CN108765448A - A kind of shrimp seedling analysis of accounts method based on improvement TV-L1 models - Google Patents
A kind of shrimp seedling analysis of accounts method based on improvement TV-L1 models Download PDFInfo
- Publication number
- CN108765448A CN108765448A CN201810522933.5A CN201810522933A CN108765448A CN 108765448 A CN108765448 A CN 108765448A CN 201810522933 A CN201810522933 A CN 201810522933A CN 108765448 A CN108765448 A CN 108765448A
- Authority
- CN
- China
- Prior art keywords
- image
- shrimp seedling
- illumination
- avg
- connected region
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 241000238557 Decapoda Species 0.000 title claims abstract description 106
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000004458 analytical method Methods 0.000 title claims abstract description 11
- 230000006872 improvement Effects 0.000 title description 6
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 238000005286 illumination Methods 0.000 claims description 62
- 238000012545 processing Methods 0.000 claims description 18
- 238000001914 filtration Methods 0.000 claims description 16
- 230000008569 process Effects 0.000 claims description 13
- 238000011282 treatment Methods 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 230000000877 morphologic effect Effects 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 230000003628 erosive effect Effects 0.000 claims description 2
- 238000003709 image segmentation Methods 0.000 claims description 2
- 238000004321 preservation Methods 0.000 claims description 2
- 238000010187 selection method Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 6
- 230000035699 permeability Effects 0.000 abstract description 2
- 238000002474 experimental method Methods 0.000 description 4
- 230000011218 segmentation Effects 0.000 description 4
- 230000007797 corrosion Effects 0.000 description 3
- 238000005260 corrosion Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 235000013399 edible fruits Nutrition 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 241000251468 Actinopterygii Species 0.000 description 2
- 241000894006 Bacteria Species 0.000 description 2
- 241001344131 Magnaporthe grisea Species 0.000 description 2
- 241001465754 Metazoa Species 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 210000003128 head Anatomy 0.000 description 2
- 238000003706 image smoothing Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 241000927735 Penaeus Species 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000001000 micrograph Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- 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
- G06T2207/10004—Still image; Photographic 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
- G06T2207/30242—Counting objects in image
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention belongs to image procossing and applied technical fields, it is related to a kind of based on the shrimp seedling analysis of accounts method for improving TV-L1 models, for the enumeration problem for the shrimp seedling image that light halation in actual life is rung, image preprocessing is carried out using improved TV-L1 models to the image collected, the connected region of image after pretreatment is marked, find out the average area of shrimp seedling after pretreatment, shrimp seedling quantity is determined further according to the ratio of marked connected region area and average area in set relatively section, realize on there are adhesion and be illuminated by the light influence shrimp seedling quantity automatic detection, it is strong to solve shrimp body permeability, easily it is illuminated by the light influence, the problem of target is not readily separated with background, the random noise of image is eliminated simultaneously；Its method is simple, easy to operate, can directly handle single image and parameter setting is simple, counts accurate.
Description
Technical field：
The invention belongs to image procossing and applied technical fields, are related to a kind of based on the shrimp seedling counting for improving TV-L1 models
Analysis method carries out accurate metering in the case of illumination invariant to shrimp seedling.
Background technology：
As shrimp seedling cultivation scale constantly expands, the links such as raising, transport, sale need to carry out quantitative scoring to shrimp seedling
Number.Traditional method of counting mostly uses the samplings such as cup, bowl, bucket, is counted in the method for artificial naked eyes, there is very big mistake with the method
Difference takes time and effort and has a degree of damage to shrimp seedling.All kinds of shrimp seedling photoelectric counters, the fishes and shrimps seedling meter developed in recent years
Though number instrument etc. is convenient, easily influenced by channel size and shrimp seedling size, and expensive.Machine vision technique Yin Qifei connects
It touches, high-precision, quantifiable advantage are at present gradually applied to animal automatic identification counted fields, if document Liu generation is brilliant, king
General, Chen Jun wait based on the recognition methods of movement shrimp seedling [J] the agricultural engineerings for improving principal component analysis and AdaBoost algorithms
Report, 2017,33 (1):212-218；The behavioral value using machine vision technique development aquatic animal and knowledge there are also scholar
It Ji Shuo not study, such as document Ma H, Tsai T F, Liu C C.Real-time monitoring of water quality
Using temporal trajectory of live fish [M] .Pergamon Press, Inc. 2010 and Zhou Yingli,
Zeng Libo, Liu Juntang wait bacterium colony automatic counting method and its realization [J] data acquisition and procession of the based on image procossing,
2003,18(4):460-464；Document is imperial together, Jiang Yu, Li Zehua, waits magnaporthe grisea spores of the based on Micrograph image processing certainly
Dynamic detection and method of counting [J] Journal of Agricultural Engineering, 2015,31 (12):186-193 is proposed based on range conversion and Gauss
The improvement watershed algorithm of filtering detaches adhesion spore, meets magnaporthe grisea spore detection and count requirement automatically,
Zhou Yingli etc. proposes one kind and being based on range conversion and watershed algorithm separation of synechia bacterium colony, is identified using Edge track
It counts.These methods are mainly suitable for not being illuminated by the light influence and the typical object identification of feature counts.Related shrimp seedling inspection at present
It surveys the research report counted to can be counted on one's fingers, the research that shrimp seedling counts is primarily present problems with：When shrimp seedling have it is small,
The features such as translucency is strong is easily illuminated by the light influence, keeps its boundary information unintelligible；Second is that since shrimp seedling is distributed in different water depth, add
Translucency it is strong, make acquisition image there are adhesion, be of different shades the problems such as, utilize existing method carry out count results
Error is larger.
