CN106204537B - Live pig image partition method under a kind of complex environment - Google Patents
Live pig image partition method under a kind of complex environment Download PDFInfo
- Publication number
- CN106204537B CN106204537B CN201610493686.1A CN201610493686A CN106204537B CN 106204537 B CN106204537 B CN 106204537B CN 201610493686 A CN201610493686 A CN 201610493686A CN 106204537 B CN106204537 B CN 106204537B
- Authority
- CN
- China
- Prior art keywords
- image
- live pig
- obtains
- segmented
- light source
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses live pig image partition methods under a kind of complex environment.Acquisition live pig image in real time obtains light source reference point, establishes rectangular coordinate system, and live pig image grayscale to be split is converted, and obtains live pig grayscale image, and then handle and obtain live pig gradient map, and segmentation, which is extracted, obtains initial segmentation image;To initial segmentation closing operation, then connected domain is removed, obtains secondary splitting image;It calculates and obtains object, object area, intensity of illumination and distance, expansion structure member is biased in building, computational shadowgraph penalty coefficient carries out processing to secondary splitting image and obtains four segmented images, obtains five segmented images according to four segmented image combination secondary splitting image procossings.The method of the present invention effectively can be partitioned into pig body in the more complicated pig house live pig image of environment, and it can handle to continuous batch and there is self adjustment capability, suitable for carrying out image procossing to continuous prolonged live pig monitoring image, help to realize automatically-monitored to live pig progress.
Description
Technical field
The present invention relates to a kind of live pig image partition methods, more particularly, to the live pig image segmentation under a kind of complex environment
Method.
Background technique
Pig-breeding is gradually converted into more environmentally protective intensive style from traditional extensive breeding way.In order to make
Pig-breeding more efficiently, pig is monitored and detect its physiological status seem more it is necessary to.And tradition is to the situation of pig
It is monitored all by manually carrying out, since pig house environment is sultry, the smell is awful, and there have for a long time to the body of staff to be unfavorable
It influences, in addition, staff can not also accomplish to detect constantly, it is slower to the abnormal conditions reflection of live pig.And utilize machine
Vision system obtains live pig image and carries out the state-detection of live pig instead of manual type by image processing techniques simultaneously
Preferably selection.
In recent years, due to technological progress, have using the research that machine vision carries out animal state detection considerable degree of
Progress (pig abnormal breathing detection [D] Jiangsu University of the Wu Zhilei based on machine vision, 2011).However what most researchs used
Environment is all more simple, or has considerably artificial reforming, thus the number for facilitating image segmentation and subsequent processing, and using
According to being often short-term, so that the variation of various aspects is all unobvious in image, method used in image procossing is also more simple.And mesh
The environment on preceding many pig farms is complicated, lacks effective image partition method and splits pig body portion.
Summary of the invention
In order to solve the problems, such as background technique, the purpose of the present invention is to provide the live pigs under a kind of complex environment
Image partition method, by recycling its matter after carrying out background difference, Threshold segmentation acquisition primary segmentation image to live pig image
The heart and image light source information carry out shadow compensation, to obtain the image after live pig segmentation, traditional images processing method are avoided to exist
Failure under complex environment.
The technical solution adopted by the present invention to solve the technical problems is:
1) acquisition live pig image in real time is put into the fixed position of video camera, light source reference Point C L is obtained by image procossing;
It 2) is that y-axis establishes rectangular coordinate system using the vertical direction of image to be split as x-axis, horizontal direction;
3) for live pig image to be split, gradation conversion is carried out, obtains live pig grayscale image I, with live pig gray level image I
It subtracts background image BK and obtains difference image DI, and live pig grayscale image I is handled to obtain live pig gradient map;
4) gray value is greater than the pixel that gradient value in gray threshold and live pig gradient map is greater than Grads threshold on difference image
Point is labeled as 1, and rest of pixels point is labeled as 0, obtains initial segmentation image Bw1;Gray threshold and Grads threshold are being embodied
40 and 0.8 can be taken respectively.
5) to initial segmentation image Bw1Closed operation is carried out using morphology operations structural elements M, then removes area and is less than
The connected domain of area threshold obtains secondary splitting image Bw2;Area threshold can take 50000 in specific implementation respectively.
6) it is directed to secondary splitting image Bw2The object mass center CB and object area AB of its target object are calculated, and takes light source
Gray value at reference Point C L is as intensity of illumination L;
7) distance D is calculated, then expansion structure member MS is biased in building, is calculated and is obtained shadow compensation coefficient S;
8) result is obtained to the secondary splitting image Bw of step 5) using step 7)2It carries out processing and obtains four segmented images
Bw4, according to four segmented image Bw4In conjunction with secondary splitting image Bw2It is further processed and obtains five segmented image Bw5, as most
Whole segmentation result.
The live pig image is the top view photograph or side view photo of live pig.
Light source reference Point C L in the step 1) is to obtain in the following ways: around-the clock live pig image is turned
Gray level image is changed, the variance yields of each pixel gray-value variation in 24 hours is calculated, where taking the maximum pixel of variance yields
Position is as light source reference Point C L.
Gradient specifically is calculated using the following equation to each pixel in live pig grayscale image I in the step 3), thus
Obtain live pig gradient map:
In formula, Grad (i, j) is the gradient value for the pixel that row coordinate is i, column coordinate is j, and f (i, j) is the gray scale
Value, l are Laplace operator, and m indicates that the row coordinate of operator, n indicate the column coordinate of operator;
Laplace operator l uses the matrix of following formula:
Distance D in the step 7) refer between shade angle [alpha] and object mass center CB and light source reference Point C L away from
From D, specifically calculated with following formula:
In formula, α indicates shade angle, and D indicates the distance between object mass center CB and light source reference Point C L, xCBExpression pair
As the x coordinate of mass center CB, yCBIndicate the y-coordinate of object mass center CB, xCLIndicate the x coordinate of light source reference Point C L, yCLIndicate light source
The y-coordinate of reference Point C L.
Deviation expansion structure member MS in the step 7) is specifically constructed in the following ways: one full 0 matrix of setting, entirely
The side length of 0 matrix be rounded (D/10+1) value, using matrix center as origin, to the left lower section make ray so that ray and
Angle between image level direction is shade angle [alpha], then sets 1 for the matrix element with ray intersection, to obtain partially
To expansion structure member MS.
The folder of the shade angle [alpha] line and image level direction between light source reference Point C L and object mass center CB
Angle.
Shadow compensation coefficient S in the step 7) is specifically to be calculated using the following equation:
In formula, S0It is benchmark penalty coefficient;SABIt is reference area coefficient, measures size, AB indicates object area;aAB
It is area attenuation coefficient, control area changes the influence to penalty coefficient;SDIt is reference distance coefficient, measures apart from length;aDIt is
Range attenuation coefficient, command range change the influence to penalty coefficient;SLIt is benchmark illumination tensor, measures light intensity, these coefficients
Occurrence depending on actual conditions.
The step 8) specifically:
8.1) using the deviation expansion structure member MS of step 7) to the secondary splitting image Bw of step 5)2Carry out expansive working
Obtain segmented image Bw three times3, with segmented image Bw three times3Processing is masked to live pig grayscale image I and obtains secondary live pig gray scale
Figure, by secondary splitting image Bw2Processing is masked to secondary live pig grayscale image after reverse phase and obtains live pig grayscale image three times, by three
Secondary live pig grayscale image divides institute multiplied by the laggard row threshold division of background image BK is subtracted after the shadow compensation coefficient S of step 7) again
It obtains image and adds secondary splitting image Bw2For four segmented image Bw4;
If 8.2) four segmented image Bw4With secondary splitting image Bw2It is identical, then next step is carried out, otherwise with four
Secondary segmented image Bw4Instead of secondary splitting image Bw2It repeats afterwards 8.1) until four segmented image Bw4With secondary splitting image Bw2
It is identical;
8.3) to four segmented image Bw4It recycles morphology operations structural elements M to carry out morphology operations and obtains five times points
Cut image Bw5, as final segmentation result.
10. live pig image partition method under a kind of complex environment according to claim 1, it is characterised in that:
The step 5) and 8.3) in morphology operations structural elements M specifically use the matrix of following formula:
The step 3) and 8.1) in background image BK use initial background image in first frame processing, below
Background image of the real-time update as next frame processing is carried out when the processing of every frame: five segmentations obtained after present frame is handled
Image Bw5Dilation operation is carried out with circular configuration member, circular configuration member is specially that edge is 0, the structural elements that border circular areas is 1,
Obtain context update template Bw6, by context update template Bw6Live pig grayscale image I is masked after reverse phase, is obtained among first
Mask images, by context update template Bw6The background image BK of present frame processing is masked to obtain the second reticle mask
First reticle image and the second reticle image are added by image, obtain new background image BK as next frame at
The background image of reason.
The initial background image is to obtain in the following ways: by continuous 10 seconds live pig image converting gradation images,
The mean value of each pixel gray-value variation in 10 seconds is calculated, to obtain initial background image.
The beneficial effects of the present invention are:
The present invention is by recycling its matter after carrying out background difference, Threshold segmentation acquisition primary segmentation image to live pig image
The heart and image light source information carry out shadow compensation, to obtain the image after live pig segmentation, traditional images processing method are avoided to exist
Failure under complex environment.
It is partitioned into pig body in the live pig image that the method for the present invention can be obtained effectively in the more complicated pig house of environment, and
It can with continuous batch handle data and there is self adjustment capability to a certain extent, be suitable for supervising continuous prolonged live pig
It controls image and carries out image procossing, help to realize automatically-monitored to live pig progress.
Detailed description of the invention
Fig. 1 is treatment process method flow block diagram of the invention.
Fig. 2 is the original image of processing of the invention.
Fig. 3 is the secondary splitting image of processing of the invention.
Fig. 4 is the schematic diagram of angle used in shadow compensation of the invention and distance.
Fig. 5 is the deviation expansion structure member example used of the invention.
Fig. 6 is the final segmented image of processing of the invention.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and examples.
Step as shown in Figure 1, detailed process is as follows for the embodiment of the present invention:
1) the fixed position of video camera is put to the live pig image for acquiring depression angle as shown in Figure 2 in real time, obtains light source
Reference Point C L and initial background image BK:
By around-the clock live pig image converting gradation image, the side of each pixel gray-value variation in 24 hours is calculated
Difference takes the position where the maximum pixel of variance yields as light source reference Point C L.
By continuous 10 seconds live pig image converting gradation images, the mean value of each pixel gray-value variation in 10 seconds is calculated,
As background image BK.
It 2) is that y-axis establishes rectangular coordinate system using the vertical direction of image to be split as x-axis, horizontal direction;
3) for live pig image to be split, gradation conversion is carried out, obtains live pig grayscale image I, with live pig gray level image I
It subtracts background image BK and obtains difference image DI, and gradient is calculated to each pixel in live pig grayscale image I and obtains live pig gradient
Figure, is calculated as follows gradient
In formula, Grad (i, j) is the gradient value for the pixel that row coordinate is i, column coordinate is j, and f (i, j) is the gray scale
Value, l are Laplace operator, and m indicates that the row coordinate of operator, n indicate the column coordinate of operator.
Laplace operator l uses the matrix of following formula:
4) gray value on difference image is greater than the picture that gradient value on gray threshold and live pig gradient map is greater than Grads threshold
Vegetarian refreshments is labeled as 1, and rest of pixels point is labeled as 0, and gray threshold and Grads threshold can take 40 and 0.8 in specific implementation respectively, obtains
To initial segmentation image Bw1;
5) to initial segmentation image Bw1Closed operation is carried out using morphology operations structural elements M, then removes area and is less than
The connected domain of area threshold, area threshold can take 50000 in specific implementation respectively, obtain secondary splitting image Bw2, such as Fig. 3 institute
Show, for convenience of observing, wherein the part of black is that the part for the prospect that is divided into is labeled as 1 pixel, and white part is
Labeled as 0 part;
6) it is directed to secondary splitting image Bw2The object mass center CB and object area AB of its target object are calculated, and takes light source
Gray value at reference Point C L is as intensity of illumination L;
7) as shown in figure 4, be calculated as follows between shade angle [alpha] and object mass center CB and light source reference Point C L away from
From D.
In formula, α indicates shade angle, and D indicates the distance between object mass center CB and light source reference Point C L, xCBIndicate object
The x coordinate of mass center CB, yCBIndicate the y-coordinate of object mass center CB, xCLIndicate the x coordinate of light source reference Point C L, yCLIndicate light source ginseng
The y-coordinate of examination point CL.
Then expansion structure member MS is biased in building, sets a full 0 matrix, and the side length of full 0 matrix is the (D/10+ being rounded
1) value, using matrix center as origin, ray is made in lower section to the left, so that the angle between ray and image level direction is yin
Shadow angle [alpha], then 1 is set by the matrix element with ray intersection, to obtain being biased to expansion structure member MS, Fig. 5 being that α is
45 °, D structural elements example generated when being 100.
Then it is calculated as follows and obtains shadow compensation coefficient S,
In formula, S0It is benchmark penalty coefficient;SABIt is reference area coefficient, measures size, AB indicates object area;aAB
It is area attenuation coefficient, control area changes the influence to penalty coefficient;SDIt is reference distance coefficient, measures apart from length;aDIt is
Range attenuation coefficient, command range change the influence to penalty coefficient;SLIt is benchmark illumination tensor, measures light intensity.Base in calculating
Quasi- penalty coefficient S03 are taken, reference area coefficient SAB10000 are taken, area attenuation coefficient aAB2 are taken, reference distance coefficient SDTake 1000,
Range attenuation coefficient aD0.5 is taken, reference light shines coefficient SLTake 150.
8) using the deviation expansion structure member MS of step 7) to the secondary splitting image Bw of step 5)2Expansive working is carried out to obtain
To segmented image Bw three times3, with segmented image Bw three times3Processing is masked to live pig grayscale image I and obtains secondary live pig gray scale
Figure, by secondary splitting image Bw2Processing is masked to secondary live pig grayscale image after reverse phase and obtains live pig grayscale image three times, by three
Secondary live pig grayscale image divides institute multiplied by the laggard row threshold division of background image BK is subtracted after the shadow compensation coefficient S of step 7) again
It obtains image and adds secondary splitting image Bw2For four segmented image Bw4;
If four segmented image Bw4With secondary splitting image Bw2It is identical, then next step is carried out, otherwise with four times points
Cut image Bw4Instead of secondary splitting image Bw2It repeats afterwards 8.1) until four segmented image Bw4With secondary splitting image Bw2Completely
It is identical;
To four segmented image Bw4It recycles morphology operations structural elements M to carry out morphology operations and obtains five segmentation figures
As Bw5, as shown in fig. 6, as final segmentation result, relative to Fig. 3, before the part of pig body shady face is obviously divided into
Scape, general outline are also fuller.
The step 5) of embodiment and 8.3) in morphology operations structural elements M specifically use the matrix of following formula:
Embodiment step 3) and 8) in background image BK use initial background image in first frame processing, below
Background image of the real-time update as next frame processing is carried out when the processing of every frame: five segmentations obtained after present frame is handled
Image Bw5Dilation operation is carried out with circular configuration member, circular configuration member is specially that edge is 0, the structural elements that border circular areas is 1,
Obtain context update template Bw6, by context update template Bw6Live pig grayscale image I is masked after reverse phase, is obtained among first
Mask images, by context update template Bw6The background image BK of present frame processing is masked to obtain the second reticle mask
First reticle image and the second reticle image are added by image, obtain new background image BK as next frame at
The background image of reason, to adapt to the image segmentation operations under continuous complex environment.
Claims (9)
1. live pig image partition method under a kind of complex environment, it is characterised in that:
1) acquisition live pig image in real time is put into the fixed position of video camera, light source reference Point C L is obtained by image procossing;
It 2) is that y-axis establishes rectangular coordinate system using the vertical direction of image to be split as x-axis, horizontal direction;
3) for live pig image to be split, gradation conversion is carried out, obtains live pig grayscale image I, is subtracted with live pig gray level image I
Background image BK obtains difference image DI, and is handled to obtain live pig gradient map to live pig grayscale image I;
4) gray value on difference image is greater than the pixel that gradient value on gray threshold and live pig gradient map is greater than Grads threshold
Labeled as 1, rest of pixels point is labeled as 0, obtains initial segmentation image Bw1;
5) to initial segmentation image Bw1Closed operation is carried out using morphology operations structural elements M, then removes area less than area
The connected domain of threshold value obtains secondary splitting image Bw2;
6) it is directed to secondary splitting image Bw2The object mass center CB and object area AB of its target object are calculated, and takes light source reference point
Gray value at CL is as intensity of illumination L;
7) distance D is calculated, then expansion structure member MS is biased in building, is calculated and is obtained shadow compensation coefficient S;
Distance D in the step 7) refers to shade angle [alpha] and the distance between object mass center CB and light source reference Point C L D,
Specifically calculated with following formula:
In formula, α indicates shade angle, and D indicates the distance between object mass center CB and light source reference Point C L, xCBIndicate object mass center
The x coordinate of CB, yCBIndicate the y-coordinate of object mass center CB, xCLIndicate the x coordinate of light source reference Point C L, yCLIndicate light source reference point
The y-coordinate of CL;
8) result is obtained to the secondary splitting image Bw of step 5) using step 7)2It carries out processing and obtains four segmented image Bw4,
According to four segmented image Bw4In conjunction with secondary splitting image Bw2It is further processed and obtains five segmented image Bw5, as final
Segmentation result.
2. live pig image partition method under a kind of complex environment according to claim 1, it is characterised in that: the live pig
Image is the top view photograph or side view photo of live pig.
3. live pig image partition method under a kind of complex environment according to claim 1, it is characterised in that: the step 1)
In light source reference Point C L be to obtain in the following ways: around-the clock live pig image converting gradation image calculates each
The variance yields of pixel gray-value variation in 24 hours takes the position where the maximum pixel of variance yields as light source reference point
CL。
4. live pig image partition method under a kind of complex environment according to claim 1, it is characterised in that: the step 3)
In be specifically gradient is calculated using the following equation to each pixel in live pig grayscale image I, to obtain live pig gradient map:
In formula, Grad (i, j) is the gradient value for the pixel that row coordinate is i, column coordinate is j, and f (i, j) is the gray value, l
For Laplace operator, m indicates that the row coordinate of operator, n indicate the column coordinate of operator;
Laplace operator l uses the matrix of following formula:
5. live pig image partition method under a kind of complex environment according to claim 1, it is characterised in that: the step 7)
In deviation expansion structure member MS specifically construct in the following ways: one full 0 matrix of setting, the side length of full 0 matrix are to be rounded
(D/10+1) value, using matrix center as origin, to the left lower section make ray so that the angle between ray and image level direction
For shade angle [alpha], then by the matrix element with ray intersection 1 is set, to obtain being biased to expansion structure member MS.
6. live pig image partition method under a kind of complex environment according to claim 1 or 5, it is characterised in that: described
The angle of shade angle [alpha] line and image level direction between light source reference Point C L and object mass center CB.
7. live pig image partition method under a kind of complex environment according to claim 1, it is characterised in that: the step 7)
In shadow compensation coefficient S be specifically be calculated using the following equation:
In formula, S0It is benchmark penalty coefficient;SABIt is reference area coefficient, measures size, AB indicates object area;aABIt is face
Product attenuation coefficient;SDIt is reference distance coefficient;aDIt is range attenuation coefficient;SLIt is benchmark illumination tensor.
8. live pig image partition method under a kind of complex environment according to claim 1, it is characterised in that: the step 8)
Specifically:
8.1) using the deviation expansion structure member MS of step 7) to the secondary splitting image Bw of step 5)2Expansive working is carried out to obtain
Segmented image Bw three times3, with segmented image Bw three times3Processing is masked to live pig grayscale image I and obtains secondary live pig grayscale image,
By secondary splitting image Bw2Processing is masked to secondary live pig grayscale image after reverse phase and obtains live pig grayscale image three times, it will three times
Live pig grayscale image divides gained multiplied by the laggard row threshold division of background image BK is subtracted after the shadow compensation coefficient S of step 7) again
Image adds secondary splitting image Bw2For four segmented image Bw4;
If 8.2) four segmented image Bw4With secondary splitting image Bw2It is identical, then next step is carried out, otherwise with four times points
Cut image Bw4Instead of secondary splitting image Bw2It repeats afterwards 8.1) until four segmented image Bw4With secondary splitting image Bw2Completely
It is identical;
8.3) to four segmented image Bw4It recycles morphology operations structural elements M to carry out morphology operations and obtains five segmented images
Bw5, as final segmentation result.
9. live pig image partition method under a kind of complex environment according to claim 8, it is characterised in that: the step 5)
The matrix of following formula is specifically used with the morphology operations structural elements M in 8.3):
The step 3) and 8.1) in background image BK first frame processing when use initial background image, in every frame below
Background image of the real-time update as next frame processing: five segmented images obtained after present frame is handled is carried out when processing
Bw5Dilation operation is carried out with circular configuration member, obtains context update template Bw6, by context update template Bw6To live pig after reverse phase
Grayscale image I is masked, and obtains the first reticle image, by context update template Bw6To the Background of present frame processing
As BK is masked to obtain the second reticle image, the first reticle image and the second reticle image are added, obtained
Background image to new background image BK as next frame processing.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610493686.1A CN106204537B (en) | 2016-06-24 | 2016-06-24 | Live pig image partition method under a kind of complex environment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610493686.1A CN106204537B (en) | 2016-06-24 | 2016-06-24 | Live pig image partition method under a kind of complex environment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106204537A CN106204537A (en) | 2016-12-07 |
CN106204537B true CN106204537B (en) | 2019-05-21 |
Family
ID=57462364
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610493686.1A Active CN106204537B (en) | 2016-06-24 | 2016-06-24 | Live pig image partition method under a kind of complex environment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106204537B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107986127A (en) * | 2017-11-20 | 2018-05-04 | 江苏省特种设备安全监督检验研究院 | The stranded detection method of passenger in a kind of elevator |
CN109886984B (en) * | 2019-01-22 | 2021-01-08 | 浙江大学 | Image accurate segmentation method using foreground and background gray difference and deep learning network |
CN109886271B (en) * | 2019-01-22 | 2021-01-26 | 浙江大学 | Image accurate segmentation method integrating deep learning network and improving edge detection |
CN110539312A (en) * | 2019-08-29 | 2019-12-06 | 南京禹智智能科技有限公司 | Efficient and accurate livestock and poultry meat dividing robot |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2005034618A1 (en) * | 2003-10-10 | 2005-04-21 | Ab Svenska Mätanalys | Method and device for the monitoring of pigs |
CN104252709A (en) * | 2014-07-14 | 2014-12-31 | 江苏大学 | Multiple-target foreground detection method for look-down group-housed pigs in look-down state under complicated background |
CN104504704A (en) * | 2014-12-24 | 2015-04-08 | 江苏大学 | Multiple overlooked herded pig target extraction method capable of self-adaptive multi-threshold segmentation of blocks |
-
2016
- 2016-06-24 CN CN201610493686.1A patent/CN106204537B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2005034618A1 (en) * | 2003-10-10 | 2005-04-21 | Ab Svenska Mätanalys | Method and device for the monitoring of pigs |
CN104252709A (en) * | 2014-07-14 | 2014-12-31 | 江苏大学 | Multiple-target foreground detection method for look-down group-housed pigs in look-down state under complicated background |
CN104504704A (en) * | 2014-12-24 | 2015-04-08 | 江苏大学 | Multiple overlooked herded pig target extraction method capable of self-adaptive multi-threshold segmentation of blocks |
Also Published As
Publication number | Publication date |
---|---|
CN106204537A (en) | 2016-12-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106204537B (en) | Live pig image partition method under a kind of complex environment | |
Xiong et al. | Visual detection of green mangoes by an unmanned aerial vehicle in orchards based on a deep learning method | |
CN110232666A (en) | Underground piping image rapid defogging method based on dark primary priori | |
CN103702015B (en) | Exposure control method for human face image acquisition system under near-infrared condition | |
CN1738426A (en) | Video motion goal division and track method | |
CN103295010B (en) | A kind of unitary of illumination method processing facial image | |
CN104732578B (en) | A kind of building texture optimization method based on oblique photograph technology | |
CN104881644A (en) | Face image acquisition method under uneven lighting condition | |
CN105976337B (en) | A kind of image defogging method based on intermediate value guiding filtering | |
CN105959510B (en) | A kind of video rapid defogging method | |
JP2013141210A (en) | Image defogging apparatus, image defogging method, and image processing system | |
CN110378924A (en) | Level set image segmentation method based on local entropy | |
CN106713701B (en) | A kind of crowd hazards collecting method and system based on image processing techniques | |
CN109451292B (en) | Image color temperature correction method and device | |
CN102496155A (en) | Underwater optical image processing method for optimizing C-V (chan-vese) model | |
CN108109138B (en) | Method for self-adaptive light uniformization of high-light area of mirror-like object | |
CN111161228B (en) | Button surface defect detection method based on transfer learning | |
CN112862721A (en) | Underground pipeline image defogging method based on dark channel and Retinex | |
US11080861B2 (en) | Scene segmentation using model subtraction | |
US10108877B2 (en) | System for capturing pupil and method thereof | |
CN110148097B (en) | Color correction method of cataract image | |
CN106960421A (en) | Evening images defogging method based on statistical property and illumination estimate | |
CN112598777A (en) | Haze fusion method based on dark channel prior | |
CN107220983B (en) | A kind of live pig detection method and system based on video | |
Yu et al. | Synchronous measurement of 3D morphology and illuminance of plant leaves based on binocular vision |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20211123 Address after: 230000 room 1201, building 1, University Sanchuang Park, No. 478, Jiulong Road, economic and Technological Development Zone, Hefei, Anhui Province Patentee after: Hefei Shenmu Information Technology Co.,Ltd. Address before: 310027 No. 38, Zhejiang Road, Hangzhou, Zhejiang, Xihu District Patentee before: ZHEJIANG University |
|
TR01 | Transfer of patent right |