CN106897991A - Microbiologic population's recognition methods based on Live Wire segmentations - Google Patents
Microbiologic population's recognition methods based on Live Wire segmentations Download PDFInfo
- Publication number
- CN106897991A CN106897991A CN201710007720.4A CN201710007720A CN106897991A CN 106897991 A CN106897991 A CN 106897991A CN 201710007720 A CN201710007720 A CN 201710007720A CN 106897991 A CN106897991 A CN 106897991A
- Authority
- CN
- China
- Prior art keywords
- node
- image
- microbiologic population
- pixel
- formula
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10056—Microscopic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Abstract
The invention discloses a kind of microbiologic population's recognition methods based on Live Wire segmentations, first, in image segmentation part, Live Wire algorithms are used.This is a kind of interactive image segmentation algorithm, it is intended to expertise is introduced in the case that manual intervention is reduced into minimum degree, so as to obtain preferable result in complex situations.On the other hand, the process of segmentation can be completely controlled in cutting procedure due to operator, so the algorithm can be prevented effectively from over-segmentation on the premise of image information is not lost.Secondly, when partitioning portion is screened, SVMs (SVM) has been used, has can be good at solving non-linear and high-dimensional pattern-recognition, the accuracy of the interference such as pollution has been rejected so as to greatly improve.
Description
Technical field
Know the invention belongs to microorganism detection field, more particularly to a kind of microbiologic population based on Live-Wire segmentations
Other method.
Background technology
In fields such as food, medical treatment, industry, environmental protection, the Micro biological Tests to product and sample have great importance.Its
In, microbiologic population's identification is a mostly important step in Micro biological Tests.In recent years, with the hair of computer vision technique
Exhibition, the microbiologic population's recognition methods based on image procossing with its it is automatic, quick, objective the features such as to become both domestic and external one big
Study hotspot.In general, the microbiologic population's identification based on image procossing is broadly divided into image segmentation and partitioning portion screening
Two steps.At present, the method for the more commonly used segmentation microbiologic population image has dividing method based on threshold value and based on shape
Watershed algorithm of state etc..Wherein effect most preferably watershed algorithm, because the closing of its profile, registration.And for
The screening of partitioning portion then mainly passes through first to obtain the morphological feature of doubtful group, then to some parameter settings of morphological feature
Threshold value rejects the purpose of the interference such as pollution to reach.
Watershed algorithm has good response for faint edge, the trickle gray scale of the noise, body surface in image
Change, the phenomenon that can all produce over-segmentation to split.And in order to eliminate over-segmentation, generally require to carry out some special places
Reason, and this normally results in the loss of parts of images information.Additionally, watershed algorithm is automatic dividing method, lack expert
The intervention of knowledge, possibly preferable result cannot be obtained in complex situations.In terms of partitioning portion is screened, due to the form for polluting
Feature has uncertainty, and the accuracy that the interference such as pollution are rejected using threshold value is poor.
The content of the invention
The purpose of the present invention is to solve the shortcomings of the prior art, there is provided a kind of microorganism based on Live-Wire segmentations
Group's recognition methods.
The purpose of the present invention is achieved through the following technical solutions：A kind of microorganism based on Live-Wire segmentations
Group's recognition methods, comprises the following steps：
(1) microbiologic population's image is read, image is pre-processed；
(2) Live-wire segmentations are carried out to pretreated image, obtains some images fragments split：
(2.1) according to one weighted digraph of picture construction to be split；
(2.2) shortest path in weighted digraph between 2 points is searched using optimum route search algorithm as object
Border segment, carry out edge segmentation；
(3) microbiologic population that will be identified is instructed with SVM with pollution image as training set to training set
Practice, obtain microbiologic population and water pollution classification model, the images fragment obtained to step 2 using the disaggregated model for training is carried out
Classification, so as to recognize microbiologic population.