Invention content：
It is an object of the invention to overcome disadvantage of the existing technology, for the shrimp of light halation sound in actual life
The enumeration problem of seedling image, design provide a kind of illumination invariant shrimp seedling connected component labeling face based on improved TV-L1 models
Product analysis of accounts method, accurate metering analysis is carried out in the case of illumination invariant to shrimp seedling.
In order to achieve the above-mentioned object of the invention, present invention progress shrimp seedling analysis of accounts method includes the following steps：
(1) acquisition of image：Camera is first set in viewfinder range, camera position is adjusted, adjustment holder makes image just
Really display selects suitable camera scene mode further according to actual conditions, and sets camera to automatic adjustment focal length and light
Circle, automatic white balance are sampled under different illumination intensity indoors, and capture and preservation obtain observed image；
(2) image preprocessing：Detailed process is：
(2-1) improves the foundation of TV-L1 models：When being influenced there are larger inhomogeneous illumination in observed image, observation chart
The content of picture is often coated over dark region, and the image enhancement technique based on logarithmic transformation can effectively make low value gray scale
Image detail enhanced, by take logarithm in the way of to existing TV-L1 model refinements, specially：
It is obtained after carrying out logarithm enhancing to observed image f：
F=log f (1)
Using F as object to be studied,
E (u)=∫Ω|▽u|dxdy+∫Ωλ||u-F||dxdy (2)
In formula, f is the observed image of Noise, and u is the smoothed image obtained after TV-L1 model treatments, and ▽ is image
Gradient operator, λ are punishment parameter, and Ω is whole image region；Solution gradient drop equation obtains Jacobi iteration, and using existing
There are upwind difference and centered difference in technology to obtain：
Wherein,
Fi,jPosition is the pixel value of (i, j) in expression image,Restored picture position is after indicating nth iteration
The pixel value of (i, j), λ are punishment parameter, control the similarity of u and F, and step-length h values are 1, indicate neighbor pixel in image
Distance, ε values be 10-7, it is 0 to prevent denominator；
The processing of (2-2) image normalization：It is strong according to shrimp seedling image translucency, be not easy background and target image segmentation
Feature, using improved TV-L1 models as low-pass filtering operator estimate illumination component, define log-domain in observed image with
Using the quotient images of the smooth image obtained after improvement TV-L1 model treatments as the result images of unitary of illumination, specifically
For：
Unitary of illumination method analogue observation image F is the product of illumination component l and reflecting component r, by estimating illumination
Component, and then obtain reflecting component and estimate as the unrelated amount of illumination,
F=rl (8)
It uses improved TV-L1 models as low-pass filtering operator E, obtains illumination component and be estimated as
* signal convolution is indicated, re-defining logarithm quotient images (LQI) is
Defining quotient images (QI) is
In formula, u is obtained smoothed image after TV-L1 model treatments；The effect of quotient images unitary of illumination comes from vacation
If：The slow low-frequency component of wide variation is influenced as a result, with class low-pass filtering operator filtering by illumination variation in image
Obtained image is considered the image for reflecting illumination variation, therefore, quotient's operation pixel-by-pixel is done using smoothed image
Realize the normalization for illumination, the illumination invariant to obtain image indicates；
(2-3) morphological image process：First select the flat type disc structure element that a radius is 4 to illumination normalizing
Change treated image and carry out expansion process, then to select the flat type disc structure element that a radius is 9 to carry out image rotten
