CN109492544A - A method of classified by enhancing optical microscopy to animal origin - Google Patents
A method of classified by enhancing optical microscopy to animal origin Download PDFInfo
- Publication number
- CN109492544A CN109492544A CN201811224702.2A CN201811224702A CN109492544A CN 109492544 A CN109492544 A CN 109492544A CN 201811224702 A CN201811224702 A CN 201811224702A CN 109492544 A CN109492544 A CN 109492544A
- Authority
- CN
- China
- Prior art keywords
- fiber
- image
- measured
- several
- classified
- 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
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000000399 optical microscopy Methods 0.000 title claims abstract description 60
- 230000002708 enhancing effect Effects 0.000 title claims abstract description 57
- 239000000835 fiber Substances 0.000 claims abstract description 167
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 30
- 230000003628 erosive effect Effects 0.000 claims abstract description 9
- 238000002203 pretreatment Methods 0.000 claims abstract description 9
- 238000005259 measurement Methods 0.000 claims description 15
- 230000000007 visual effect Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 4
- 230000001537 neural effect Effects 0.000 claims description 4
- 238000005260 corrosion Methods 0.000 claims description 3
- 230000007797 corrosion Effects 0.000 claims description 3
- 210000000988 bone and bone Anatomy 0.000 claims description 2
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 210000002268 wool Anatomy 0.000 abstract description 4
- 239000000203 mixture Substances 0.000 abstract description 3
- 238000004458 analytical method Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 abstract description 2
- 230000004927 fusion Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 1
- 239000000571 coke Substances 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001889 high-resolution electron micrograph Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/698—Matching; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/695—Preprocessing, e.g. image segmentation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Image Analysis (AREA)
Abstract
The method that the invention discloses a kind of to classify to animal origin by enhancing optical microscopy, comprising: select clear image from several focus storehouse images；Fusion output weighted image；Binaryzation is carried out to weighted image；Identify fiber to be measured；Exposure mask is established for each fiber to be measured in bianry image；Erosion algorithm is executed to exposure mask, until retaining the wide skeleton of 1 pixel；The skeleton without brachyplast is extracted from several skeletons, collects coordinate and is sorted；Make skeleton by profile changeover straight line, and the pixel of fiber to be measured is repositioned on skeleton, obtains fibre image；Pre-treatment step is carried out to fibre image and several fiber segments input convolutional neural networks are classified and exported, and/or several fibre images are inputted into half heuritic approach module classification and are exported.The present invention combines the high-resolution pictures of optical microscopy with analysis image procossing, deep learning, automatically or semi-automatically to detect and the mixture for wool fiber of classifying.
Description
Technical field
The present invention relates to the technical fields of Fibre sorting, more particularly to a kind of enhancing optical microscopy that passes through is to animal origin
The method classified.
Background technique
Stringent supervision of the current natural fiber classification workflow by iso standard and the Chinese Industrial Standards (CIS) of extension.In order to
Distinguish the source (such as the suede from wool or suede from yak) of a kind of natural fiber and another natural fiber, step
Sequence Image Acquisition, measurement fibre diameter are usually carried out by human expert and identifies whole arrangements of scale and carries out.This is
One time-consuming task needs the professional knowledge and the spirit of utter devotion of expert.Especially for fibre blend, need with higher
Precision is identified.
The fiber of some high quality is only difficult to differentiate between by optical microscopy, it is therefore desirable to use scanning electron microscope
(SEM), and corresponding expert is seeked advice to obtain determining answer, this needs the longer time
In fact, having occurred the method for carrying out natural fiber classification using vision at present, but swept using high-resolution
Electron micrograph image is retouched urgently to develop the method that natural fiber is classified.
Summary of the invention
In view of this, being classified by enhancing optical microscopy to animal origin the purpose of the present invention is to provide a kind of
Method.