Further, described step 2.1 is specially：By the pixel in image to be split as the node in digraph,
Between adjacent pixel while as connecting node while.Define a cost function on side, its value and makes strong as the weights on side
Edge has less cost value, and non-edge has larger cost value.It is believed that connecting the cost value on the side of non-adjacent pixel
For+∞.Two cost value formula such as formula (1) of adjacent pixels：
T (p, q)=ωG×fG(q)+ωZ×fZ(q)+ωD×fD(p,q) (1)
T (p, q) is local cost of the pixel p to its adjacent pixels q in formula；ωG、ωZAnd ωDIt is weight coefficient；fG(q)、
fZ(q) and fDQ () is Gradient Features function, Laplace zero passages characteristic function and the smoothness constraints function at respective pixel q, such as
Shown in formula (2)~(4)：
G (q) and L (q) are respectively the amplitude and Laplace values of the gradient at pixel q in formula；D (q) is pixel q in image
The unit normal vector at place.To ∏ p, q has 0≤t (p, q)≤1.
Further, described step 2.2 specifically includes following sub-step：
(2.2.1) chooses start node s and mesh interested man-machine interactively on the weighted digraph that step 2.1 builds
Mark node e；
(2.2.2) makes shortest path flag node set M={ s }, unmarked node set U=A-M, interim shortest path
Flag node set T=Φ；Wherein, A is the set of image all pixels point；
(2.2.3) will be newly joined the node i of M as extended source, obtain the shortest path of all adjacent node J to s of i
Footpath, and the unmarked node in J is added in T, i.e. T=T ∪ (J ∩ U)；
(2.2.4) the node x of selection with minimum d in T, is added in M, i.e. M=M ∪ x；Simultaneously x from T
Remove, i.e. T=T-x；D is shortest path of certain point to starting point s；
(2.2.5) is if destination node e is not belonging to M, repeat step 2.2.3~2.2.4；If e belongs to M, carry out
Next step；
(2.2.6) is obtained from start node s to destination node e's from its directional information of destination node e tracebacks
Shortest path；Image is split according to shortest path；
(2.2.7) is if also have other start node s and destination node e interested, repeatedly 2.2.1~2.2.6；It is no
Then, segmentation terminates, and obtains some images fragments split.
Further, in described step 3, SVM includes to the feature that training set is trained：Area A, girth P, length
L, width W, equivalent diameter D, form factor C.Shown in the formula of equivalent diameter D and form factor C such as formula (5)~(6).
C=4 π A/P2 (6)
The beneficial effects of the invention are as follows：
1) split microbiologic population's image by Live-Wire algorithms, minimized while expertise is introduced
Manual intervention, is effectively improved the accuracy and specific aim of microbiologic population's image segmentation.
2) present invention is screened using SVM to partitioning portion, and the sample of low-dimensional input space linearly inseparable is turned
Turning to high-dimensional feature space makes its linear separability, effectively overcomes tired in the examination that brings of uncertainty of pollution shape
It is difficult.
Brief description of the drawings
Fig. 1 is microbiologic population recognition methods overall flow of the present invention based on Live-Wire segmentations；
Fig. 2 is the idiographic flow of Live-Wire segmentation figure pictures.
Specific embodiment
The present invention is described in further detail with specific embodiment below in conjunction with the accompanying drawings.
As shown in figure 1, a kind of microbiologic population's recognition methods based on Live-Wire segmentations that the present invention is provided, including
Following steps：
(1) microbiologic population's image is read, image is pre-processed, for example：The removal of culture dish edge, image gray processing
Treatment；
(2) Live-wire segmentations are carried out to pretreated image, obtains some images fragments split.Live‐
Wire algorithms are a kind of interactive image segmentation algorithms, it is intended to introduce special in the case that manual intervention is reduced into minimum degree
Family's knowledge, so as to obtain preferable result in complex situations.On the other hand, because operator can be complete in cutting procedure
The process of segmentation is controlled, so the algorithm can be prevented effectively from over-segmentation on the premise of image information is not lost, while
There is preferably performance during segmentation complexity microbiologic population's image.Specifically include following sub-step：
(2.1) according to one weighted digraph of picture construction to be split：By the pixel in image to be split as digraph
In node, between adjacent pixel while as connecting node while.On side define a cost function, its value as side power
Value, and make strong edge that there is less cost value, non-edge has larger cost value.It is believed that connecting the side of non-adjacent pixel
Cost value be+∞.Two cost value formula such as formula (1) of adjacent pixels：
T (p, q)=ωG×fG(q)+ωZ×fZ(q)+ωD×fD(p,q) (1)
T (p, q) is local cost of the pixel p to its adjacent pixels q in formula；ωG、ωZAnd ωDIt is weight coefficient；fG(q)、
fZ(q) and fDQ () is Gradient Features function, Laplace zero passages characteristic function and the smoothness constraints function at respective pixel q, such as
Shown in formula (2)~(4).
G (q) and L (q) are respectively the amplitude and Laplace values of the gradient at pixel q in formula；D (q) is pixel q in image
The unit normal vector at place.So, to ∏ p, q has 0≤t (p, q)≤1.
(2.2) shortest path in weighted digraph between 2 points is searched using optimum route search algorithm as object
Border segment, carry out edge segmentation.As shown in Fig. 2 specifically including following sub-step：
(2.2.1) chooses start node s and mesh interested man-machine interactively on the weighted digraph that step 2.1 builds
Mark node e；
(2.2.2) makes shortest path flag node set M={ s }, unmarked node set U=A-M, interim shortest path
Flag node set T=Φ；Wherein, A is the set of image all pixels point；
(2.2.3) will be newly joined the node i of M as extended source, obtain the shortest path of all adjacent node J to s of i
Footpath, and the unmarked node in J is added in T, i.e. T=T ∪ (J ∩ U)；
(2.2.4) the node x of selection with minimum d in T, is added in M, i.e. M=M ∪ x；Simultaneously x from T
Remove, i.e. T=T-x；D is shortest path of certain point to starting point s；
(2.2.5) is if destination node e is not belonging to M, repeat step 2.2.3~2.2.4；If e belongs to M, carry out
Next step；
(2.2.6) is obtained from start node s to destination node e's from its directional information of destination node e tracebacks
Shortest path；
(2.2.7) is split according to shortest path to image；
(2.2.8) is if also have other start node s and destination node e interested, repeatedly 2.2.1~2.2.7；It is no
Then, segmentation terminates, and obtains some images fragments split.
(3) microbiologic population that will be identified is instructed with SVM with pollution image as training set to training set
Practice, obtain microbiologic population and water pollution classification model, the images fragment obtained to step 2 using the disaggregated model for training is carried out
Classification, so as to recognize microbiologic population.
Due to reasons such as misoperations, a large amount of pollutions are might have in microbiologic population's image.If at this moment directly taken preceding
The image that face has been split is come if carrying out microorganism detection, it is more likely that can be pollution also to calculating in group, thus seriously
Have impact on the accuracy and reliability of result.So for the image split, in addition it is also necessary to further screened, to reject
The distracters such as pollution.The present invention uses SVM in this step.
SVM is a kind of mode identification method, and it uses structural risk minimization, takes into account training error and extensive energy
Power, many distinctive advantages are shown in small sample, dimension non-linear, high, local minimum isotype identification is solved the problems, such as.
It is directed to linear separability situation and is analyzed, during for linearly inseparable, will be low by using non-linear map
The sample of dimension input space linearly inseparable is converted into high-dimensional feature space makes its linear separability, so that high-dimensional feature space
Linear analysis is carried out using linear algorithm to the nonlinear characteristic of sample to be possibly realized.
Training set uses the objective microbe group and pollution image being identified, and test set (partitioning portion)
Each single item then in non-label state, it is necessary to be predicted using the forecast model for training.Training set and test set
Feature extraction include the following shape facility of doubtful group：Area A, girth P, length L, width W, equivalent diameter D, shape
Factor C.Shown in the formula of equivalent diameter D and form factor C such as formula (5)~(6).
C=4 π A/P2 (6)。
Claims (4)
1. a kind of microbiologic population's recognition methods based on Live-Wire segmentations, it is characterised in that comprise the following steps：
(1) microbiologic population's image is read, image is pre-processed；
(2) Live-wire segmentations are carried out to pretreated image, obtains some images fragments split：
(2.1) according to one weighted digraph of picture construction to be split；
(2.2) side of the shortest path in weighted digraph between 2 points as object is searched using optimum route search algorithm
Area under a person's administration, carries out edge segmentation；
(3) microbiologic population that will be identified is trained with SVM with pollution image as training set to training set, is obtained
To microbiologic population and water pollution classification model, the images fragment that step 2 is obtained is classified using the disaggregated model for training,
So as to recognize microbiologic population.
2. a kind of microbiologic population's recognition methods based on Live-Wire segmentations according to claim 1, its feature exists
In described step 2.1 is specially：By the pixel in image to be split as the node in digraph, the side between adjacent pixel
As the side of connecting node.Define a cost function on side, its value as side weights, and it is less to have strong edge
Cost value, non-edge has larger cost value.It is believed that the cost value for connecting the side of non-adjacent pixel is+∞.Two adjoinings
The cost value formula such as formula (1) of pixel：
T (p, q)=ωG×fG(q)+ωZ×fZ(q)+ωD×fD(p,q) (1)
T (p, q) is local cost of the pixel p to its adjacent pixels q in formula；ωG、ωZAnd ωDIt is weight coefficient；fG(q)、fZ(q)
And fDQ () is the Gradient Features function at respective pixel q, Laplace zero passages characteristic function and smoothness constraints function, such as formula
(2) shown in~(4)：
G (q) and L (q) are respectively the amplitude and Laplace values of the gradient at pixel q in formula；D (q) is in image at pixel q
Unit normal vector.To ∏ p, q has 0≤t (p, q)≤1.
3. a kind of microbiologic population's recognition methods based on Live-Wire segmentations according to claim 1, its feature exists
In described step 2.2 specifically includes following sub-step：
(2.2.1) chooses start node s and target section interested man-machine interactively on the weighted digraph that step 2.1 builds
Point e；
(2.2.2) makes shortest path flag node set M={ s }, unmarked node set U=A-M, interim shortest path mark
Node set T=Φ；Wherein, A is the set of image all pixels point；
(2.2.3) will be newly joined the node i of M as extended source, obtain the shortest path of all adjacent node J to s of i, and
Unmarked node in J is added in T, i.e. T=T ∪ (J ∩ U)；
(2.2.4) the node x of selection with minimum d in T, is added in M, i.e. M=M ∪ x；X is removed from T simultaneously,
That is T=T-x；D is shortest path of certain point to starting point s；
(2.2.5) is if destination node e is not belonging to M, repeat step 2.2.3~2.2.4；If e belongs to M, carry out next
Step；
(2.2.6) is obtained from start node s to the most short of destination node e from its directional information of destination node e tracebacks
Path；Image is split according to shortest path；
(2.2.7) is if also have other start node s and destination node e interested, repeatedly 2.2.1~2.2.6；Otherwise,
Segmentation terminates, and obtains some images fragments split.
4. a kind of microbiologic population's recognition methods based on Live-Wire segmentations according to claim 1, its feature exists
In in described step 3, SVM includes to the feature that training set is trained：It is area A, girth P, length L, width W, equivalent straight
Footpath D, form factor C.Shown in the formula of equivalent diameter D and form factor C such as formula (5)~(6).
C=4 π A/P2 (6)。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710007720.4A CN106897991A (en) | 2017-01-05 | 2017-01-05 | Microbiologic population's recognition methods based on Live Wire segmentations |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710007720.4A CN106897991A (en) | 2017-01-05 | 2017-01-05 | Microbiologic population's recognition methods based on Live Wire segmentations |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106897991A true CN106897991A (en) | 2017-06-27 |
Family
ID=59198518
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710007720.4A Pending CN106897991A (en) | 2017-01-05 | 2017-01-05 | Microbiologic population's recognition methods based on Live Wire segmentations |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106897991A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111179288A (en) * | 2019-12-20 | 2020-05-19 | 浙江理工大学 | Interactive contrast blood vessel segmentation method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105447527A (en) * | 2015-12-31 | 2016-03-30 | 四川木牛流马智能科技有限公司 | Method and system for classifying environmental microorganisms by image recognition technology |
| CN105608472A (en) * | 2015-12-31 | 2016-05-25 | 四川木牛流马智能科技有限公司 | Method and system for carrying out fully automatic classification of environmental microorganisms |
| CN105844285A (en) * | 2016-02-25 | 2016-08-10 | 中国农业大学 | Cucumber disease identification method and apparatus based on image information |
-
2017
- 2017-01-05 CN CN201710007720.4A patent/CN106897991A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105447527A (en) * | 2015-12-31 | 2016-03-30 | 四川木牛流马智能科技有限公司 | Method and system for classifying environmental microorganisms by image recognition technology |
| CN105608472A (en) * | 2015-12-31 | 2016-05-25 | 四川木牛流马智能科技有限公司 | Method and system for carrying out fully automatic classification of environmental microorganisms |
| CN105844285A (en) * | 2016-02-25 | 2016-08-10 | 中国农业大学 | Cucumber disease identification method and apparatus based on image information |
Non-Patent Citations (2)
| Title |
|---|
| WILLIAM A. BARRETT等: "Interactive live-wire boundary extraction", 《MEDICAL IMAGE ANALYSIS》 * |
| 芶小珊: "乳腺X线图像预处理方法的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111179288A (en) * | 2019-12-20 | 2020-05-19 | 浙江理工大学 | Interactive contrast blood vessel segmentation method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106682633B (en) | The classifying identification method of stool examination image visible component based on machine vision | |
| CN107346436A (en) | A kind of vision significance detection method of fused images classification | |
| CN109190567A (en) | Abnormal cervical cells automatic testing method based on depth convolutional neural networks | |
| CN109977780A (en) | A kind of detection and recognition methods of the diatom based on deep learning algorithm | |
| CN109919177B (en) | Feature selection method based on hierarchical deep network | |
| CN109740686A (en) | A kind of deep learning image multiple labeling classification method based on pool area and Fusion Features | |
| CN107016405A (en) | A kind of insect image classification method based on classification prediction convolutional neural networks | |
| CN107292314A (en) | A kind of lepidopterous insects species automatic identification method based on CNN | |
| CN107341812A (en) | A kind of sequence Lung neoplasm image partition method based on super-pixel and Density Clustering | |
| CN107944410A (en) | A kind of cross-cutting facial characteristics analytic method based on convolutional neural networks | |
| CN106127197A (en) | A kind of saliency object detection method based on notable tag sorting | |
| CN106611420A (en) | SAR image segmentation method based on deconvolution network and sketch direction constraint | |
| CN111476307B (en) | Lithium battery surface defect detection method based on depth field adaptation | |
| CN109544564A (en) | A kind of medical image segmentation method | |
| CN107545571A (en) | A kind of image detecting method and device | |
| CN103745233B (en) | The hyperspectral image classification method migrated based on spatial information | |
| CN108549901A (en) | A kind of iteratively faster object detection method based on deep learning | |
| CN108764330A (en) | SAR image sorting technique based on super-pixel segmentation and convolution deconvolution network | |
| CN106991049A (en) | A kind of Software Defects Predict Methods and forecasting system | |
| CN107818321A (en) | A kind of watermark date recognition method for vehicle annual test | |
| CN107590821A (en) | A kind of method for tracking target and system based on track optimizing | |
| CN108960342A (en) | Based on the image similarity calculation method for improving SoftMax loss function | |
| CN107292346A (en) | A kind of MR image hippocampus partitioning algorithms learnt based on Local Subspace | |
| CN106529604A (en) | Adaptive image tag robust prediction method and system | |
| CN106203448A (en) | A kind of scene classification method based on Nonlinear Scale Space Theory |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170627 |
|
| WD01 | Invention patent application deemed withdrawn after publication |