Operation is lost, the shrimp seedling being sticked together is separated；
(2-4) image binaryzation：Using document Ostu N, Nobuyuki O, Otsu N.A threshold
selection method from gray-level histogram IEEE transactions on systems[J]
.IEEE Transactions on Systems Man&Cybernetics,1979,9(1):62-66. between disclosed maximum kind
Variance method determines optimal threshold, divides image, is obtained to the image progress image binaryzation obtained after step (2-3) processing pre-
Treated image；
(3) shrimp seedling counts：First the connected region of image after pretreatment is marked, finds out shrimp seedling after pretreatment
Average area determines shrimp seedling further according to the ratio of marked connected region area and average area in set relatively section
Quantity, detailed process are：Connected region first is marked with sequential labelling method, it is a to calculate connected region number, then is calculated separately
The area for going out connected region sorts it from small to large, is stored in array S, assignment n1=0, n2=0；Then before countingIt is a
The area mean value of connected region is denoted as Avg；According to the Avg found out in advance, the connected region between 0.5Avg~1.5Avg is counted
Domain area, averaged Avg again, and be compared with original average value, Avg is constantly updated, this step is repeated, until
Average area Avg no longer changes, and acquires average value Avg, takes out element S [i] in S successively；If when S [i] < 0.3Avg, counting
It is 0；S if [i] > Avg, the quantity of shrimp seedlingN1=n1+K, wherein round are to round up to take
Integral function；If 0.3Avg≤S [i]≤Avg, n2=n2+1；As i > a, N=n1+n2 completes shrimp seedling and counts.
Compared with prior art, the present invention having studied detection and the method for counting of shrimp seedling with image processing techniques, realize
On there are adhesion and be illuminated by the light influence shrimp seedling quantity automatic detection, solve that shrimp body permeability is strong, is easily illuminated by the light shadow
The problem of sound, target and background are not readily separated, while eliminating the random noise of image；Its method is simple, easy to operate, energy
It directly handles single image and parameter setting is simple, count accurate.
Description of the drawings：
Fig. 1 is the workflow schematic block diagram of the present invention.
Fig. 2 is that shrimp seedling of the present invention counts flow diagram.
Fig. 3 is experimental facilities schematic diagram used in the embodiment of the present invention.
Fig. 4 is the shrimp seedling image of acquisition of the embodiment of the present invention, wherein (a1)-(a4) is shrimp seedling image under pale yellow light,
(b1)-(b4) is shrimp seedling image under half-light, and (c1)-(c4) is shrimp seedling image under more half-light, and (d1)-(d4) is shrimp under bright light
Seedling image.
Fig. 5 is the embodiment of the present invention to image preprocessing as a result, wherein (a) is sparse shrimp seedling image 5 (a) LQI processing
As a result, (b) being sparse shrimp seedling image 5 (a) histogram equalization processing as a result, (c) being sparse shrimp seedling image 5 (b) LQI
Processing as a result, (d) be sparse shrimp seedling image 5 (b) histogram equalization processing result.
Fig. 6 is that the embodiment of the present invention distinguishes each connected region area of the shrimp seedling image under 16 width difference illumination in Fig. 4
It is sorting in a manner of from small to large in line chart as a result, wherein abscissa indicates the number of connected region, ordinate indicates
Connected region area；Wherein (a) is shrimp seedling image under pale yellow light, is (b) shrimp seedling image under half-light, is (c) shrimp under more half-light
Seedling image (d) is shrimp seedling image under bright light.
Fig. 7 is the connected region centre plane of the shrimp seedling image under 16 width difference illumination in Fig. 4 described in the embodiment of the present invention
Product.
Specific implementation mode：
The invention will be further described by way of example and in conjunction with the accompanying drawings.
