CN108090426A - A kind of group rearing pig personal identification method based on machine vision - Google Patents

A kind of group rearing pig personal identification method based on machine vision Download PDF

Info

Publication number
CN108090426A
CN108090426A CN201711284288.XA CN201711284288A CN108090426A CN 108090426 A CN108090426 A CN 108090426A CN 201711284288 A CN201711284288 A CN 201711284288A CN 108090426 A CN108090426 A CN 108090426A
Authority
CN
China
Prior art keywords
mrow
pig
msub
mtd
feature
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
Application number
CN201711284288.XA
Other languages
Chinese (zh)
Inventor
朱伟兴
黄炜嘉
李新城
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu University
Original Assignee
Jiangsu University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jiangsu University filed Critical Jiangsu University
Priority to CN201711284288.XA priority Critical patent/CN108090426A/en
Publication of CN108090426A publication Critical patent/CN108090426A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00Scenes; Scene-specific elements
    • G06V20/40Scenes; Scene-specific elements in video content
    • G06V20/41Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
    • G06V20/42Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/24Classification techniques
    • G06F18/241Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411Classification 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
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00Arrangements for image or video recognition or understanding
    • G06V10/20Image preprocessing
    • G06V10/28Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00Scenes; Scene-specific elements
    • G06V20/40Scenes; Scene-specific elements in video content
    • G06V20/46Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Data Mining & Analysis (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Analysis (AREA)

Abstract

The present invention provides a kind of group rearing pig personal identification methods based on machine vision.Group rearing pig is gathered first overlooks video;Then pig individual is extracted from the single-frame images of video;Then in a frequency domain, multiple dimensioned multidirectional feature of pig individual is extracted;In the spatial domain, the small minutia of pig individual is extracted;Again, Feature Dimension Reduction is carried out to the feature of extraction, and is cascaded into feature vector;Grader is finally established, is trained and identifies.Ear tag is observed and uses compared with using artificial, the identification of pig individual is carried out with untouchable using machine vision and mode identification technology, the mode of intrusive mood is not required to place electron ear tage, pig stress reaction is not easily caused, also the problems such as ear tag avoided is lost, cost of labor is reduced, technology guarantee is provided subsequently to automatically analyze livestock and poultry behavior.

Description

A kind of group rearing pig personal identification method based on machine vision
Technical field
The present invention relates to the technologies such as machine vision, pattern-recognition, and in particular to the monitoring of group rearing pig regards under a kind of vertical view state In frequency, the group rearing pig personal identification method based on machine vision.
Background technology
It is to automatically analyze the important step of livestock and poultry behavior to carry out identification to animal individual.From the pig image collected In, it is observed that pig body surface such as back, the decorative pattern in flank portion.These decorative patterns may be the different shape of muscle, skin Caused by spot that the different lines for moving towards to be formed of skin surface hair and skin surface color change are formed etc..It is lost The collective effect of biography factor and environmental factor, different geometries is presented in different position for they;They depend on kind, raise The process of supporting and growth and development situation.These decorative patterns can be described by textural characteristics, and be provided to distinguish different pigs Foundation.Texture is a kind of visual signature independent of color or brightness change, can describe neighborhood of pixels space in image The rule of distribution.The present invention provides a kind of method using machine vision, and acquisition group rearing pig overlooks video, extracts pig body table The method of machine vision and pattern-recognition is applied to group rearing pig identification by the textural characteristics in face.Compared with using manually sight It examines and using ear tag, carries out the identification of pig individual with untouchable using machine vision and mode identification technology, be not required to The mode of intrusive mood is wanted to place electron ear tage, does not easily cause pig stress reaction, the problems such as ear tag that also avoids is lost, is reduced Cost of labor provides technology guarantee subsequently to automatically analyze livestock and poultry behavior.
The content of the invention
The purpose of the present invention is by machine vision and mode identification technology, regarded by the group rearing pig collected Frequently, the identification of pig individual identity is realized.
The technical solution adopted in the present invention is:Group rearing pig personal identification method based on machine vision, avoids to invade The mode for entering formula places electron ear tage on pig body, and the feature of pig individual subject need to be only gathered from video monitoring image, With untouchable, comprise the following steps:
(1) gather group rearing pig and overlook video;
(2) pig individual is extracted from the single-frame images of video;
(3) in a frequency domain, multiple dimensioned multidirectional feature of pig individual is extracted;
(4) in the spatial domain, the minor detail feature of pig individual is extracted;
(5) Feature Dimension Reduction is carried out to the feature of extraction, and is cascaded into feature vector;
(6) grader is established, is trained and identifies.
Step (1) described content, is described in detail below:Experiment with pig house install image capturing system, to group rearing pig into Row video monitoring.Video camera is mounted on directly over pig house, and video is overlooked in acquisition, can be collected under the vertical view state comprising background The RGB color video of group rearing pig.
Step (2) described content, is described in detail below:Group rearing pig video sequence framing to acquisition, to each two field picture Objective extraction is carried out, obtains single pig individual images, specific method is as follows:
1. select meet experiment condition can completely isolate all pig individual goals without adhesion picture frame;
2. and then the vertical view group rearing pig multiple target extracting method of adaptive piecemeal multi-threshold is used to obtain the prospect of single pig Image;
3. in order to obtain the pig individual original image for being suitable for feature extraction, using the result of pig individual perspective detection as mould Plate carries out template computing with original image frame;
4. centered on pig barycenter, intercept the square sub blocks that can include complete pig Individual Size and be normalized to phase Same size.
Step (3) described content, is described in detail below:Pig individual images are transformed into frequency domain, pass through Gabor amplitude responses Value description pig body surface decorative pattern, step are as follows:Pig individual images are carried out with the Gabor transformation in different scale and direction, it will Pig individual images I () and Gabor kernel functions Ψu,v() carries out convolutional calculation, is represented using equation below:
Wherein u represents direction, and v represents scale, and * represents convolution algorithm.If input picture and u × v different directions and ruler The bidimensional Gabor wavelet convolution of degree can obtain u × v Gabor amplitude response images.
Then, vectorial to pig individual images extraction Gabor characteristic, step is as follows:
1. the gray value of individual Gabor amplitude response image is connected into a row vector by row head and the tail;
2. then the row vector that u × v Gabor amplitude responses images are formed is cascaded successively, to merge different spaces Frequency, spatial position and set direction;
This row vector is just as the Gabor characteristic extracted to pig individual images.
Step (4) described content, is described in detail below:The micro-structure details in pig individual images is calculated, records pig The different rules that pixel changes in image, are as follows:
1. original image is divided into the sub-block of same size;
2. and then with sub-block center pixel IcGray value for threshold value, the pixel I with its neighborhoodkGray value compared Compared with the process compared carries out binary conversion treatment, and the point that will be greater than threshold value puts 1, and the point less than threshold value is set to 0, and utilizes equation below:
3. result of the comparison does weighted sum according to the position difference of pixel and is treated as a binary sequence, as this The local binary patterns encoded radio f of sub-blockl(xc,yc).Computational methods are shown below:
4. using from 0 to 1 or 1 to 0 saltus step sum is no more than situation twice as one mode, it is left situation One mode is merged into jointly.
Then, it is as follows to pig individual images extraction local binary patterns characterization step:
1. by local binary patterns code pattern fl(xc,yc) it is divided into m sub-block { R1,R2,…,Rm};
2. and then calculate the histogram H of each sub-blocki,j, it is shown below:
3. the histogram of all sub-blocks cascades up again, the local binary patterns feature H of pig image is formed, using such as Lower formula
H=[Hi,j], i=0,1 ..., n-1, j=0,1 ..., m-1
Wherein n represents the different patterns that local binary patterns description is formed.
Step (5) described content, is described in detail below:
1. is taken by 1 sampled point every k row, carries out space down-sampling every k rows for the filtered images of Gabor;2. then Dimensionality reduction is carried out using principal component analytical method to the Gabor characteristic vector after down-sampling;
3. to the local binary patterns feature of extraction, dimensionality reduction is carried out using principal component analytical method;
4. Gabor characteristic and local binary patterns feature vector are cascaded, obtain to current pig individual images from frequency domain and The feature of spatial domain extraction.
Step (6) described content, is described in detail below:The picture of each pig is all corresponding with the label manually marked, base In support vector machines, the method for " one-to-one " is used to carry out multicategory classification.Between the pig feature of arbitrary two classes difference label Support vector machine classifier is established, vertical k (k-1)/2 grader of building together, as the model obtained from this group of training set.
Then, pig individual identification process is described in detail below:To the feature vector extracted from a certain pig image, make It is judged respectively with all graders established in the training process, it is just one ticket of note which pig is judging result, which belong to, Most input pig image is classified as the class for possessing most polls at last.
The beneficial effects of the invention are as follows:
The present invention realizes the identity of pig individual under actual breeding environment, the video of group rearing pig is analyzed and handled Identification.Since uneven illumination is even in actual pig farm environment, contrast is different everywhere, with the movement of pig, in breeding environment The location of middle difference, the picture illumination difference and noise problem of acquisition are also inevitable.The present invention can be certain The influence of uneven illumination and noise is overcome in degree, obtains accurate discrimination.Method based on machine vision, overcomes Traditional artificial observation and the limitation of ear tag mode improve the time-consuming and laborious of artificial view mode and to pig individual generation Interference, also avoids the invasive of ear tag mode, reduces cost.
Description of the drawings
Fig. 1 is the group rearing pig personal identification method flow chart the present invention is based on machine vision;
Fig. 2 is the example of video acquisition platform of the present invention and the video image frame of acquisition:(a) video acquisition platform, (b) are adopted Certain picture frame collected;
Fig. 3 is the example that the present invention obtains single pig individual process from group rearing pig video frame:(a) certain image after framing Frame, (b) foreground detection as a result, (c) to single pig Objective extraction as a result, (d) normalization after pig individual images (e) when 7 pig individual coloured images of previous frame extraction;
Fig. 4 is the example that the present invention extracts pig individual images multiple dimensioned Orientation Features:(a) pig individual images, (b) The filtered amplitude responses of Gabor;
Fig. 5 is example of the present invention to pig individual images extraction minor detail feature:(a) pig individual images are calculated local Binary pattern encodes, the histogram of (b) local binary patterns list block, during (c) local binary patterns code pattern piecemeal 9 × 9 Cascade histogram.
Specific embodiment
The present invention is described in detail with reference to the accompanying drawings and detailed description, but protection scope of the present invention is not It is limited to this.
Fig. 1 is the flow chart of the group rearing pig personal identification method based on machine vision, with reference to the figure, is further illustrated The each several part specific embodiment being specifically related to.
(1) gather group rearing pig and overlook video.
Specific method is as follows:As shown in Fig. 2, by reconstructing experiment pig house, the Image Acquisition of video is overlooked in installation shooting System carries out video monitoring to group rearing pig.Video camera, which is mounted on directly over 3 meters of pig house from the ground, overlooks acquisition video, can be with Collect the RGB color video of group rearing pig under the vertical view state comprising background.
(2) pig individual is extracted from the single-frame images of video.
Specific method is as follows:As shown in figure 3, the group rearing pig video sequence framing to acquisition, mesh is carried out to each two field picture Mark extraction, obtains single pig individual images.
1. select meet experiment condition can completely isolate all pig individual goals without adhesion picture frame;
2. and then the vertical view group rearing pig multiple target extracting method of adaptive piecemeal multi-threshold is used to obtain the prospect of single pig Image;
3. in order to obtain the pig individual original image for being suitable for feature extraction, using the result of pig individual perspective detection as mould Plate carries out template computing with original image frame;
4. centered on pig barycenter, intercept the square sub blocks that can include complete pig Individual Size and be normalized to phase Same size.
(3) in a frequency domain, multiple dimensioned multidirectional feature of pig individual is extracted.
Specific method is as follows:Pig individual images are transformed into frequency domain, pig body table is described by Gabor amplitude response values Face decorative pattern, step are as follows:Pig individual images are carried out with the Gabor transformation in different scale and direction, by pig individual images I () with Gabor kernel functions Ψu,v() carries out convolutional calculation, is represented using equation below:
Wherein u represents direction, and v represents scale, and * represents convolution algorithm.The decorative pattern of pig body surface by muscle difference The common shapes such as the spot that shape, the different lines for moving towards to be formed of skin surface hair and skin surface color change are formed Into with different scale and direction.As shown in figure 4, the bidimensional Gabor for setting input picture and 5 × 8 different directions and scale is small Ripple convolution can obtain 5 × 8=40 Gabor characteristic images.
Then, vectorial to pig individual images extraction Gabor characteristic, step is as follows:
1. the gray value of individual Gabor amplitude response image is connected into a row vector by row head and the tail;
2. then the row vector that u × v Gabor amplitude responses images are formed is cascaded successively, to merge different spaces Frequency, spatial position and set direction;
This row vector is just as the Gabor characteristic extracted to pig individual images.
(4) in the spatial domain, the minor detail feature of pig individual is extracted.
Specific method is as follows:As shown in Fig. 5 (a), first, the micro-structure details in pig individual images is calculated, record pig The different rules that pixel changes in image, are as follows:
1. original image is divided into the sub-block of same size;
2. and then with sub-block center pixel IcGray value for threshold value, the pixel I with its neighborhoodkGray value compared Compared with the process compared carries out binary conversion treatment, and the point that will be greater than threshold value puts 1, and the point less than threshold value is set to 0, and utilizes equation below:
3. result of the comparison does weighted sum according to the position difference of pixel and is treated as a binary sequence, as this The local binary patterns encoded radio f of sub-blockl(xc,yc).Computational methods are shown below:
In formula, IcRepresent the central point of sub-block, gray value is (xc,yc), IkIt is IcThe pixel of surrounding.Will from 0 to 1 or For the saltus step sum of person 1 to 0 no more than situation twice respectively as one mode, remaining situation merges into one mode jointly.
4. using from 0 to 1 or 1 to 0 saltus step sum is no more than situation twice as one mode, it is left situation One mode is merged into jointly.
Then, as shown in Fig. 5 (b), (c), local binary patterns feature is extracted to pig individual images, is as follows:
1. by local binary patterns code pattern fl(xc,yc) it is divided into m sub-block { R1,R2,…,Rm};
2. and then calculate the histogram H of each sub-blocki,j, it is shown below:
In formula, flIt is local binary patterns coded image, (x, y) represents the coordinate of pixel, and it is (x, y) to coordinate that Q, which is, Pixel whether appear in sub-block RjIt is middle judged as a result, M represent Q functions judgement content, Hi,jIt is single sub-block Local binary patterns histogram.
3. the histogram of all sub-blocks cascades up again, the local binary patterns feature H of pig image is formed, using such as Lower formula
H=[Hi,j], i=0,1 ..., n-1, j=0,1 ..., m-1
Wherein n represents the different patterns that local binary patterns description is formed.
(5) Feature Dimension Reduction is carried out to the feature of extraction, and is cascaded into feature vector.
Specific method is as follows:
1. is taken by 1 sampled point every k row, carries out space down-sampling every k rows for the filtered images of Gabor;
2. and then dimensionality reduction is carried out using principal component analytical method to the Gabor characteristic vector after down-sampling;
3. to the local binary patterns feature of extraction, dimensionality reduction is carried out using principal component analytical method;
4. Gabor characteristic and local binary patterns feature vector are cascaded, obtain to current pig individual images from frequency domain and The feature of spatial domain extraction.
(6) grader is established, is trained and identifies.
Specific method is as follows:Training process is described in detail below:The picture of each pig is all corresponding with the mark manually marked Label based on support vector machines, use the method for " one-to-one " to carry out multicategory classification.In the pig feature of arbitrary two classes difference label Between establish support vector machine classifier.This experiment totally 7 class pig picture, vertical 7 (7-1)/2=21 grader of building together, as The model obtained from this group of training set.
Pig individual identification process is described in detail below:To the feature vector extracted from a certain pig image, using 21 graders established in training process respectively judge it is just one ticket of note which pig is judging result, which belong to, finally Input pig image is classified as the class for possessing most polls.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " illustrative examples ", The description of " example ", " specific example " or " some examples " etc. means to combine specific features, the knot that the embodiment or example describe Structure, material or feature are contained at least one embodiment of the present invention or example.In the present specification, to above-mentioned term Schematic representation may not refer to the same embodiment or example.Moreover, specific features, structure, material or the spy of description Point can in an appropriate manner combine in any one or more embodiments or example.
Although an embodiment of the present invention has been shown and described, it will be understood by those skilled in the art that:Not In the case of departing from the principle of the present invention and objective a variety of change, modification, replacement and modification can be carried out to these embodiments, this The scope of invention is limited by claim and its equivalent.

Claims (8)

1. the group rearing pig personal identification method based on machine vision, which is characterized in that comprise the following steps:
1) gather group rearing pig and overlook video;
2) pig individual is extracted from the single-frame images of video;
3) multiple dimensioned multidirectional feature of pig individual in a frequency domain, is extracted;
4) the minor detail feature of pig individual in the spatial domain, is extracted;
5) Feature Dimension Reduction is carried out to the feature of extraction, and is cascaded into feature vector;
6) grader is established, is trained and identifies.
2. the group rearing pig personal identification method based on machine vision according to claim 1, which is characterized in that bag in step 3) It includes and pig individual images is transformed into frequency domain, pig body surface decorative pattern is described by Gabor amplitude response values, step is as follows:It is right Pig individual images carry out the Gabor transformation in different scale and direction, by pig individual images I () and Gabor kernel functions Ψu,v () carries out convolutional calculation, is represented using equation below:
<mrow> <msub> <mi>G</mi> <mrow> <mi>u</mi> <mo>,</mo> <mi>v</mi> </mrow> </msub> <mrow> <mo>(</mo> <mi>z</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>I</mi> <mrow> <mo>(</mo> <mi>z</mi> <mo>)</mo> </mrow> <mo>*</mo> <msub> <mi>&amp;Psi;</mi> <mrow> <mi>u</mi> <mo>,</mo> <mi>v</mi> </mrow> </msub> <mrow> <mo>(</mo> <mi>z</mi> <mo>)</mo> </mrow> <mo>=</mo> <msub> <mi>A</mi> <mrow> <mi>u</mi> <mo>,</mo> <mi>v</mi> </mrow> </msub> <mrow> <mo>(</mo> <mi>z</mi> <mo>)</mo> </mrow> <msup> <mi>e</mi> <mrow> <msub> <mi>i&amp;theta;</mi> <mrow> <mi>u</mi> <mo>,</mo> <mi>v</mi> </mrow> </msub> <mrow> <mo>(</mo> <mi>z</mi> <mo>)</mo> </mrow> </mrow> </msup> </mrow>
Wherein u represents direction, and v represents scale, and * represents convolution algorithm, if input picture and u × v different directions and scale Bidimensional Gabor wavelet convolution can obtain u × v Gabor amplitude response images.
3. the group rearing pig personal identification method based on machine vision according to claim 1, which is characterized in that in step 3) also Including vectorial to pig individual images extraction Gabor characteristic, step is as follows:
The gray value of individual Gabor amplitude response image is connected into a row vector by row head and the tail;
Then the row vector that u × v Gabor amplitude responses images are formed is cascaded successively, it is empty to merge different spatial frequencys Between position and set direction;
This row vector is just as the Gabor characteristic extracted to pig individual images.
4. the group rearing pig personal identification method based on machine vision according to claim 1, which is characterized in that bag in step 4) The micro-structure details calculated in pig individual images is included, records the different rules that pixel changes in pig image, step is as follows:
Original image is divided into the sub-block of same size;
Then with sub-block center pixel IcGray value for threshold value, the pixel I with its neighborhoodkGray value be compared, compare Process carry out binary conversion treatment, the point that will be greater than threshold value puts 1, and the point less than threshold value is set to 0, and utilizes equation below:
<mrow> <mi>Z</mi> <mrow> <mo>(</mo> <msub> <mi>I</mi> <mi>k</mi> </msub> <mo>,</mo> <msub> <mi>I</mi> <mi>c</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mn>1</mn> <mo>,</mo> </mrow> </mtd> <mtd> <mrow> <mi>i</mi> <mi>f</mi> </mrow> </mtd> <mtd> <mrow> <msub> <mi>I</mi> <mi>k</mi> </msub> <mo>-</mo> <msub> <mi>I</mi> <mi>c</mi> </msub> <mo>&amp;GreaterEqual;</mo> <mn>0</mn> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mn>0</mn> <mo>,</mo> </mrow> </mtd> <mtd> <mrow> <mi>i</mi> <mi>f</mi> </mrow> </mtd> <mtd> <mrow> <msub> <mi>I</mi> <mi>k</mi> </msub> <mo>-</mo> <msub> <mi>I</mi> <mi>c</mi> </msub> <mo>&lt;</mo> <mn>0</mn> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>,</mo> <mi>k</mi> <mo>=</mo> <mn>0</mn> <mo>,</mo> <mn>1</mn> <mo>,</mo> <mo>...</mo> <mo>,</mo> <mn>7</mn> </mrow>
Result of the comparison does weighted sum according to the position difference of pixel and is treated as a binary sequence, as the sub-block Local binary patterns encoded radio fl(xc,yc), computational methods are shown below:
<mrow> <msub> <mi>f</mi> <mi>l</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>c</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>c</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>0</mn> </mrow> <mn>7</mn> </munderover> <mi>Z</mi> <mrow> <mo>(</mo> <msub> <mi>I</mi> <mi>k</mi> </msub> <mo>,</mo> <msub> <mi>I</mi> <mi>c</mi> </msub> <mo>)</mo> </mrow> <msup> <mn>2</mn> <mi>k</mi> </msup> </mrow>
In formula, IcRepresent the central point of sub-block, gray value is (xc,yc), IkIt is IcThe pixel of surrounding.It will be from 0 to 1 or 1 Saltus step sum to 0 is no more than situation twice respectively as one mode, and remaining situation merges into one mode jointly.
5. the group rearing pig personal identification method based on machine vision according to claim 1, which is characterized in that in step 4) also Including extracting local binary patterns characterization step to pig individual images:
By local binary patterns code pattern fl(xc,yc) it is divided into m sub-block { R1,R2,…,Rm};
Then the histogram H of each sub-block is calculatedi,j, it is shown below:
<mrow> <msub> <mi>H</mi> <mrow> <mi>i</mi> <mo>,</mo> <mi>j</mi> </mrow> </msub> <mo>=</mo> <munder> <mi>&amp;Sigma;</mi> <mrow> <mi>x</mi> <mo>,</mo> <mi>y</mi> </mrow> </munder> <mi>Q</mi> <mo>{</mo> <msub> <mi>f</mi> <mi>l</mi> </msub> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>i</mi> <mo>}</mo> <mi>Q</mi> <mo>{</mo> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>&amp;Element;</mo> <msub> <mi>R</mi> <mi>j</mi> </msub> <mo>}</mo> <mo>,</mo> </mrow>
<mrow> <mi>Q</mi> <mo>{</mo> <mi>M</mi> <mo>}</mo> <mo>=</mo> <mfenced open = "{" close = "}"> <mtable> <mtr> <mtd> <mrow> <mn>1</mn> <mo>,</mo> </mrow> </mtd> <mtd> <mrow> <mi>M</mi> <mo>=</mo> <mi>t</mi> <mi>r</mi> <mi>u</mi> <mi>e</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mn>0</mn> <mo>,</mo> </mrow> </mtd> <mtd> <mrow> <mi>M</mi> <mo>=</mo> <mi>f</mi> <mi>a</mi> <mi>l</mi> <mi>s</mi> <mi>e</mi> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>,</mo> <mi>i</mi> <mo>=</mo> <mn>0</mn> <mo>,</mo> <mn>1</mn> <mo>,</mo> <mo>...</mo> <mo>,</mo> <mi>n</mi> <mo>-</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> <mo>=</mo> <mn>0</mn> <mo>,</mo> <mn>1</mn> <mo>,</mo> <mo>...</mo> <mo>,</mo> <mi>m</mi> <mo>-</mo> <mn>1</mn> </mrow>
In formula, flIt is local binary patterns coded image, (x, y) represents the coordinate of pixel, and Q is the pixel for (x, y) to coordinate Whether point appears in sub-block RjIt is middle judged as a result, M represent Q functions judgement content, Hi,jIt is local the two of single sub-block It is worth pattern histogram.Then, the histogram of all sub-blocks is cascaded up, forms the local binary patterns feature H of pig image, Utilize equation below
H=[Hi,j], i=0,1 ..., n-1, j=0,1 ..., m-1
Wherein n represents the different patterns that local binary patterns description is formed.
6. the group rearing pig personal identification method based on machine vision according to claim 1, which is characterized in that step 5) is specific Process is as follows:
5.1) is taken by 1 sampled point every k row, carries out space down-sampling every k rows for the filtered images of Gabor;
5.2) and then to the Gabor characteristic vector after down-sampling dimensionality reduction is carried out using principal component analytical method;
5.3) to the local binary patterns feature of extraction, dimensionality reduction is carried out using principal component analytical method;
5.4) Gabor characteristic and local binary patterns feature vector are cascaded, obtained to current pig individual images from frequency domain and sky Between domain extract feature.
7. the group rearing pig personal identification method based on machine vision according to claim 1, which is characterized in that built in step 6) What vertical grader was trained is described in detail below:The picture of each pig is all corresponding with the label manually marked, based on support Vector machine carries out multicategory classification using man-to-man method, classification is established between the pig feature of arbitrary two classes difference label Device, vertical k (k-1)/2 grader of building together, as the model obtained from this group of training set.
8. the group rearing pig personal identification method based on machine vision according to claim 1, which is characterized in that pig in step 6) Individual identification process detailed process is as follows:To the feature vector extracted from a certain pig image, using in the training process All graders established respectively judge it is just one ticket of note which pig is judging result, which belong to, most inputs pig at last Image is classified as the class for possessing most polls.
CN201711284288.XA 2017-12-07 2017-12-07 A kind of group rearing pig personal identification method based on machine vision Pending CN108090426A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711284288.XA CN108090426A (en) 2017-12-07 2017-12-07 A kind of group rearing pig personal identification method based on machine vision

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711284288.XA CN108090426A (en) 2017-12-07 2017-12-07 A kind of group rearing pig personal identification method based on machine vision

Publications (1)

Publication Number Publication Date
CN108090426A true CN108090426A (en) 2018-05-29

Family

ID=62174024

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711284288.XA Pending CN108090426A (en) 2017-12-07 2017-12-07 A kind of group rearing pig personal identification method based on machine vision

Country Status (1)

Country Link
CN (1) CN108090426A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110222725A (en) * 2019-05-15 2019-09-10 广州影子科技有限公司 Pig checking method, pig veritify device and pig verifying system
CN110402840A (en) * 2019-07-25 2019-11-05 深圳市阿龙电子有限公司 Live pig monitoring terminal and live pig monitoring system based on image recognition
CN110826371A (en) * 2018-08-10 2020-02-21 京东数字科技控股有限公司 Animal identification method, device, medium and electronic equipment
CN111242035A (en) * 2020-01-14 2020-06-05 江苏大学 Local feature-based group-breeding pig identity identification method
CN118333080A (en) * 2024-06-13 2024-07-12 双胞胎(集团)股份有限公司 Digital cultivation management method and digital cultivation management system
CN118365847A (en) * 2023-09-13 2024-07-19 张宇琦 Training method for virtual electronic ear tag addition model, ear tag addition method and device

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101154265A (en) * 2006-09-29 2008-04-02 中国科学院自动化研究所 Method for recognizing iris with matched characteristic and graph based on partial bianry mode
CN105654141A (en) * 2016-01-06 2016-06-08 江苏大学 Isomap and SVM algorithm-based overlooked herded pig individual recognition method
CN106778784A (en) * 2016-12-20 2017-05-31 江苏大学 Pig individual identification and drinking behavior analysis method based on machine vision
CN107437068A (en) * 2017-07-13 2017-12-05 江苏大学 Pig individual discrimination method based on Gabor direction histograms and pig chaeta hair pattern

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101154265A (en) * 2006-09-29 2008-04-02 中国科学院自动化研究所 Method for recognizing iris with matched characteristic and graph based on partial bianry mode
CN105654141A (en) * 2016-01-06 2016-06-08 江苏大学 Isomap and SVM algorithm-based overlooked herded pig individual recognition method
CN106778784A (en) * 2016-12-20 2017-05-31 江苏大学 Pig individual identification and drinking behavior analysis method based on machine vision
CN107437068A (en) * 2017-07-13 2017-12-05 江苏大学 Pig individual discrimination method based on Gabor direction histograms and pig chaeta hair pattern

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110826371A (en) * 2018-08-10 2020-02-21 京东数字科技控股有限公司 Animal identification method, device, medium and electronic equipment
CN110222725A (en) * 2019-05-15 2019-09-10 广州影子科技有限公司 Pig checking method, pig veritify device and pig verifying system
CN110402840A (en) * 2019-07-25 2019-11-05 深圳市阿龙电子有限公司 Live pig monitoring terminal and live pig monitoring system based on image recognition
CN110402840B (en) * 2019-07-25 2021-12-17 深圳市阿龙电子有限公司 Live pig monitoring terminal and live pig monitoring system based on image recognition
CN111242035A (en) * 2020-01-14 2020-06-05 江苏大学 Local feature-based group-breeding pig identity identification method
CN111242035B (en) * 2020-01-14 2023-08-22 江苏大学 Group pig raising identity recognition method based on local characteristics
CN118365847A (en) * 2023-09-13 2024-07-19 张宇琦 Training method for virtual electronic ear tag addition model, ear tag addition method and device
CN118333080A (en) * 2024-06-13 2024-07-12 双胞胎(集团)股份有限公司 Digital cultivation management method and digital cultivation management system
CN118333080B (en) * 2024-06-13 2024-09-13 双胞胎(集团)股份有限公司 Digital cultivation management method and digital cultivation management system

Similar Documents

Publication Publication Date Title
CN108090426A (en) A kind of group rearing pig personal identification method based on machine vision
Pérez-Zavala et al. A pattern recognition strategy for visual grape bunch detection in vineyards
CN105938564B (en) Rice disease identification method and system based on principal component analysis and neural network
CN107016357B (en) Video pedestrian detection method based on time domain convolutional neural network
Aquino et al. A new methodology for estimating the grapevine-berry number per cluster using image analysis
CN108520226B (en) Pedestrian re-identification method based on body decomposition and significance detection
Roscher et al. Automated image analysis framework for high-throughput determination of grapevine berry sizes using conditional random fields
CN107909005A (en) Personage&#39;s gesture recognition method under monitoring scene based on deep learning
CN107909081B (en) Method for quickly acquiring and quickly calibrating image data set in deep learning
CN106778609A (en) A kind of electric power construction field personnel uniform wears recognition methods
CN112136505A (en) Fruit picking sequence planning method based on visual attention selection mechanism
CN109215034A (en) A kind of Weakly supervised image, semantic dividing method for covering pond based on spatial pyramid
Tamilselvi et al. Unsupervised machine learning for clustering the infected leaves based on the leaf-colours
CN112464730B (en) Pedestrian re-identification method based on domain-independent foreground feature learning
CN104036250B (en) Video pedestrian detection and tracking
CN107153819A (en) A kind of queue length automatic testing method and queue length control method
CN112883915A (en) Automatic wheat ear identification method and system based on transfer learning
CN104008404B (en) Pedestrian detection method and system based on significant histogram features
CN108960276B (en) Sample expansion and consistency discrimination method for improving spectral image supervision classification performance
Sood et al. Image quality enhancement for Wheat rust diseased images using Histogram equalization technique
Zeng et al. Nearest neighbor based digital restoration of damaged ancient Chinese paintings
CN108256557A (en) The hyperspectral image classification method integrated with reference to deep learning and neighborhood
CN103150573B (en) Based on the nerve dendritic spine image classification method of many resolving power fractal characteristic
CN110569716A (en) Goods shelf image copying detection method
Shireesha et al. Citrus fruit and leaf disease detection using DenseNet

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20180529