To achieve the goals above, the technical scheme adopted by the invention is as follows:
A method of classified by enhancing optical microscopy to animal origin, wherein include:
Step S1: several focus storehouse images of fiber to be measured are provided, and are selected from several focus storehouse images
Several clear images out, the identical pixel in position is respective pixel in several clear images；
Step S2: choosing the respective pixel in each clear image, calculates the maximum in each respective pixel
The weighted sum of value and minimum value, and merge output weighted image；
Step S3: binaryzation is carried out to the weighted image, obtains bianry image；
Step S4；Identify the fiber to be measured in the bianry image；
Step S5: exposure mask is established for each fiber to be measured in the bianry image；
Step S6: being directed to each fiber to be measured, executes erosion algorithm to the exposure mask, until it is wide to retain 1 pixel
Skeleton；
Step S7: the skeleton without brachyplast is extracted from several skeletons；
Step S8: the coordinate of the skeleton without the brachyplast and sequence are collected；
Step S9: establishing a coordinate set, and the skeleton for not having brachyplast is respectively mapped to the coordinate set so that institute
It is straight line that skeleton, which is stated, by profile changeover, and according to the coordinate relationship between the pixel and the skeleton of the fiber to be measured, will
The pixel of the fiber to be measured is repositioned on the skeleton, forms the fibre image being straightened, and executes step S10.1 extremely
Step S10.2 and/or step S11.1；
Step S10.1: carrying out pre-treatment step to the fibre image, obtains several with identical height and same widths
Fiber segment；
Step S10.2: several fiber segments are inputted into convolutional neural networks, the convolutional neural networks are to the fibre
Dimension segment is classified and is exported；
Step S11.1: several fibre images are inputted into half heuritic approach module, the half heuritic approach module
The fibre image is classified and exported；
Wherein, the convolutional neural networks include it is several for detect visual characteristic convolution blocks and it is several for point
The intensive block of class, the convolutional neural networks are trained by fibre image training set and fibre image verifying collection；
Wherein, the half heuritic approach module is preset with the feature vector standard data set of fiber, and described half is heuristic
Algoritic module detects the feature vector of the fiber to be measured and compared with described eigenvector standard data set pair, to be divided
Class simultaneously exports.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein in the step S10.1
In, the pre-treatment step includes:
Step S10.11: the height of the fibre image is extended into the first specified pixel；
Step S10.12: the fibre image is cut into the part that several width are the second specified pixel, to obtain
The fiber segment that several height are the first specified pixel, width is the second specified pixel.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein described eigenvector is used
In the physical characteristic or visual characteristic of description fiber, described eigenvector includes: first for describing the bulk of fiber
Measurement group, the second measurement group of boundary for describing fiber, the third measurement group of flaky shape for describing fiber
With the fourth amount group of the color for describing fiber.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein the convolutional Neural net
Network and the half heuritic approach module are preset with several type labels respectively, convolutional neural networks and described half heuristic
Algoritic module exports the probability that the fiber to be measured is classified as a wherein type label respectively.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein the convolutional Neural net
Network and the half heuritic approach module are by the label as a result of the type label with maximum probability.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein further include:
Step S12 judges the result label and the half heuritic approach module that the convolutional neural networks export
Whether the result label of output is identical, S13 is thened follow the steps if they are the same, if not identical then follow the steps S14；
Step S13: being the result label by the Fibre sorting to be measured；
Step S14: being unidentified result by the Fibre sorting to be measured.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein calculate several cokes
The population variance of the second dervative of point storehouse image, and according to the minimum in the population variance of the second dervative of the focus storehouse image
Value selects the clear image.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein use OTSU algorithm pair
The weighted image carries out binaryzation.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein in the step S4,
By judge boundary whether meet the standard of area, the standard of aspect ratio and/or boundary to skeleton distance standard to identify
The fiber to be measured.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein in the step S5,
It inverts the bianry image and expansion algorithm is carried out to the bianry image of reversion, with each fiber to be measured to extraction
Establish the exposure mask.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein in the step S7,
Threshold value by setting the corrosion distance of the erosion algorithm removes the brachyplast, to extract the bone without the brachyplast
Frame；
Or, by only choosing long shoot, to extract the skeleton without the brachyplast.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein in the step S8,
Coordinate and the sequence of the skeleton are collected according to the connectivity of the image of the skeleton.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein in the step S9,
Coordinate in the coordinate set is the minimum range vector between the pixel and the skeleton of the fiber to be measured.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein be the convolutional Neural
The first minimum threshold is arranged in the probability of network output, if the probability of convolutional neural networks output is less than described first most
Small threshold value, then the convolutional neural networks are by the result label labeled as unidentified.
The above-mentioned method classified by enhancing optical microscopy to animal origin, wherein heuristic for described half
The second minimum threshold is arranged in the probability of algoritic module, if the probability of half heuritic approach module output is less than described the
Two minimum thresholds, then the half heuritic approach module is by the result label labeled as unidentified.
Due to using above-mentioned technology, the good effect for being allowed to have compared with prior art is the present invention:
(1) present invention combines the high-resolution pictures of optical microscopy with analysis image procossing, deep learning, so as to
The mixture of automatically or semi-automatically detection and wool fiber of classifying, passes through automatic work or semi-automatic work improves wool mixing
The speed of object classification, time efficiency and accuracy.
Detailed description of the invention
Fig. 1 is the first embodiment of the invention by enhancing the method that optical microscopy classifies to animal origin
Flow chart.