Embodiment：
The present embodiment experiment uses shrimp seedling for Penaeus Vannmei, and experimental facilities is as shown in Figure 1, including laptop, circle
Shape whiteware bowl, camera support and slr camera etc., experimental situation is：MATLAB R2014B, processor are：Intel(R)
Core (TM) i5-4590CPU@3.30GHZ 3.30GHZ, installation memory (RAM)：4.00GB carries out analysis of accounts to shrimp seedling
Detailed process is：
The first step：Image Acquisition, specific image acquisition process are：
(1) camera is set in viewfinder range, adjusts camera position, adjustment holder makes image correctly show；
(2) camera is set, and selects suitable camera scene mode；
(3) camera is set as automatic adjustment focal length and aperture, automatic white balance carry out under different illumination intensity indoors
Sampling；
(4) it captures and preserves, obtained shrimp seedling image is as shown in figure 4, as can be seen from Figure 4 acquisition shrimp seedling image has
Following feature：(1) image irradiation of shrimp seedling background is uneven, acquires image in different time and place, thus generates not
With the incident light of angle, causes shrimp seedling image background different, increase identification difficulty；(2) due to shrimp seedling self structure and light transmission
Property strong feature, cause the imaging gray value of shrimp seedling head, trunk, tail portion different, the imaging of especially trunk and tail portion is intimate
It combines together with background, increase extraction characteristic point and removes the difficulty of background；(3) part shrimp seedling is densely distributed, and exists viscous
Even, it increases segmentation and counts difficulty；
Second step：Image preprocessing
3.1 unitary of illumination
(1) TV-L1 models
TV-L1 models are the classical models in the image restoration fields of propositions such as Chan T F, Esedoglu S, main to solve
Restore the inversely processing problem of original image certainly from observed image, it is a kind of thin to preserve image while image smoothing
Section is the Regularization method of target；Chen et al. proposes to carry out the calculation of image layered and background verification using TV-L1 models
Method, therefore, to image filtering, can be imitated with reaching the image smoothing that edge is kept using TV-L1 models as low pass operator
Fruit：
E (u)=∫Ω|▽u|dxdy+∫Ωλ||u-f||dxdy (12)
In formula, f is the observed image of Noise, and u is image after restoring, and ▽ is image gradient operator, and λ is that punishment is joined
Number.First item is known as smooth item on the right of equation, plays smooth interaction to image, can remove noise；Section 2 is number on the right of equation
According to item, the similarity degree of image and observed image after restoring is indicated；
Equation is dropped by gradient：
?
After Jacobi is iterated to calculate：
Wherein,
In this way, observed image f obtains smoothed image u after iteration convergence；
(2) improved TV-L1 models
When being influenced there are larger inhomogeneous illumination in image, the content of image is often coated over dark region, base
The image detail of low value gray scale can be effectively set to be enhanced in the image enhancement technique of logarithmic transformation, therefore the present embodiment carries
Go out by take logarithm in the way of to TV-L1 model refinements.
For observed image f, obtained after carrying out logarithm enhancing：
F=log f (20)
It is used in combination F as object to be studied.
E (u)=∫Ω|▽u|dxdy+∫Ωλ||u-F||dxdy (21)
It is obtained through gradient drop equation solution：
Wherein c1、c2、c3、c4Value see formula (16) formula (17) formula (18) formula (19)；
(3) unitary of illumination
The unitary of illumination algorithm of image is that influencing for illumination is removed from inhomogeneous illumination image by image transformation
The image changed to no light.Existing researcher proposes many unitary of illumination algorithms at present.Wang et al. is proposed from quotient graph
As model, the model using the quotient images of observed image and Gaussian smoothing image as unitary of illumination as a result, this method can answer
For arbitrary single image, but Gaussian smoothing cannot keep the edge details in low frequency illumination field, the present embodiment root well
, be not easy divide background and target image the characteristics of strong according to shrimp seedling image translucency, proposition using improved TV-L1 models as
Low-pass filtering operator estimate illumination component, define log-domain in observed image with using improvement TV-L1 model treatments after obtain
Smooth image result images of the quotient images as unitary of illumination：
Unitary of illumination method analogue observation image F is the product of illumination component l and reflecting component r on the whole, is passed through
Estimate illumination component, and then obtain reflecting component as the unrelated amount estimation of illumination, therefore, the present embodiment has
F=rl (22)
It uses improved TV-L1 models as low-pass filtering operator E, obtains illumination component and be estimated as
* signal convolution is indicated.Defining logarithm quotient images (LQI) is
Defining quotient images (QI) is
In formula, u is obtained smoothed image after TV-L1 model treatments, and the effect of quotient images unitary of illumination comes from vacation
If：The slow low-frequency component of wide variation is influenced as a result, with class low-pass filtering operator filtering by illumination variation in image
Obtained image is considered the image for reflecting illumination variation, therefore, it is real to do quotient's operation pixel-by-pixel with smoothed image
The normalization for illumination is showed, the illumination invariant to obtain image indicates.
It uses TV-L1 models to estimate illumination variation image as low-pass filtering operator in image log domain, can eliminate common low
Logical operator acts on the halation phenomenon that high-frequency region leaves, while removing partial noise, and has certain edge retention performance,
Parameter setting is also fairly simple simultaneously, and only there are one parameter lambdas, can obtain clear image by adjusting λ, the numerical value of λ is bigger, right
The punishment of data item is bigger, and image is more clear, and image can be enable to preferably keep edge spy when it works with smooth item one
Property；During actual tests using histogram equalization in the prior art can only it is limited improve image contrast, and
TV-L1 models are used to be carried out as low-pass filtering operator from quotient's (observed image is quotient with smoothed image) merely though handling energy
Illumination effect is removed, but there are partial noises, the pixel difference between taking logarithm that can reduce gray value image is away from extending
Low value gray compression high level gray scale, then handle from quotient, improvement with obvious effects.
3.2 morphological image process
The application of mathematical morphology can simplify image data, and while keeping their basic configuration characteristics, removal is not
Coherent structure, the present embodiment find that there is also adhesion, edges for shrimp seedling after on image eliminates halation and partial impurities influence
The problems such as burr, therefore the present embodiment is in the image basis that unitary of illumination is handled, first select a radius be 4 it is flat
Type disc structure element to image carry out expansion process, then select a radius be 9 flat type disc structure element to image
Erosion operation is carried out, is separated the shrimp seedling being sticked together with this；
3.3 image binaryzation
It, must be by the shrimp seedling image binaryzation of different gray-scale pixels, in binarization most to keep subsequent counter more acurrate
It is important that threshold value On The Choice can be automatically obtained proper if by the way of threshold value iteration by program
Threshold value, although many experiments show that the mode of threshold value iteration is simple, but there is no good in the slight part of image
Discrimination, in certain specific images, the variation of tiny data can cause the larger change of segmentation effect, find in testing,
The maximum variance between clusters proposed by OTSU0Easy to operate, processing speed is fast, and preferable to the treatment of details of image when segmentation,
Therefore the present embodiment determines optimal threshold using this method, divides image；
4 shrimp seedlings count
Based on existing characteristics of image, connected region number is directly read as final data to pretreated image
As a result, it is clear that be difficult to obtain higher accuracy rate, shrimp seedling is not of uniform size, and have there are adhesion, it is small if corrosion scale is excessive
Area shrimp seedling can be eliminated, if corrosion scale is too small, adhesion region is difficult to separate, in order to handle the size of shrimp seedling feature area
First connected region is marked for the problem of differing, the present embodiment, finds out the average area of shrimp seedling after pretreatment, further according to
The set ratio of marked connected region area and average area relatively in section determines shrimp seedling quantity.
The calculating of 4.1 average areas, is as follows:
(1) connected component labeling algorithm is utilized, image is traversed using 8 connected regions, seeks connected region number, be denoted as
a；
(2) connected region is arranged by ascending order, before only countingThe area mean value of a connected region is denoted as Avg
(3) according to the Avg found out in advance, the connected region area between 0.5Avg~1.5Avg is counted, is sought again
Average value Avg, and being compared with original average value constantly updates Avg, repeats this step, until average area Avg no longer
Variation；
Processing strategy when 4.2 counting
When shrimp seedling counts, for different size of connected region, processing mode is as follows：
(1) during Morphological scale-space, shrimp seedling eyes are detached with body, and cause counting error.When connected region face
When product S < 0.3Avg, not count；
(2) in pre-processing, part shrimp seedling excessive corrosion, for the part of 0.3Avg≤S in image≤Avg, according to one
Shrimp seedling calculates；
(3) area of shrimp seedling adhesion problems, connected region should generally meet Avg < S, and the quantity K of shrimp seedling is according to formula at this time
(26) it calculates：
Wherein, K round numbers, S [i] are connected region area, and round is round function；
4.3 count flow
It is as follows that shrimp seedling specifically counts flow：
(1) connected region is marked with sequential labelling method, it is a to calculate connected region number；
(2) area for calculating separately out connected region sorts it from small to large, is stored in array S, assignment n1=0, n2
=0；
(3) according to the method for update Avg used in 4.1, average Avg, takes out element S [i] in S successively；
(4) if when S [i] < 0.3Avg, it is counted as 0；If S [i] > Avg,n1
=n1+K；If 0.3Avg≤S [i]≤Avg, n2=n2+1；
(5) as i > a, N=n1+n2, algorithm terminates, and completes shrimp seedling and counts.
5 experimental results and analysis
Carry out that treated that the results are shown in Figure 5 using the shrimp seedling automatic counting method of the present embodiment, wherein (a) and (c)
It is sparse shrimp and intensive shrimp to the pretreated result images of LQI, is (b) and (d) sparse shrimp and intensive shrimp histogram equalization
Change pretreated result images, influence is notable, and Fig. 6 is that 16 width in Fig. 4 are different it can be seen that experimental result is illuminated by the light by image
It is that each connected region area of shrimp seedling image under illumination sorts in line chart in a manner of from small to large respectively as a result, its
Middle abscissa indicates that the number of connected region, ordinate indicate connected region area；The present embodiment will to LQI treated knot
Fruit image and the result images of histogram equalization processing have been contrast experiment, 16 width difference illumination that experiment sample is Fig. 4,
Form, the results are shown in Table 1 for the shrimp seedling picture count of number,
Table 1:Shrimp seedling count results
Wherein relative error formula is
The calculation formula of Average Accuracy is
The experimental results showed that：In 16 width images of test, the number of each image shrimp seedling is 4-167, at LQI
The Average Accuracy 99.276% that shrimp seedling detects after reason is higher than the shrimp seedling of histogram equalization processing detection average rate
27.76%；The result detected to shrimp seedling after LQI processing is shown:There are 11 width images that any error does not occur, this few width image
The characteristics of be:Image after segmentation is refined into isolated profile, will not be had an impact to counting；There is the counting knot of 3 width images
The characteristics of 1 to 4 fewer than artificial counting of fruit, this image is that some shrimp seedlings are distributed in edge, is caused after being pre-processed to LQI
The relatively other shrimp seedling pixels of shrimp seedling than thin, be easy to be ignored in adaptive threshold fuzziness；The meter of other residual images
Result is counted then 2 more than artificial counting, and there are the shrimp seedlings that quantity is 32 in (a1-a4) and (b1-b4) in Fig. 4, but
But there is difference in automatic count results, illustrate that experimental result is illuminated by the light, shrimp seedling is distributed, the influence etc. of shrimp seedling adhesion degree, warp
Analysis is found, in image acquisition process, because of shooting angle reason, causes shrimp seedling not of uniform size, more relatively large shrimp seedling,
It is easy head being divided into two parts with body during morphology, and this two parts size is identical, easily causes counting error.
Claims (1)
1. a kind of based on the shrimp seedling analysis of accounts method for improving TV-L1 models, it is characterised in that include the following steps：
(1) acquisition of image：Camera is first set in viewfinder range, camera position is adjusted, adjustment holder makes image correctly show
Show, selects suitable camera scene mode further according to actual conditions, and set camera to automatic adjustment focal length and aperture, automatically
White balance is sampled under different illumination intensity indoors, and capture and preservation obtain observed image；
(2) image preprocessing：Detailed process is：
(2-1) improves the foundation of TV-L1 models：When being influenced there are larger inhomogeneous illumination in observed image, observed image
Content is often coated over dark region, and the image enhancement technique based on logarithmic transformation can effectively make the image of low value gray scale
Details is enhanced, by take logarithm in the way of to existing TV-L1 model refinements, specially：
It is obtained after carrying out logarithm enhancing to observed image f：
F=logf (1)
Using F as object to be studied,
In formula, f is the observed image of Noise, and u is obtained smoothed image after TV-L1 model treatments,It is calculated for image gradient
Son, λ are punishment parameter, and Ω is whole image region；Solution gradient drop equation obtains Jacobi iteration, and using in the prior art
Upwind difference and centered difference obtain：
Wherein,
Fi,jPosition is the pixel value of (i, j) in expression image,Restored picture position is (i, j) after indicating nth iteration
Pixel value, λ are punishment parameter, control the similarity of u and F, and step-length h values are 1, indicate the distance of neighbor pixel in image, ε
Value is 10-7, it is 0 to prevent denominator；
The processing of (2-2) image normalization：, be not easy background and target image segmentation the characteristics of strong according to shrimp seedling image translucency,
Estimate illumination component using improved TV-L1 models as low-pass filtering operator, the observed image defined in log-domain changes with utilization
Result images of the quotient images of the smooth image obtained after into TV-L1 model treatments as unitary of illumination, specially：
Unitary of illumination method analogue observation image F is the product of illumination component l and reflecting component r, by estimating illumination component,
And then obtain reflecting component and estimate as the unrelated amount of illumination,
F=rl (8)
It uses improved TV-L1 models as low-pass filtering operator E, obtains illumination component and be estimated as
* signal convolution is indicated, re-defining logarithm quotient images LQI is
Defining quotient images QI is
In formula, u is obtained smoothed image after TV-L1 model treatments；The effect of quotient images unitary of illumination comes from hypothesis：Figure
The slow low-frequency component of wide variation is to be influenced by illumination variation as a result, being obtained with class low-pass filtering operator filtering as in
Image is considered the image for reflecting illumination variation, therefore, does quotient's operation pixel-by-pixel using smoothed image and realizes needle
Normalization to illumination, the illumination invariant to obtain image indicate；
(2-3) morphological image process：First select the flat type disc structure element that a radius is 4 to unitary of illumination processing
Image afterwards carries out expansion process, then the flat type disc structure element that a radius is 9 is selected to carry out erosion operation to image,
The shrimp seedling being sticked together is separated；
(2-4) image binaryzation：Using document Ostu N, Nobuyuki O, Otsu N.A threshold selection
method from gray-level histogram IEEE transactions on systems[J].IEEE
Transactions on Systems Man&Cybernetics,1979,9(1):62-66. disclosed maximum variance between clusters
It determines optimal threshold, divides image, the image progress image binaryzation obtained after step (2-3) processing is obtained pretreated
Image；
(3) shrimp seedling counts：First the connected region of image after pretreatment is marked, finds out being averaged for shrimp seedling after pretreatment
Area determines shrimp seedling quantity further according to the ratio of marked connected region area and average area in set relatively section,
Detailed process is：Connected region first is marked with sequential labelling method, it is a to calculate connected region number, then calculates separately out connection
The area in region sorts it from small to large, is stored in array S, assignment n1=0, n2=0；Then before countingA connected region
Area mean value be denoted as Avg；According to the Avg found out in advance, the connected region area between 0.5Avg~1.5Avg is counted, weight
New averaged Avg, and be compared with original average value, Avg is constantly updated, this step is repeated, until average area Avg
No longer change, acquire average value Avg, takes out element S [i] in S successively；When S if [i] < 0.3Avg, it is counted as 0；S if [i] >
Avg, the then quantity of shrimp seedlingN1=n1+K, wherein round are round function；If
0.3Avg≤S [i]≤Avg, then n2=n2+1；As i > a, N=n1+n2 completes shrimp seedling and counts.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810522933.5A CN108765448B (en) | 2018-05-28 | 2018-05-28 | Shrimp larvae counting analysis method based on improved TV-L1 model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810522933.5A CN108765448B (en) | 2018-05-28 | 2018-05-28 | Shrimp larvae counting analysis method based on improved TV-L1 model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108765448A true CN108765448A (en) | 2018-11-06 |
| CN108765448B CN108765448B (en) | 2022-01-07 |
Family
ID=64002960
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810522933.5A Active CN108765448B (en) | 2018-05-28 | 2018-05-28 | Shrimp larvae counting analysis method based on improved TV-L1 model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108765448B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112257564A (en) * | 2020-10-20 | 2021-01-22 | 青岛丰禾星普科技有限公司 | Aquatic product quantity statistical method, terminal equipment and storage medium |
| CN113139934A (en) * | 2021-03-26 | 2021-07-20 | 上海师范大学 | Rice grain counting method |
| CN113812370A (en) * | 2021-11-02 | 2021-12-21 | 广东科贸职业学院 | Shrimp larvae counting device and method based on visual identification technology |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW384451B (en) * | 1996-05-02 | 2000-03-11 | Lin Yi Hung | Counting system for shrimp and fish fry using mechanical visualization |
| US20070165961A1 (en) * | 2006-01-13 | 2007-07-19 | Juwei Lu | Method And Apparatus For Reducing Motion Blur In An Image |
| CN103700117A (en) * | 2013-11-21 | 2014-04-02 | 北京工业大学 | Robust optical flow field estimating method based on TV-L1 variation model |
| US20150016669A1 (en) * | 2013-07-15 | 2015-01-15 | Digitalglobe, Inc. | Automated remote car counting |
| CN106296749A (en) * | 2016-08-05 | 2017-01-04 | 天津大学 | RGB D image eigen decomposition method based on L1 norm constraint |
| WO2017054775A1 (en) * | 2015-09-30 | 2017-04-06 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for determining a breast region in a medical image |
| CN106959133A (en) * | 2017-03-23 | 2017-07-18 | 贵州省机械电子产品质量监督检验院 | A kind of LED illumination product photo-biological safety detection system |
| WO2017221259A1 (en) * | 2016-06-23 | 2017-12-28 | S Jyothi | Automatic recognition of indian prawn species |
-
2018
- 2018-05-28 CN CN201810522933.5A patent/CN108765448B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW384451B (en) * | 1996-05-02 | 2000-03-11 | Lin Yi Hung | Counting system for shrimp and fish fry using mechanical visualization |
| US20070165961A1 (en) * | 2006-01-13 | 2007-07-19 | Juwei Lu | Method And Apparatus For Reducing Motion Blur In An Image |
| US20150016669A1 (en) * | 2013-07-15 | 2015-01-15 | Digitalglobe, Inc. | Automated remote car counting |
| CN103700117A (en) * | 2013-11-21 | 2014-04-02 | 北京工业大学 | Robust optical flow field estimating method based on TV-L1 variation model |
| WO2017054775A1 (en) * | 2015-09-30 | 2017-04-06 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for determining a breast region in a medical image |
| WO2017221259A1 (en) * | 2016-06-23 | 2017-12-28 | S Jyothi | Automatic recognition of indian prawn species |
| CN106296749A (en) * | 2016-08-05 | 2017-01-04 | 天津大学 | RGB D image eigen decomposition method based on L1 norm constraint |
| CN106959133A (en) * | 2017-03-23 | 2017-07-18 | 贵州省机械电子产品质量监督检验院 | A kind of LED illumination product photo-biological safety detection system |
Non-Patent Citations (7)
| Title |
|---|
| GANG LIU: "High-order TVL1-based images restoration and spatially adapted regularization parameter selection", 《COMPUTERS & MATHEMATICS WITH APPLICATIONS》 * |
| T. CHEN: "Illumination normalization for face recognition and uneven background correction using total variation based image models", 《2005 IEEE COMPUTER SOCIETY CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| 刘世晶: "基于改进主成分分析和AdaBoost算法的运动虾苗识别力法", 《农业工程学报》 * |
| 张熠: "基于总变分模型的光照不变人脸识别算法", 《万方数据知识服务平台》 * |
| 杨慧赞: "基于图像处理的鱼卵计数方法研究", 《万方数据知识服务平台》 * |
| 王硕: "基于计算机视觉的大菱鲆鱼苗计数方法研究", 《万方数据知识服务平台》 * |
| 陈基伟: "基于数字图像处理的棒材计数方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112257564A (en) * | 2020-10-20 | 2021-01-22 | 青岛丰禾星普科技有限公司 | Aquatic product quantity statistical method, terminal equipment and storage medium |
| CN112257564B (en) * | 2020-10-20 | 2022-07-12 | 青岛丰禾星普科技有限公司 | Aquatic product quantity statistical method, terminal equipment and storage medium |
| CN113139934A (en) * | 2021-03-26 | 2021-07-20 | 上海师范大学 | Rice grain counting method |
| CN113139934B (en) * | 2021-03-26 | 2024-04-30 | 上海师范大学 | Rice grain counting method |
| CN113812370A (en) * | 2021-11-02 | 2021-12-21 | 广东科贸职业学院 | Shrimp larvae counting device and method based on visual identification technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108765448B (en) | 2022-01-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104834912B (en) | A kind of weather recognition methods and device based on image information detection | |
| CN109447945B (en) | Quick counting method for basic wheat seedlings based on machine vision and graphic processing | |
| Dewi et al. | Classification of polycystic ovary based on ultrasound images using competitive neural network | |
| Sanchez-Torres et al. | Automatic measurement of fish weight and size by processing underwater hatchery images | |
| Pang et al. | Automatic segmentation of crop leaf spot disease images by integrating local threshold and seeded region growing | |
| Sunny et al. | An efficient citrus canker detection method based on contrast limited adaptive histogram equalization enhancement | |
| CN112598713A (en) | Offshore submarine fish detection and tracking statistical method based on deep learning | |
| Qing et al. | Automated detection and identification of white-backed planthoppers in paddy fields using image processing | |
| CN111179216B (en) | Crop disease identification method based on image processing and convolutional neural network | |
| CN108765448A (en) | A kind of shrimp seedling analysis of accounts method based on improvement TV-L1 models | |
| Jadhav et al. | Soybean leaf disease detection and severity measurement using multiclass SVM and KNN classifier | |
| CN111879735B (en) | Rice appearance quality detection method based on image | |
| Feng et al. | A separating method of adjacent apples based on machine vision and chain code information | |
| CN105678245A (en) | Target position identification method based on Haar features | |
| Gupta et al. | Automated segmentation of powdery mildew disease from cherry leaves using image processing | |
| CN113204990B (en) | Machine learning method and device based on intelligent fishing rod | |
| Ni et al. | Convolution neural network based automatic corn kernel qualification | |
| Zhu et al. | Corn leaf diseases diagnostic techniques based on image recognition | |
| CN116246174B (en) | Sweet potato variety identification method based on image processing | |
| CN113269690A (en) | Method and system for detecting diseased region of blade | |
| Miao et al. | Research on Soybean Disease Identification Method Based on Deep Learning | |
| Gou et al. | A novel retinal vessel extraction method based on dynamic scales allocation | |
| Zhang et al. | Automatic Segmentation of Remote Sensing Images on Water Bodies Based on Image Enhancement. | |
| Büyükarıkan et al. | Convolutional neural network-based apple images classification and image quality measurement by light colors using the color-balancing approach | |
| CN113269756A (en) | Retina blood vessel segmentation method and device based on multi-scale matched filtering and particle swarm optimization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |