circle and abnormal hole normalized dirt detection curve is constructed by taking a mixed weighting curve as a weight, dirt is segmented and positioned through curve singularity detection, and a dirt discrimination cleanliness index is constructed, so that the problem of detecting the cleanliness of the universal spinneret holes is solved.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, the description is relatively simple because of corresponding to the method disclosed in the embodiment, and the relevant points refer to the description of the method section.
The above description is only illustrative of the preferred embodiments of the present invention and is not intended to limit the scope of the present invention, and any alterations and modifications made by those skilled in the art based on the above disclosure shall fall within the scope of the appended claims.
Claims (8)
1. The utility model provides a general spinneret orifice washs cleanliness detection method which is characterized in that, general spinneret orifice washs cleanliness detection method includes:
extracting an edge contour closed curve and a curvature curve of the edge contour closed curve of the standard clean hole high gradient threshold value, and constructing a parameterized segmented closed curve standard detection template containing control points;
in the dirt detection stage, mapping the parameterized segmented closed curve standard detection template onto a contour to be detected through non-rigid registration;
constructing a dirt detection curve, wherein the dirt detection curve is formed by weighting the parameterized segmented closed curve standard detection template, the distance curve of the contour to be detected based on nearest neighbor searching and the mixed curvature curve;
performing floor benchmark correction on the dirt detection curve to form a global threshold value so as to divide and locate dirt and quantify a cleanliness index;
reconstructing a parameterized piecewise closed curve standard detection template based on control points comprises:
Refining based on 8 neighborhood and searching edge crossing points, and breaking the edge crossing points;
searching curve end points, tracking the contour track, counting the length, and if the length is smaller than a set threshold value, reversely tracking along the end points and erasing the track;
searching the longest curve segment, marking the starting point and the end point position of the longest curve segment, searching the starting point of the nearest adjacent curve segment from the end point position, marking the searched curve segment, and connecting the end point-starting point pair by using a straight line;
continuously searching unmarked nearest neighbor curve segments from the last nearest neighbor curve segment end point, continuously connecting and iterating in a circulating way until the searched start point is the start point of the initial longest line segment, and ending the iteration;
the non-rigid registration includes:
the method comprises the steps of adopting a nonlinear least square parameter optimization method, regulating the position of a line segment control F of a parameterized piecewise closed curve standard detection template, carrying out non-rigid registration on the parameterized piecewise closed curve standard detection template to an edge contour closed curve of a high gradient threshold value of a hole to be detected, which is reconstructed by a first edge, so that the distance mean square error of each point of the contour curve to be detected projected to a template reconstruction piecewise closed curve { F, P, Q } is minimum;
is provided withFor the F point of the contour to be measured, the projection distance function of straight line segment of the parameterized segmented closed curve standard detection template i is +. >Standard detection of parameterized segmented closed curve for k point of contour to be detectedA projection distance function of the arc section of the template j; the nonlinear least square optimization objective function based on the parameterized piecewise closed curve standard detection template control F is as follows:
setting F for optimized template control point τ The optimization method adopts a ceres-solver algorithm library, wherein P is a straight line segment index set, and Q is an arc segment index set.
2. The universal spinneret orifice cleaning cleanliness detection method as claimed in claim 1, further comprising:
extracting a standard hole profile and a curvature curve thereof through standard clean hole image edge analysis, and carrying out segmentation and fitting on a straight line segment and an arc segment of the hole profile based on a double-threshold segmentation algorithm of the curvature curve;
dividing and fitting according to the straight line segment and the circular arc segment of the standard hole profile, and reconstructing a parameterized segmented closed curve standard detection template based on control points;
extracting an edge contour closed curve of a high gradient threshold value of a hole to be detected, calculating the length of the edge contour closed curve of the high gradient threshold value of the hole to be detected and the area surrounded by the edge contour closed curve, detecting significant dirt, if the detection result is unqualified, detecting the next hole to be detected, otherwise, detecting non-significant dirt;
Extracting an edge contour closed curve of a low gradient threshold value of a hole to be detected, mapping the edge contour closed curve of a high gradient threshold value of the standard clean hole to the edge contour closed curve of the high gradient threshold value of the hole to be detected through non-rigid registration, constructing a nearest neighbor search-based distance curve between the edge contour closed curve of the high gradient threshold value of the standard clean hole and the edge contour curve of the low gradient threshold value of the hole to be detected, and constructing a dirt detection curve by taking a mixed weighting curve as a weight;
positioning and quantifying dirt by using global threshold segmentation through floor reference calibration processing of the dirt detection curve, constructing a cleanliness index for dirt discrimination, and judging whether hole dirt detection is qualified or not; round holes, flat holes, trefoil holes and cross holes are used as research objects, and result analysis and conclusion are given.
3. The method of claim 2, wherein extracting the edge profile closed curve for the standard clean hole height gradient threshold comprises:
extracting a hole edge profile curve by adopting a Canny edge operator, respectively detecting strong and weak edges by adopting a double-gradient threshold method, and only retaining the weak edges connected with the strong edges;
Setting a first gradient threshold sigma thr1 And a second gradient threshold sigma thr2 The strong contour edge and the strong and weak combined contour edge are respectively extracted, the first gradient threshold value is smaller than the second gradient threshold value, and the first gradient threshold value and the second gradient threshold value respectively correspond to Canny detection double-ladder threshold value [ sigma ] thr1 ,3*σ thr1 ][ sigma ] thr2 ,3*σ thr2 ];
The extracted image edges are a first edge and a second edge respectively, wherein the first edge is used for constructing parameterized segmented closed curve standard detection and non-rigid registration, and significant dirt detection, and the second edge is used for non-significant dirt detection.
4. The method for detecting the cleaning cleanliness of a universal spinneret orifice according to claim 3, wherein reconstructing an edge profile closed curve of the high gradient threshold of the orifice to be detected from the first edge is as follows:
which is a kind ofFor the coordinates m of the contour points 1 Counting the number of the contour points;
gauge L 1 The indexes of the circumferential P and the filling surface S of the steel plate are counted, the indexes of the qualified and clean N holes N > 1024 of the same-hole type machining are counted, and the average mu is calculated p 、μ s Sum standard sigma p 、σ s Index, taking the optimal threshold value as tau p =μ p ±α*σ p 、τ s =μ s ±α*σ s I.e. |P-. Mu. p |≤α*σ p And |S-. Mu. s |≤α*σ s And if the cleanliness is qualified, entering the subsequent non-obvious dirt detection, otherwise, directly carrying out the next hole detection alpha=3.5.
5. The universal spinneret orifice cleaning cleanliness detection method as claimed in claim 4, further comprising:
The nonlinear registration of the contour to be detected is carried out by adopting parameterized straight line segment and circular arc segment piecewise curve models based on control point positioning, and the nonlinear registration specifically comprises the following steps: calculating a closed contour curvature curve, setting a proper double threshold value according to curvature distribution characteristics, dividing the contour curve into a plurality of straight lines and large arc sections, automatically positioning each sectional control point position, decomposing the closed curve into straight line sections and arc sections which are connected end to end, fitting each divided line section by adopting a least square method according to corresponding straight line or arc parameters, calculating the intersection point of each adjacent line section after fitting as a new control point, and reconstructing the whole new parameterized sectional closed curve as a standard matching template.
6. The universal spinneret orifice cleaning cleanliness detection method as claimed in claim 5, further comprising:
extracting an edge contour closed curve L of the high gradient threshold value of the hole to be detected 1 Ith trackingAnd a certain radius range r before and after 2 Inner furthest contour +.>And->Fitting a circle by adopting a least square algorithm, and calculating c by taking the radius of the fitted circle as the curvature i Gauge L 1 The overall profile curvature distribution of (C = { C) i ,i∈[1,m 1 ]};/>
Setting a curvature dividing threshold value c thr1 And c thr2 Segment length threshold g thr ;c i >c thr1 Marked as straight line segment, c i ≤c thr1 And c i >c thr2 Marking as a large arc section, otherwise marking as a small arc section;
the control points of each divided line segment are respectively the starting position and the ending position, the starting position and the ending position of two adjacent divided line segments share the same position, and meanwhile, the arc segment is added with a midpoint control point which is positioned at the center of the arc and used for positioning the direction of the arc compared with the straight line segment;
after curvature segmentation, carrying out length statistics on segmented line segments to obtain a length set G= { G i ,i∈[1,f g ]Gram of (g) i Representing the length f of each segment g The number of the segments is divided;
searching for minimum segmentation line segment length g kmin For example g kmin <g thr The kmin segment is discarded,
the common control points of two adjacent segmentation line segments are respectively replaced by the midpoints of the current segmentation line segments;
carrying out length statistics, minimum length judgment and fusion again, and carrying out loop iteration until all the segmentation line segments meet the conditions; if only one arc segment is obtained after the curvature is divided, the hole pattern is a conventional circle, and the hole pattern is divided into two connected arc segments; here the curvature segmentation parameter takes c thr1 =400,c thr2 =120,g thr =40;
The parameterized piecewise closed curve standard detection template is represented as follows:
F={f i =(f i,x ,f i,y ),i∈[1,f n ]},P={p i =(s,e),i∈[1,f l ],s∈[1,f n ],e∈[1,f n ],s≠e},
Q={q i =(s,c,e),i∈[1,f c ],s∈[1,f n ],e∈[1,fn],s≠e,s≠c,c≠e},
wherein F is all characteristic point sets, P is straight line segment index set, Q is arc segment index set, Fitting a curvature radius set for the arc segments, wherein s is the number of the initial characteristic points, e is the number of the final characteristic points, c is the number of the midpoint characteristic points of the arc, and f n F is the total number of the whole feature points l F is the total number of straight line segments c Is the total number of arc line segments.
7. The method for detecting the cleaning cleanliness of a universal spinneret orifice according to claim 6, wherein the edge profile closed curve of the low gradient threshold of the orifice to be detected extracted from the second edge is:
which is a kind ofFor the coordinates m of the contour points 2 Counting the number of the contour points; from F τ
The reconstructed closed curve is:
which is a kind ofFor the coordinates m of the contour points 3 The number of points of the reconstructed curve;
calculating a reconstruction curve L 3 The distance maximum value of all search points (j, i) at the point i in (i) is taken as the distance curve value of the point as follows:
the curvature weight of the straight line segment is 1, the arc segment weighting function adopts a gradient interpolation method, and the straight line segment and the arc segment weighting function are as follows:
its D l Is L 3 Middle/section straight line area point set, D c Is L 3 A middle c section arc area point set;
the blend curvature weighting function is as follows:
which is a convolution operator, a gaussian function g (i) is used to smooth the coefficient jumps in the curve weighted curvature transition region, μ 1 =0,σ 1 =3.0;
The single control point regional gaussian weighting function is as follows:
f i control point L for f 3 A position index;
The mixture gaussian weighting function is as follows:
sigma of which 2 As a Gaussian function standard deviation sigma 2 =2.5;
The hybrid curvature weighting function multiplied by the single control point region gaussian weighting function is as follows:
W t (i)=W t1 (i)·W t2 (i) (8)
dirt detection curve L weighted by distance curve 5 The following are provided:
L 5 ={W(i)=W t (i)·D(i),i∈[1,m 3 ]} (9)
the general spinneret orifice cleaning cleanliness detection method further comprises the following steps:
eliminating the non-uniformity of the floor reference by adopting tophat conversion, and setting the global dirt threshold to be 2.5;
let i dirt fingers P i Is { W i ,H i ,S i },i∈[1,n]N is the number of local dirt, and the overall cleanliness index of the spinneret orifice is calculated as follows:
h is of thr =15,w thr =4;
In actual detection, a cleanliness threshold P meeting the process requirements is set thr ,P>P thr And when the hole is judged to be unqualified, automatic hole blowing is firstly carried out, then the hole is re-detected, and if the hole is still unqualified, manual cleaning is carried out.
8. The utility model provides a general spinneret orifice washs cleanliness factor detecting system which characterized in that, general spinneret orifice washs cleanliness factor detecting system includes closed curve extraction module, template mapping module, dirt detection curve module and benchmark correction module, wherein:
the closed curve extraction module is configured to extract an edge contour closed curve of a standard clean hole high gradient threshold value and a curvature curve thereof, and construct a parameterized segmented closed curve standard detection template containing control points;
In a dirt detection stage, the template mapping module is configured to map the parameterized piecewise closed curve standard detection template onto a contour to be detected through non-rigid registration;
the dirt detection curve module is configured to construct a dirt detection curve, wherein the dirt detection curve is formed by weighting the parameterized segmented closed curve standard detection template, the distance curve of the profile to be detected based on nearest neighbor search and the mixed curvature curve;
the reference correction module is configured to form a global threshold by floor reference correction of the soil detection curve to segment and locate soil and quantify a cleanliness index.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010412672.9A CN111627053B (en) | 2020-05-15 | 2020-05-15 | Method and system for detecting cleaning cleanliness of universal spinneret orifice |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010412672.9A CN111627053B (en) | 2020-05-15 | 2020-05-15 | Method and system for detecting cleaning cleanliness of universal spinneret orifice |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111627053A CN111627053A (en) | 2020-09-04 |
CN111627053B true CN111627053B (en) | 2023-06-02 |
Family
ID=72259030
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010412672.9A Active CN111627053B (en) | 2020-05-15 | 2020-05-15 | Method and system for detecting cleaning cleanliness of universal spinneret orifice |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111627053B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113865508B (en) * | 2021-09-28 | 2023-04-07 | 南京航空航天大学 | Automatic detection device and method for through hole rate of sound lining of honeycomb sandwich composite material |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102141374A (en) * | 2010-12-23 | 2011-08-03 | 苏州天准精密技术有限公司 | Image type spinneret plate automatic detector |
CN103063159A (en) * | 2012-12-31 | 2013-04-24 | 南京信息工程大学 | Part size measurement method based on charge coupled device (CCD) |
CN105547182A (en) * | 2015-12-09 | 2016-05-04 | 中国科学院声学研究所东海研究站 | Spinneret plate detection equipment and method |
CN107830813A (en) * | 2017-09-15 | 2018-03-23 | 浙江理工大学 | The longaxones parts image mosaic and flexural deformation detection method of laser wire tag |
CN108917593A (en) * | 2018-05-14 | 2018-11-30 | 南京工业大学 | Intelligent measurement system and method based on element configuration of workpiece to be measured |
CN110175999A (en) * | 2019-05-30 | 2019-08-27 | 广东工业大学 | A kind of position and posture detection method, system and device |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110242310A1 (en) * | 2010-01-07 | 2011-10-06 | University Of Delaware | Apparatus and Method for Electrospinning Nanofibers |
US20120109399A1 (en) * | 2012-01-01 | 2012-05-03 | Bao Tran | Energy resource conservation systems and methods |
-
2020
- 2020-05-15 CN CN202010412672.9A patent/CN111627053B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102141374A (en) * | 2010-12-23 | 2011-08-03 | 苏州天准精密技术有限公司 | Image type spinneret plate automatic detector |
CN103063159A (en) * | 2012-12-31 | 2013-04-24 | 南京信息工程大学 | Part size measurement method based on charge coupled device (CCD) |
CN105547182A (en) * | 2015-12-09 | 2016-05-04 | 中国科学院声学研究所东海研究站 | Spinneret plate detection equipment and method |
CN107830813A (en) * | 2017-09-15 | 2018-03-23 | 浙江理工大学 | The longaxones parts image mosaic and flexural deformation detection method of laser wire tag |
CN108917593A (en) * | 2018-05-14 | 2018-11-30 | 南京工业大学 | Intelligent measurement system and method based on element configuration of workpiece to be measured |
CN110175999A (en) * | 2019-05-30 | 2019-08-27 | 广东工业大学 | A kind of position and posture detection method, system and device |
Non-Patent Citations (1)
Title |
---|
几种轮廓曲率估计角点检测算法研究;张世征;CNKI;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN111627053A (en) | 2020-09-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111681206B (en) | Method for detecting size of special-shaped hole of spinneret plate | |
CN109490316B (en) | Surface defect detection algorithm based on machine vision | |
CN107255641B (en) | A method of Machine Vision Detection is carried out for self-focusing lens surface defect | |
WO2020248439A1 (en) | Crown cap surface defect online inspection method employing image processing | |
CN105067638B (en) | Tire fetal membrane face character defect inspection method based on machine vision | |
CN115908411B (en) | Concrete curing quality analysis method based on visual detection | |
CN115861291B (en) | Chip circuit board production defect detection method based on machine vision | |
CN104794721B (en) | A kind of quick optic disk localization method based on multiple dimensioned spot detection | |
CN107239742B (en) | Method for calculating scale value of instrument pointer | |
CN116309600B (en) | Environment-friendly textile quality detection method based on image processing | |
CN109540925B (en) | Complex ceramic tile surface defect detection method based on difference method and local variance measurement operator | |
CN115100206B (en) | Printing defect identification method for textile with periodic pattern | |
CN114994060B (en) | Intelligent detection system and method for magnetic ring appearance defects under machine vision | |
CN117315289A (en) | Aeroengine blade contour edge detection method based on image processing | |
CN116358449A (en) | Aircraft rivet concave-convex amount measuring method based on binocular surface structured light | |
CN111627053B (en) | Method and system for detecting cleaning cleanliness of universal spinneret orifice | |
CN109359653B (en) | Cotton leaf adhesion lesion image segmentation method and system | |
CN115880280A (en) | Detection method for quality of steel structure weld joint | |
CN118365635B (en) | Visual inspection method and system for surface defects of packaging film | |
CN106767425A (en) | A kind of vision measuring method of bearing snap spring gap | |
CN112396580B (en) | Method for detecting defects of round part | |
CN117496247A (en) | Method for recognizing shortened image of pathological morphological feature crypt of inflammatory bowel disease | |
CN113080843B (en) | Meibomian gland image-based gland extraction method and quantitative analysis method | |
CN116228776A (en) | Electromechanical equipment welding defect identification method and system | |
CN111784634B (en) | Corner detection method based on Harris-CPDA |
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 |