Fig. 2 is the second embodiment of the invention by enhancing the method that optical microscopy classifies to animal origin
Flow chart.
Fig. 3 is the 3rd embodiment of the invention by enhancing the method that optical microscopy classifies to animal origin
Flow chart.
Fig. 4 is the 3rd embodiment of the invention by enhancing the method that optical microscopy classifies to animal origin
Flow chart.
Fig. 5 is the fourth embodiment of the invention by enhancing the method that optical microscopy classifies to animal origin
Flow chart.
Fig. 6 is the signal of the step A1 of the method for the invention classified by enhancing optical microscopy to animal origin
Figure.
Fig. 7 is the signal of the step A1 of the method for the invention classified by enhancing optical microscopy to animal origin
Figure.
Fig. 8 is the signal of the step A2 of the method for the invention classified by enhancing optical microscopy to animal origin
Figure.
Fig. 9 is the signal of the step A2 of the method for the invention classified by enhancing optical microscopy to animal origin
Figure.
Figure 10 is showing for the step A3 of the method for the invention classified by enhancing optical microscopy to animal origin
It is intended to.
Figure 11 is the step A4 extremely step of the method for the invention classified by enhancing optical microscopy to animal origin
The schematic diagram of rapid A7.
Figure 12 is the step A8 extremely step of the method for the invention classified by enhancing optical microscopy to animal origin
The schematic diagram of rapid A9.
Specific embodiment
The present invention will be further explained below with reference to the attached drawings and specific examples, but not as the limitation of the invention.
First embodiment:
Fig. 1 is the first embodiment of the invention by enhancing the method that optical microscopy classifies to animal origin
Flow chart, Fig. 6 are the signals of the step A1 of the method for the invention classified by enhancing optical microscopy to animal origin
Figure, Fig. 7 are the schematic diagram of the step A1 of the method for the invention classified by enhancing optical microscopy to animal origin, figure
8 be the schematic diagram of the step A2 of the method for the invention classified by enhancing optical microscopy to animal origin, and Fig. 9 is this
The schematic diagram of the step A2 for the method for invention classified by enhancing optical microscopy to animal origin, Figure 10 are the present invention
The method classified by enhancing optical microscopy to animal origin step A3 schematic diagram, Figure 11 is of the invention logical
The schematic diagram of the step A4 to step A7 for the method that enhancing optical microscopy classifies to animal origin is crossed, Figure 12 is the present invention
The method classified by enhancing optical microscopy to animal origin step A8 to the schematic diagram of step A9, refer to figure
1, shown in Fig. 6 to Figure 12, classifying by enhancing optical microscopy to animal origin for the first preferred embodiment is shown
Method, comprising:
Step A1: providing several focus storehouse images of fiber to be measured, and selects clearly from several focus storehouse images
Clear image, the identical pixel in position is respective pixel in several clear images；
Step A2: choosing the respective pixel in each group of clear image, calculate maximum value in each group of respective pixel and
The weighted sum of minimum value, and merge output weighted image；
Step A3: binaryzation is carried out to weighted image, obtains bianry image；
Step A4；Identify the fiber to be measured in bianry image；
Step A5: exposure mask is established for each fiber to be measured in bianry image；
Step A6: being directed to each fiber to be measured, executes erosion algorithm to exposure mask, until retaining the wide skeleton of 1 pixel；
Step A7: the skeleton without brachyplast is extracted from several skeletons；
Step A8: the coordinate of the skeleton without brachyplast and sequence are collected；
Step A9: establishing a coordinate set, and the skeleton for not having brachyplast is respectively mapped to the coordinate set so that institute
It is straight line that skeleton, which is stated, by profile changeover, and according to the coordinate relationship between the pixel and the skeleton of the fiber to be measured, will
The pixel of the fiber to be measured is repositioned on the skeleton, forms the fibre image being straightened；
Step A10.1: pre-treatment step is carried out to fibre image, obtains several fibres with identical height and same widths
Tie up segment；
Step A10.2: several fiber segments are inputted into convolutional neural networks, convolutional neural networks divide fiber segment
Class simultaneously exports result label；
Wherein, convolutional neural networks include it is several for detect visual characteristic convolution blocks and it is several for classification
Intensive block, convolutional neural networks are trained by fibre image training set and fibre image verifying collection.
Second embodiment:
Fig. 2 is the second embodiment of the invention by enhancing the method that optical microscopy classifies to animal origin
Flow chart, it is shown in Figure 2, show carrying out by enhancing optical microscopy to animal origin for second of preferred embodiment
The method of classification, comprising:
Step B1: providing several focus storehouse images of fiber to be measured, and selects clearly from several focus storehouse images
Clear image, the identical pixel in position is respective pixel in several clear images；
Step B2: choosing the respective pixel in each group of clear image, calculate maximum value in each group of respective pixel and
The weighted sum of minimum value, and merge output weighted image；
Step B3: binaryzation is carried out to weighted image, obtains bianry image；
Step B4；Identify the fiber to be measured in bianry image；
Step B5: exposure mask is established for each fiber to be measured in bianry image；
Step B6: being directed to each fiber to be measured, executes erosion algorithm to exposure mask, until retaining the wide skeleton of 1 pixel；
Step B7: the skeleton without brachyplast is extracted from several skeletons；
Step B8: the coordinate of the skeleton without brachyplast and sequence are collected；
Step B9: establishing a coordinate set, and the skeleton for not having brachyplast is respectively mapped to the coordinate set so that institute
It is straight line that skeleton, which is stated, by profile changeover, and according to the coordinate relationship between the pixel and the skeleton of the fiber to be measured, will
The pixel of the fiber to be measured is repositioned on the skeleton, forms the fibre image being straightened；
Step B11.1: several fibre images are inputted into half heuritic approach module, half heuritic approach module is to fibrogram
As being classified and exporting result label；
Wherein, half heuritic approach module is preset with the feature vector standard data set of fiber, half heuritic approach module
Detect the feature vector of fiber to be measured and compared with feature vector standard data set pair.
3rd embodiment:
Fig. 3 is the 3rd embodiment of the invention by enhancing the method that optical microscopy classifies to animal origin
Flow chart, Fig. 4 are the 3rd embodiments of the invention by enhancing the method that optical microscopy classifies to animal origin
Flow chart refers to shown in Fig. 3, Fig. 4, show the third preferred embodiment by enhancing optical microscopy to animal origin
The method classified, comprising:
Step C1: providing several focus storehouse images of fiber to be measured, and selects clearly from several focus storehouse images
Clear image, the identical pixel in position is respective pixel in several clear images；
Step C2: choosing the respective pixel in each group of clear image, calculate maximum value in each group of respective pixel and
The weighted sum of minimum value, and merge output weighted image；
Step C3: binaryzation is carried out to weighted image, obtains bianry image；
Step C4；Identify the fiber to be measured in bianry image；
Step C5: exposure mask is established for each fiber to be measured in bianry image；
Step C6: being directed to each fiber to be measured, executes erosion algorithm to exposure mask, until retaining the wide skeleton of 1 pixel；
Step C7: the skeleton without brachyplast is extracted from several skeletons；
Step C8: the coordinate of the skeleton without brachyplast and sequence are collected；
Step C9: establishing a coordinate set, and the skeleton for not having brachyplast is respectively mapped to the coordinate set so that institute
It is straight line that skeleton, which is stated, by profile changeover, and according to the coordinate relationship between the pixel and the skeleton of the fiber to be measured, will
The pixel of the fiber to be measured is repositioned on the skeleton, forms the fibre image being straightened, and executes step C10.1 to step
Rapid C10.2 and step C11.1；
Step C10.1: pre-treatment step is carried out to fibre image, obtains several fibres with identical height and same widths
Tie up segment；
Step C10.2: several fiber segments are inputted into convolutional neural networks, convolutional neural networks divide fiber segment
Class and output label；
Step C11.1: several fibre images are inputted into half heuritic approach module, half heuritic approach module is to fibrogram
As being classified and exporting result label；
Step C12 judges the result label of convolutional neural networks output and the result mark of half heuritic approach module output
Whether label are identical, C13 thened follow the steps if they are the same, if not identical then follow the steps C14；
Step C13: being result label by Fibre sorting to be measured；
Step C14: being unidentified result by Fibre sorting to be measured.
Wherein, convolutional neural networks include it is several for detect visual characteristic convolution blocks and it is several for classification
Intensive block, convolutional neural networks are trained by fibre image training set and fibre image verifying collection；
Wherein, half heuritic approach module is preset with the feature vector standard data set of fiber, half heuritic approach module
Detect the feature vector of fiber to be measured and compared with feature vector standard data set pair.
Further, as a kind of preferred embodiment, feature vector is used to describe the physical characteristic or visual characteristic of fiber,
Feature vector include: the first measurement group for describing the bulk of fiber, the boundary for describing fiber second
The fourth amount group of measurement group, the third measurement group of flaky shape for describing fiber and the color for describing fiber.
Further, it is preset respectively as a kind of preferred embodiment, convolutional neural networks and half heuritic approach module
There are several type labels, convolutional neural networks and half heuritic approach module export fiber to be measured respectively and be classified as one type
The probability of type label.
Further, will have as a kind of preferred embodiment, convolutional neural networks and half heuritic approach module
The type label of maximum probability label as a result.
The foregoing is merely preferred embodiments of the present invention, are not intended to limit embodiments of the present invention and protection model
It encloses.
The present invention also has on the basis of the above is implemented as follows mode:
In further embodiment of the present invention, the first measurement group describes the shape and structure in large scale of fiber,
Second measurement group has been further described through the boundary of fiber and internal microstructure, third measurement group further describe
The microstructure of the scale of the inside of fiber, fourth amount group describe the color of fiber.
In further embodiment of the present invention, it is preferable that first metric can with or include at least: the middle line of fiber
Length, along fiber arc length measure fibre diameter maximum value, along fiber arc length measure fibre diameter minimum value,
The average value of fibre diameter that measures along the arc length of fiber, the absolute standard of fibre diameter be poor, by the normalized fibre of average value
The absolute standard for tieing up diameter is poor, the spatial variations of the degree of bias of fibre diameter, the kurtosis of fibre diameter and fibre diameter.
In further embodiment of the present invention, the horizontal central line along fiber uses 1D signal, can directly obtain or calculate
Obtain several features of the second measurement group.
In further embodiment of the present invention, it is preferable that second metric can with or include at least: point shape, fiber
The covariance of tangent slope at the covariance of the bottom at the top and fiber boundary on boundary, each point on fiber boundary, fiber
The scale height of the edge of average distance, fiber between the maximum value of height and the minimum value of fiber height is averaged just tiltedly
Rate, fiber edge scale height average negative slope.
In further embodiment of the present invention, the flaky shape by normalizing 1D signal measurement fibrous inside obtains third
Several features of group metric.
In further embodiment of the present invention, it is preferable that the third metric can with or include at least: it is obtained to return
One changes the power spectral density of 1D signal.
In further embodiment of the present invention, it is preferable that the fourth amount group can with or include at least: fiber exists
The maximum of the intensity in each channel of the maximum value, fiber of the intensity in each channel in CIELab system in CIELab system
The mean intensity in each channel of two values and fiber in RGB system at the half of value.
Wherein, it is two chrominance channels, the channel R, the channel G and channel B point that the channel L, which is lightness channel, the channel a and the channel b,
It Wei not red channel, green channel and blue channel.
In further embodiment of the present invention, the first minimum threshold is arranged in the probability for convolutional neural networks output, if volume
The probability of product neural network output is less than the first minimum threshold, then convolutional neural networks are by result label labeled as unidentified.
It is that the second minimum threshold is arranged in the probability of half heuritic approach module, if partly in further embodiment of the present invention
The probability of heuritic approach module output is less than the second minimum threshold, then half heuritic approach module is by result label labeled as not
Identification.
Fourth embodiment:
Fig. 5 is the fourth embodiment of the invention by enhancing the method that optical microscopy classifies to animal origin
Flow chart, shown in Figure 5, the 4th kind of preferred embodiment is on the basis of first embodiment or 3rd embodiment, into one
Step ground includes: a kind of pre-treatment step, and pre-treatment step includes:
Step D10.11: the height of fibre image is extended into the first specified pixel；
Step D10.12: fibre image is cut into the part that several width are the second specified pixel, to obtain several
Height is the first specified pixel, the fiber segment that width is the second specified pixel.
Further, as a kind of preferred embodiment, in step D10.11, by filling empty pixel for fibre image
Height extends to the first specified pixel.
Specifically, in step D10.11, the height of fibre image is passed through into filling 0 pixel-expansion to 224 pixels.
Specifically, in step D10.12, fibre image is cut into the part that several width are 384 pixels, thus
Obtain the fiber segment having a size of 224 × 384 pixels.
5th embodiment:
5th kind of preferred embodiment is on the basis of first embodiment, second embodiment or 3rd embodiment, further
Ground includes: in step s 4, by judging whether boundary meets the standard of area, the standard of aspect ratio and/or boundary to skeleton
Distance standard to identify fiber to be measured.
In addition, in step A1/B1/C1, calculating the second order of several focus storehouse images as a kind of preferred embodiment
The population variance of derivative, and clear image is selected according to the minimum value in the population variance of the second dervative of focus storehouse image.
Further, as a kind of preferred embodiment, in step A1/B1/C1, the i-th row jth of each clear image is arranged
Pixel be respective pixel, in step A2/B2/C2, maximum value and most is chosen in all respective pixels of the i-th row jth column
Small value is weighted summation, and merges output weighted image.
In addition, in step A2/B2/C2, being carried out using OTSU algorithm to weighted image as a kind of preferred embodiment
Binaryzation.
In addition, as a kind of preferred embodiment, in step A4/B4/C4, by judging whether boundary meets area
Standard, the standard of aspect ratio and/or boundary to skeleton distance standard to identify fiber to be measured, to distinguish single fiber to be measured
With the rest part of image (including straight line, curve, line segment and overlapping part).Wherein, the mark of the standard, aspect ratio of area
The standard of quasi-, boundary to skeleton distance can be specifically defined according to the actual conditions of fiber to be measured, be surrounded for finding boundary
Region area and shape, to find the boundary of fiber to be measured.
Specifically, different modes can be used and carry out optimal fiber extraction, be with total variance and Laplace operator
Example, it is fiber to be measured that aspect ratio can be used in a wherein preferred embodiment by the pattern classification in bianry image,
In another preferred embodiment, it is fiber to be measured that the radius change on boundary, which can be used, by the pattern classification in bianry image,
To take out the chaff interferent (such as noise) in the background of bianry image.
Further, as a kind of preferred embodiment, in step A5/B5/C5, bianry image is inverted and to the two of reversion
It is worth image and carries out expansion algorithm, establishes exposure mask to each fiber to be measured of extraction.
Selectively, as a kind of preferred embodiment, in step A7/B7/C7, by the corrosion for setting erosion algorithm
The threshold value of distance removes brachyplast, to extract the skeleton without brachyplast；
Selectively, as a kind of preferred embodiment, in step A7/B7/C7, by only choosing long shoot, to extract
Skeleton without brachyplast.
In further embodiment of the present invention, by define shape, length etc. of skeleton with by skeleton divide into long shoot or
Brachyplast.
In addition, in step A8/B8/C8, being collected according to the connectivity of the image of skeleton as a kind of preferred embodiment
The coordinate of skeleton and sequence.
On the other hand, as a kind of preferred embodiment, in step A9/B9/C9, the coordinate in coordinate set is fibre to be measured
Minimum range vector between the pixel and skeleton of dimension.
The above is only preferred embodiments of the present invention, are not intended to limit the implementation manners and the protection scope of the present invention, right
For those skilled in the art, it should can appreciate that and all replace with being equal made by description of the invention and diagramatic content
It changes and obviously changes obtained scheme, should all be included within the scope of the present invention.
Claims (10)
1. a kind of method classified by enhancing optical microscopy to animal origin characterized by comprising step S1:
Several focus storehouse images of fiber to be measured are provided, and select several clear images from several focus storehouse images,
The identical pixel in position is respective pixel in several clear images；Step S2: pair in each clear image is chosen
Pixel is answered, the weighted sum of the maximum value and minimum value in each respective pixel is calculated, and merges output weighted image；
Step S3: binaryzation is carried out to the weighted image, obtains bianry image；
Step S4；Identify the fiber to be measured in the bianry image；
Step S5: exposure mask is established for each fiber to be measured in the bianry image；
Step S6: being directed to each fiber to be measured, executes erosion algorithm to the exposure mask, until retaining the wide skeleton of 1 pixel；
Step S7: the skeleton without brachyplast is extracted from several skeletons；
Step S8: the coordinate of the skeleton without the brachyplast and sequence are collected；
Step S9: establishing a coordinate set, and the skeleton for not having brachyplast is respectively mapped to the coordinate set so that the bone
Frame is straight line by profile changeover, and according to the coordinate relationship between the pixel and the skeleton of the fiber to be measured, will be described
The pixel of fiber to be measured is repositioned on the skeleton, forms the fibre image being straightened, and executes step S10.1 to step
S10.2 and/or step S11.1；
Step S10.1: pre-treatment step is carried out to the fibre image, obtains several fibres with identical height and same widths
Tie up segment；
Step S10.2: several fiber segments are inputted into convolutional neural networks, the convolutional neural networks are to the fibre plate
Duan Jinhang classifies and exports；
Step S11.1: several fibre images are inputted into half heuritic approach module, the half heuritic approach module is to institute
Fibre image is stated to be classified and exported；
Wherein, the convolutional neural networks include it is several for detect visual characteristic convolution blocks and it is several for classification
Intensive block, the convolutional neural networks are trained by fibre image training set and fibre image verifying collection；
Wherein, the half heuritic approach module is preset with the feature vector standard data set of fiber, half heuritic approach
Module detects the feature vector of the fiber to be measured and compared with described eigenvector standard data set pair, to be classified simultaneously
Output.
2. the method according to claim 1 classified by enhancing optical microscopy to animal origin, feature exist
In in the step S10.1, the pre-treatment step includes:
Step S10.11: the height of the fibre image is extended into the first specified pixel；
Step S10.12: the fibre image is cut into the part that several width are the second specified pixel, to obtain several
Height is the first specified pixel, the fiber segment that width is the second specified pixel.
3. the method according to claim 2 classified by enhancing optical microscopy to animal origin, feature exist
In described eigenvector includes: the first measurement group for describing the bulk of fiber, the boundary for describing fiber
The second measurement group, the fourth amount of the third measurement group of flaky shape for describing fiber and the color for describing fiber
Group.
4. the method according to claim 3 classified by enhancing optical microscopy to animal origin, feature exist
In the convolutional neural networks and the half heuritic approach module are preset with several type labels, the convolutional Neural respectively
Network and the half heuritic approach module export the fiber to be measured respectively and are classified as the general of a wherein type label
Rate.
5. the method according to claim 4 classified by enhancing optical microscopy to animal origin, feature exist
In the convolutional neural networks and the half heuritic approach module are using the type label with maximum probability as knot
Fruit label.
6. the method according to claim 5 classified by enhancing optical microscopy to animal origin, feature exist
In, further includes:
Step S12 judges the result label and the half heuritic approach module output of the convolutional neural networks output
The result label it is whether identical, S13 is thened follow the steps if they are the same, if not identical then follow the steps S14；
Step S13: being the result label by the Fibre sorting to be measured；
Step S14: being unidentified result by the Fibre sorting to be measured.
7. the method according to claim 1 classified by enhancing optical microscopy to animal origin, feature exist
In in the step S4, by judging whether boundary meets the standard of area, the standard of aspect ratio and/or boundary to skeleton
Distance standard to identify the fiber to be measured.
8. the method according to claim 1 classified by enhancing optical microscopy to animal origin, feature exist
In, in the step S5, invert the bianry image and to the bianry image of reversion carry out expansion algorithm, to extraction
Each fiber to be measured establish the exposure mask.
9. the method according to claim 1 classified by enhancing optical microscopy to animal origin, feature exist
In in the step S7, the threshold value removal brachyplast of the corrosion distance by setting the erosion algorithm does not have to extract
There is the skeleton of the brachyplast；
Or, by only choosing long shoot, to extract the skeleton without the brachyplast.
10. the method according to claim 1 classified by enhancing optical microscopy to animal origin, feature exist
In, in the step S9, coordinate in the coordinate set be between the pixel and the skeleton of the fiber to be measured most
Small distance vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811224702.2A CN109492544B (en) | 2018-10-19 | 2018-10-19 | Method for classifying animal fibers through enhanced optical microscope |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811224702.2A CN109492544B (en) | 2018-10-19 | 2018-10-19 | Method for classifying animal fibers through enhanced optical microscope |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109492544A true CN109492544A (en) | 2019-03-19 |
| CN109492544B CN109492544B (en) | 2023-01-03 |
Family
ID=65692273
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811224702.2A Expired - Fee Related CN109492544B (en) | 2018-10-19 | 2018-10-19 | Method for classifying animal fibers through enhanced optical microscope |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109492544B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110047074A (en) * | 2019-05-17 | 2019-07-23 | 广东工业大学 | The fiber of textile mixes content detection, reverse engineering analysis method and equipment |
| CN110648312A (en) * | 2019-09-03 | 2020-01-03 | 上海工程技术大学 | Method for identifying wool and cashmere fibers based on scale morphological characteristic analysis |
| CN114067122A (en) * | 2022-01-18 | 2022-02-18 | 深圳市绿洲光生物技术有限公司 | Two-stage binarization image processing method |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101358931A (en) * | 2008-08-22 | 2009-02-04 | 北京中棉机械成套设备有限公司 | Detecting and metering device and method for foreign fibre in cotton |
| CN105095907A (en) * | 2014-05-12 | 2015-11-25 | 浙江理工大学 | Cotton foreign fiber identifying method based on RBF neural network |
| CN106780597A (en) * | 2016-08-08 | 2017-05-31 | 大连工业大学 | It is a kind of based on image procossing to the extracting method of fiber characteristics in fibre reinforced composites |
| US9691161B1 (en) * | 2015-09-25 | 2017-06-27 | A9.Com, Inc. | Material recognition for object identification |
| CN107240141A (en) * | 2017-05-19 | 2017-10-10 | 华南理工大学 | A kind of paper fibre cellulose fiber two-dimensional structure method for reconstructing based on image procossing |
| CN107909107A (en) * | 2017-11-14 | 2018-04-13 | 深圳码隆科技有限公司 | Fiber check and measure method, apparatus and electronic equipment |
| CN108038838A (en) * | 2017-11-06 | 2018-05-15 | 武汉纺织大学 | A kind of cotton fibriia species automatic testing method and system |
| CN108090498A (en) * | 2017-12-28 | 2018-05-29 | 广东工业大学 | A kind of fiber recognition method and device based on deep learning |
-
2018
- 2018-10-19 CN CN201811224702.2A patent/CN109492544B/en not_active Expired - Fee Related
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101358931A (en) * | 2008-08-22 | 2009-02-04 | 北京中棉机械成套设备有限公司 | Detecting and metering device and method for foreign fibre in cotton |
| CN105095907A (en) * | 2014-05-12 | 2015-11-25 | 浙江理工大学 | Cotton foreign fiber identifying method based on RBF neural network |
| US9691161B1 (en) * | 2015-09-25 | 2017-06-27 | A9.Com, Inc. | Material recognition for object identification |
| CN106780597A (en) * | 2016-08-08 | 2017-05-31 | 大连工业大学 | It is a kind of based on image procossing to the extracting method of fiber characteristics in fibre reinforced composites |
| CN107240141A (en) * | 2017-05-19 | 2017-10-10 | 华南理工大学 | A kind of paper fibre cellulose fiber two-dimensional structure method for reconstructing based on image procossing |
| CN108038838A (en) * | 2017-11-06 | 2018-05-15 | 武汉纺织大学 | A kind of cotton fibriia species automatic testing method and system |
| CN107909107A (en) * | 2017-11-14 | 2018-04-13 | 深圳码隆科技有限公司 | Fiber check and measure method, apparatus and electronic equipment |
| CN108090498A (en) * | 2017-12-28 | 2018-05-29 | 广东工业大学 | A kind of fiber recognition method and device based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| XINXIN WANG ET AL.: "Fiber Image Classification Using Convolutional Neural Networks", 《THE 2017 4TH INTERNATIONAL CONFERENCE ON SYSTEMS AND INFORMATICS (ICSAI 2017)》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110047074A (en) * | 2019-05-17 | 2019-07-23 | 广东工业大学 | The fiber of textile mixes content detection, reverse engineering analysis method and equipment |
| CN110648312A (en) * | 2019-09-03 | 2020-01-03 | 上海工程技术大学 | Method for identifying wool and cashmere fibers based on scale morphological characteristic analysis |
| CN114067122A (en) * | 2022-01-18 | 2022-02-18 | 深圳市绿洲光生物技术有限公司 | Two-stage binarization image processing method |
| CN114067122B (en) * | 2022-01-18 | 2022-04-08 | 深圳市绿洲光生物技术有限公司 | Two-stage binarization image processing method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109492544B (en) | 2023-01-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP4864857B2 (en) | Measurement of mitotic activity | |
| CN103914708B (en) | Food kind detection method based on machine vision and system | |
| TWI467515B (en) | Multi-color dropout for scanned document | |
| KR100303608B1 (en) | Blood cell automatic recognition method and apparatus | |
| CN101251898B (en) | Skin color detection method and apparatus | |
| CN105893925A (en) | Human hand detection method based on complexion and device | |
| CN104899871B (en) | A kind of IC elements solder joint missing solder detection method | |
| CN110189383B (en) | Traditional Chinese medicine tongue color and fur color quantitative analysis method based on machine learning | |
| CN105574514B (en) | The raw tomato automatic identifying method in greenhouse | |
| CN109492544A (en) | A method of classified by enhancing optical microscopy to animal origin | |
| CN110687121A (en) | Intelligent online detection and automatic grading method and system for ceramic tiles | |
| CN111077150A (en) | Intelligent excrement analysis method based on computer vision and neural network | |
| EP3896650A1 (en) | Quality control system for series production | |
| CN110910394B (en) | Method for measuring resolution of image module | |
| CN115761013A (en) | A Cloth Color Difference Detection Method Based on Texture Classification | |
| KR101493900B1 (en) | Video processing method for detecting whether visibility pattern of gambling card is or not | |
| CN106157301A (en) | A kind of threshold value for Image Edge-Detection is from determining method and device | |
| KR101801266B1 (en) | Method and Apparatus for image classification | |
| CN108491846A (en) | Ripening fruits machine identification method | |
| CN111881921A (en) | Optimal grain direct selection method for rice color selector | |
| CN111832392A (en) | Flame smoke detection method and device | |
| CN109543531A (en) | A kind of method of fiber extraction and fiber vision correcting | |
| CN116612331A (en) | Method, device and storage medium for automatic detection of picture quality based on image processing | |
| CN115330721A (en) | Banana fruit comb plumpness detection method and system based on shape and color information | |
| CN110378403B (en) | Wire spool classification and identification method and system |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20230103 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |