CN108318433A - The method for detecting water injection rate in salmon using hyper-spectral image technique - Google Patents
The method for detecting water injection rate in salmon using hyper-spectral image technique Download PDFInfo
- Publication number
- CN108318433A CN108318433A CN201810068193.2A CN201810068193A CN108318433A CN 108318433 A CN108318433 A CN 108318433A CN 201810068193 A CN201810068193 A CN 201810068193A CN 108318433 A CN108318433 A CN 108318433A
- Authority
- CN
- China
- Prior art keywords
- salmon
- sample
- injection rate
- water injection
- hyper
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
Abstract
The invention discloses a kind of methods detecting water injection rate in salmon using hyper-spectral image technique, include the following steps：A：Sample collection；B, sample treatment；C, sample EO-1 hyperion wave band screens；D, sample EO-1 hyperion feature extraction；E, model is established；F, model evaluation.The method that the present invention detects water injection rate in salmon using hyper-spectral image technique, detection method operating process is without pre-processing sample, it is easy to operate, it is reproducible, analysis time is short, analytic process need to be to salmon stripping and slicing and water filling, after establishing the network model for detection, need to only be scanned to obtain required characteristic under spectrum to sample to be tested can detect salmon water injection rate by network model, for non-intrusion measurement method, testing result is quick and precisely, and it is sufficiently stable, effectively raise detection efficiency, can meet the needs of sample is quickly detected in production scene.
Description
Technical field
The present invention relates to be related to high spectrum image detection field more particularly to a kind of utilization hyper-spectral image technique detection three
The method of water injection rate in literary fish.
Background technology
In the prior art, salmon, which is also named to spread, covers fish or Sa Menyu, is one of fish raw material more common in western-style food, not
Consumption market salmon with country covers different types, and Norway's salmon is mainly Atlantic salmon, and Finland's salmon is main
It is the big specification red meat rainbow trout of cultivation, the salmon in the U.S. is mainly Alaska salmon, and dog salmon refers generally to salmon shape mesh salmon
The fish that section Pacific Ocean salmon belongs to, there are many kinds of, such as China northeast production dog salmon and bow-backed dog salmon, salmon is with very
High nutritive value, the abundant unsaturated fatty acid contained in the flesh of fish can reduce cholesterol and blood fat, can be effectively reduced
The incidence of angiocardiopathy, with the development of global economy and the continuous improvement of living standards of the people, three text of world market pair
The demand of fish is increasingly vigorous, and economic interests caused by ever-expanding demand promote national governments and fisherman constantly to expand
Production scale, however while salmon industry fast development, the fraud problem of salmon also becomes getting worse, a part
Businessman replaces salmon to cheat consumer with other flesh of fish, and also businessman's water filling in the salmon flesh of fish improves net weight to try to gain
Profit, the flesh of fish mouthfeel after water filling decline, and the content of microorganisms in the water filling flesh of fish severely exceeds, and the chemical substance in water can be to human body
Cause very major injury.
Currently, measure whether the method for water filling is mainly Organoleptic method, carried out by the observation of human eye and the touch of hand
Judge, photosensitive analytic approach is simple and easy to do, but is susceptible to the influence of people's subjective factor, and can not be to whether standard is made in water filling
True judgement explores a kind of quick nondestructive, whether the method for water filling is very necessary to detection salmon immediately due to the above reasons,
, test paper method can quickly and effectively detect salmon whether water filling, but accurate judgement can not be made to water injection rate.
Invention content
The object of the present invention is to provide a kind of methods detecting water injection rate in salmon using hyper-spectral image technique.
Present invention technical solution used for the above purpose is：It is a kind of to utilize three text of hyper-spectral image technique detection
The method of water injection rate, includes the following steps in fish：
A：Sample collection：Acquire salmon sample；
B, sample treatment:Salmon sample is sliced, is cut into small pieces, and is implanted sequentially different amounts of water；
C, sample EO-1 hyperion wave band screens：The acquisition of spectroscopic data is carried out to the salmon block of different injection amount, chosen area
To full spectral signature wave band curve, using Principal Component Analysis progress Data Dimensionality Reduction, selected characteristic wave band and by its principal component figure
As obtaining weight coefficient figure；
D, sample EO-1 hyperion feature extraction：The extraction that color moment is carried out to the principal component image of selection, is used in combination gray level co-occurrence matrixes
Extract the textural characteristics of image；
E, model is established：Obtained color characteristic and textural characteristics are input to water injection rate in data model, water filling is carried out
Measure the foundation of model；
F, model evaluation：Using established model, color characteristic and the textural characteristics of input sample obtain water injection rate discriminating, root
Accuracy rate is obtained according to model, network model is assessed.
Easily there is the sample faked using relatively conventional in the market in salmon sample in the step A.
It is 1cm X 1cm X 1cm fritters that salmon sample, which is cut into size, in the step B, totally 60 pieces.
Stripping and slicing and water filling are carried out to salmon in the step B, salmon flesh of fish water filling after tested is more than that 0.6g has water
It overflows, to the salmon flesh of fish in 0-0.6g random injections.
It is 50cm, time for exposure 9ms, wavelength model that the parameter of spectrum data gathering, which is set as object distance, in the step C
It encloses for 379nm-1038nm, spectrum spacing is 1.34nm, line speed 5mm/s.
Model foundation uses Self-organizing Maps algorithm in the step E.
A kind of method detecting water injection rate in salmon using hyper-spectral image technique of the present invention, detection method operating process
Easy to operate without being pre-processed to sample, reproducible, analysis time is short, and analytic process to salmon stripping and slicing and need to be noted
Water need to only be scanned sample to be tested to obtain required characteristic under spectrum after establishing the network model for detection
Salmon water injection rate can be detected by network model, be non-intrusion measurement method, testing result quick and precisely, and very
Stablize, effectively raises detection efficiency, can meet the needs of sample is quickly detected in production scene.
Description of the drawings
Fig. 1 is that a kind of the different of method acquisition for being detected water injection rate in salmon using hyper-spectral image technique of the present invention are noted
The averaged spectrum curve graph of the salmon sample area-of-interest of water.
Fig. 2 is that a kind of the different of method acquisition for being detected water injection rate in salmon using hyper-spectral image technique of the present invention are noted
The main component PC1 images of the salmon sample of water.
Fig. 3 is that a kind of the different of method acquisition for being detected water injection rate in salmon using hyper-spectral image technique of the present invention are noted
The main component PC2 images of the salmon sample of water.
Fig. 4 is that a kind of the different of method acquisition for being detected water injection rate in salmon using hyper-spectral image technique of the present invention are noted
The main component PC3 images of the salmon sample of water.
Fig. 5 is that a kind of the different of method acquisition for being detected water injection rate in salmon using hyper-spectral image technique of the present invention are noted
The main component PC4 images of the salmon sample of water.
Fig. 6 is that a kind of the different of method acquisition for being detected water injection rate in salmon using hyper-spectral image technique of the present invention are noted
The main component PC5 images of the salmon sample of water.
Fig. 7 is that a kind of the different of method acquisition for being detected water injection rate in salmon using hyper-spectral image technique of the present invention are noted
The main component PC6 images of the salmon sample of water.
Fig. 8 is that a kind of method detecting water injection rate in salmon using hyper-spectral image technique of the present invention passes through characteristic wave bands
Under principal component image gray scale be worth to without water filling image.
Fig. 9 is that a kind of method detecting water injection rate in salmon using hyper-spectral image technique of the present invention passes through characteristic wave bands
Under principal component image the water injection rate that is worth to of gray scale 0g-0.2g image.
Figure 10 is that a kind of method detecting water injection rate in salmon using hyper-spectral image technique of the present invention passes through characteristic wave
Image of the water injection rate that the gray scale of principal component image under section is worth in 0.2-0.4g.
Figure 11 is that a kind of method detecting water injection rate in salmon using hyper-spectral image technique of the present invention passes through characteristic wave
Image of the water injection rate that the gray scale of principal component image under section is worth in 0.4-0.6g.
Figure 12 is that a kind of method detecting water injection rate in salmon using hyper-spectral image technique of the present invention passes through self-organizing
Mapping algorithm, the classification results figure of the network model 1 of foundation.
Figure 13 is that a kind of method detecting water injection rate in salmon using hyper-spectral image technique of the present invention passes through self-organizing
Mapping algorithm, the classification results figure of the network model 2 of foundation.
Figure 14 is that a kind of method detecting water injection rate in salmon using hyper-spectral image technique of the present invention passes through self-organizing
Mapping algorithm, the classification results figure of the network model 3 of foundation.
Figure 15 is a kind of the first six master detecting the method for water injection rate in salmon using hyper-spectral image technique of the present invention
Ingredient cumulative proportion in ANOVA table.
Figure 16 is that a kind of model detecting the method for water injection rate in salmon using hyper-spectral image technique of the present invention is accurate
Rate table.
Specific implementation mode
As shown in Fig. 1 to Figure 14, using hyper-spectral image technique detect salmon in water injection rate method, specifically include with
Lower step, sample collection：It buys in batches and salmon sample and carries out stripping and slicing, fish block size is 1cm X 1cm X 1cm, and totally 60
Block；Sample measures：Sample water injection rate measures, and in batches 60 samples are injected with the water of 0-0.6g at random, is put into after each water filling
It weighs on balance, is as a result accurate to 2 significant digits（Unit g）；The acquisition of hyperspectral image data：By 80 three texts
Fish sample, which is put under high spectrum image acquisition system, acquires spectroscopic data, acquires spectroscopic data after each water filling, the high-spectrum
As acquisition system mainly include high light spectrum image-forming unit, light source, automatically controlled mobile platform and equipped with control system computer composition,
Wherein high light spectrum image-forming unit includes CCD camera and image light spectrometer two parts, and light source is provided by 150W halogen lamp, is terrible
To best spectrum picture, the systematic parameter after debugging repeatedly is as follows：Time for exposure 10ms, object distance 50cm, platform movement
Speed 5mm/s, wave-length coverage 379nm-1038nm, the area-of-interest for choosing acquisition salmon sample as shown in Figure 1 obtain entirely
Averaged spectrum curve graph under wave band carries out dimensionality reduction, six obtained principal component figures using Principal Component Analysis to spectroscopic data
As shown in Figures 2 to 7, the first six principal component contributor rate known to Figure 15 tables selects principal component image of the characteristic value more than 1 to carry out
Analysis compares PC1 images and PC2 images because the contribution rate of accumulative total of the principal component of the first two has been able to reach 90% or more,
PC1 brightness of image is excessively high, and water flooding feature reflects unobvious, and PC2 images can be good at the water flooding feature for reflecting sample, pass through
Pseudo-color conversion such as Fig. 8 to Figure 11 is carried out to the PC2 images of sample, it can be seen that the PC2 images of different injection amount have significant change,
So selection PC2 images are handled, software used in process is ENVI 5.0（Research System Inc, USA）；It establishes
Model：Contrast, correlation, mean value, the variance that the PC2 images under sample principal component analysis are extracted using gray level co-occurrence matrixes are made
For textural characteristics, the first moment, second moment of the color of PC2 images under sample principal component analysis are extracted as color characteristic, model
Neuron output is divided into 0,1,2,3 four classes according to the difference of water injection rate, and 0 indicates no water filling, and 1 indicates water injection rate in 0g-0.2g
Between, 2 indicate water injection rate between 0.2g-0.4g, and 3 indicate water injection rate between 0.4g-0.6g, not according to input parameter
Together, three groups of network models are established：Model one is established and is based on from group using the textural characteristics of salmon sample as input parameter
Knit mapping（SOM）Salmon water injection rate detection model；Model two, using the color characteristic of salmon sample as input parameter,
It establishes and is based on Self-organizing Maps（SOM）Salmon water injection rate detection model；Model three merges the textural characteristics of salmon sample
With color characteristic as input parameter, establishes and be based on Self-organizing Maps（SOM）Salmon water injection rate detection model, pass through detection
As a result comparison, the accuracy rate of computation model detection, determines optimal detection network model, and software used in the process is Matlab
2014b（The MathWorksInc, USA）, 60 sample averages are divided into 4 groups, water injection rate is distributed as 0,0-0.2g,
0.2-0.4g, 0.4-0.6g, with this exemplar texture be characterized as input structure model one, sample four neurons distribution such as
Shown in Figure 12, the corresponding number of four neurons of testing result and actual number（Every group 15）The mould can be calculated by being compared
The Detection accuracy of type, the Detection accuracy for obtaining established network model are 75.36%, similarly, special with the color of 4 groups of samples
Sign for input structure model two, obtain detection sample it is as shown in figure 13 in the distribution of 4 neurons, calculate established network mould
The Detection accuracy of type is 90.83%, can accurately detect the water injection rate of sample, special with 4 groups of exemplar textures and color blend
Sign for input structure model three, obtain detection sample it is as shown in figure 14 in the distribution of 4 neurons, calculate established network mould
The Detection accuracy of type is 96.67%；The evaluation of model：By the intuitive comparison of Figure 16 tables it is found that the line for passing through principal component image
It is 75.36% to manage one accuracy rate of SOM network models that feature is established, and accuracy rate is relatively low, is established according to the color characteristic of image
Two rate of accuracy reached of network model to 90.83%, can the accurate detection salmon flesh of fish water injection rate, pass through texture color
Three Detection accuracy of model that Fusion Features are established reaches 96.67%, very can accurately detect the water filling feelings of salmon sample
Therefore condition selects model 3 to establish salmon flesh of fish water filling detection model, the present invention can be obtained by the verification to network model
The network model for the detection salmon water filling established to method using the present invention, passes through building for textural characteristics and color characteristic
Mould results contrast, the obtained network model of color characteristic of principal component image can be very good to carry out salmon sample to be measured
Differentiating, detection process only need to carry out high spectrum image scanning to sample, and whole operation process will not destroy sample to be tested,
It is easy to operate, it can effectively improve detection speed.
Claims (6)
1. a kind of method detecting water injection rate in salmon using hyper-spectral image technique, which is characterized in that include the following steps：
A：Sample collection：Acquire salmon sample；
B, sample treatment:Salmon sample is sliced, is cut into small pieces, and is implanted sequentially different amounts of water；
C, sample EO-1 hyperion wave band screens：The acquisition of spectroscopic data is carried out to the salmon block of different injection amount, chosen area
To full spectral signature wave band curve, using Principal Component Analysis progress Data Dimensionality Reduction, selected characteristic wave band and by its principal component figure
As obtaining weight coefficient figure；
D, sample EO-1 hyperion feature extraction：The extraction that color moment is carried out to the principal component image of selection, is used in combination gray level co-occurrence matrixes
Extract the textural characteristics of image；
E, model is established：Obtained color characteristic and textural characteristics are input to water injection rate in data model, water filling is carried out
Measure the foundation of model；
F, model evaluation：Using established model, color characteristic and the textural characteristics of input sample obtain water injection rate discriminating, root
Accuracy rate is obtained according to model, network model is assessed.
2. a kind of method detecting water injection rate in salmon using hyper-spectral image technique according to claim 1, special
Sign is：Easily there is the sample faked using relatively conventional in the market in salmon sample in the step A.
3. a kind of method detecting water injection rate in salmon using hyper-spectral image technique according to claim 1, special
Sign is：It is 1cm X 1cm X 1cm fritters that salmon sample, which is cut into size, in the step B, totally 60 pieces.
4. a kind of method detecting water injection rate in salmon using hyper-spectral image technique according to claim 1, special
Sign is：Stripping and slicing and water filling are carried out to salmon in the step B, salmon flesh of fish water filling after tested is more than that have water excessive by 0.6g
Go out, to the salmon flesh of fish in 0-0.6g random injections.
5. a kind of method detecting water injection rate in salmon using hyper-spectral image technique according to claim 1, special
Sign is：It is 50cm, time for exposure 9ms, wave-length coverage that the parameter of spectrum data gathering, which is set as object distance, in the step C
For 379nm-1038nm, spectrum spacing is 1.34nm, line speed 5mm/s.
6. a kind of method detecting water injection rate in salmon using hyper-spectral image technique according to claim 1, special
Sign is：Model foundation uses Self-organizing Maps algorithm in the step E.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810068193.2A CN108318433B (en) | 2018-01-24 | 2018-01-24 | Method for detecting water injection amount in salmon by utilizing hyperspectral image technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810068193.2A CN108318433B (en) | 2018-01-24 | 2018-01-24 | Method for detecting water injection amount in salmon by utilizing hyperspectral image technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108318433A true CN108318433A (en) | 2018-07-24 |
| CN108318433B CN108318433B (en) | 2021-02-23 |
Family
ID=62887647
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810068193.2A Active CN108318433B (en) | 2018-01-24 | 2018-01-24 | Method for detecting water injection amount in salmon by utilizing hyperspectral image technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108318433B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111562273A (en) * | 2020-06-05 | 2020-08-21 | 大连工业大学 | Hyperspectrum-based fish water jet descaling slight damage visualization method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102721651A (en) * | 2012-06-18 | 2012-10-10 | 浙江大学 | Detection method and system of water content in plant leaf based on multispectral image |
| CN103063585A (en) * | 2013-01-05 | 2013-04-24 | 石河子大学 | Rapid nondestructive lemon and fruit maturity testing device and testing system establishment method |
| CN103235095A (en) * | 2013-04-18 | 2013-08-07 | 北京工商大学 | Water-injected meat detection method and device |
| CN103257118A (en) * | 2013-04-22 | 2013-08-21 | 华南理工大学 | Fish tenderness hyperspectral detection method based on characteristic wave band |
| CN103674864A (en) * | 2013-11-12 | 2014-03-26 | 浙江大学 | Fish water content distribution detection method based on hyperspectral imaging technology |
| CN104880427A (en) * | 2015-05-29 | 2015-09-02 | 华南理工大学 | Rapid pork product moisture content detection method |
| CN104949927A (en) * | 2015-06-01 | 2015-09-30 | 浙江大学 | Hyperspectral detection method of dried scallop moisture content |
-
2018
- 2018-01-24 CN CN201810068193.2A patent/CN108318433B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102721651A (en) * | 2012-06-18 | 2012-10-10 | 浙江大学 | Detection method and system of water content in plant leaf based on multispectral image |
| CN103063585A (en) * | 2013-01-05 | 2013-04-24 | 石河子大学 | Rapid nondestructive lemon and fruit maturity testing device and testing system establishment method |
| CN103235095A (en) * | 2013-04-18 | 2013-08-07 | 北京工商大学 | Water-injected meat detection method and device |
| CN103257118A (en) * | 2013-04-22 | 2013-08-21 | 华南理工大学 | Fish tenderness hyperspectral detection method based on characteristic wave band |
| CN103674864A (en) * | 2013-11-12 | 2014-03-26 | 浙江大学 | Fish water content distribution detection method based on hyperspectral imaging technology |
| CN104880427A (en) * | 2015-05-29 | 2015-09-02 | 华南理工大学 | Rapid pork product moisture content detection method |
| CN104949927A (en) * | 2015-06-01 | 2015-09-30 | 浙江大学 | Hyperspectral detection method of dried scallop moisture content |
Non-Patent Citations (7)
| Title |
|---|
| DI WU: "Application of visible and near infrared hyperspectral imaging for non-invasively measuring distribution of water-holding capacity in salmon flesh", 《TALANTA》 * |
| 劳东青 等: ""计算机视觉在作物水分亏缺诊断中的应用进展"", 《塔里木大学学报》 * |
| 孙俊 等: ""基于高光谱图像的生菜叶片水分预测研究"", 《光谱学与光谱分析》 * |
| 孙俊: "《光谱技术在农作物 农产品信息无损检测中的应用》", 30 June 2017 * |
| 张良培 等: "《高光谱遥感》", 31 October 2005 * |
| 潘玉成: "人工神经网络在茶叶品质评定中的应用", 《广东茶业》 * |
| 詹白勺 等: ""基于高光谱成像技术的三文鱼肉水分含量的可视化研究"", 《光谱学与光谱分析》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111562273A (en) * | 2020-06-05 | 2020-08-21 | 大连工业大学 | Hyperspectrum-based fish water jet descaling slight damage visualization method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108318433B (en) | 2021-02-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zheng et al. | Detection of adulteration with duck meat in minced lamb meat by using visible near-infrared hyperspectral imaging | |
| CN108956604B (en) | Method for identifying eriocheir sinensis quality based on hyperspectral image technology | |
| CN106596416A (en) | Chilled fresh meat quality non-destructive testing method based on hyperspectral imaging technology | |
| Xiong et al. | Potential of hyperspectral imaging for rapid prediction of hydroxyproline content in chicken meat | |
| CN102147402A (en) | Machine vision technology based method for rapidly detecting egg freshness | |
| CN101561402B (en) | Machine vision-based real-time detection and grading method and grading device for pork appearance quality | |
| CN103076288B (en) | A kind of flesh of fish self-grading device based on computer vision and method | |
| CN103257118A (en) | Fish tenderness hyperspectral detection method based on characteristic wave band | |
| Liu et al. | Prediction of chemical composition and geographical origin traceability of Chinese export tilapia fillets products by near infrared reflectance spectroscopy | |
| CN106340000A (en) | Bone age assessment method | |
| Hahn et al. | Carrot volume evaluation using imaging algorithms | |
| CN105181912A (en) | Method for detection of freshness during rice storage | |
| Wishkerman et al. | Image analysis-based classification of pigmentation patterns in fish: A case study of pseudo-albinism in Senegalese sole | |
| CN108519339A (en) | A kind of blade cadmium content Vis-NIR spectral signature modeling methods based on WT-LSSVR | |
| CN114677671A (en) | Automatic identifying method for old ribs of preserved szechuan pickle based on multispectral image and deep learning | |
| CN110108649A (en) | The fast non-destructive detection method of oil crops quality based on terahertz light spectral technology | |
| CN110243805B (en) | Fish bone detection method based on Raman hyperspectral imaging technology | |
| Wang et al. | Nondestructive determination of IMP content in chilled chicken based on hyperspectral data combined with chemometrics | |
| CN108318433A (en) | The method for detecting water injection rate in salmon using hyper-spectral image technique | |
| CN102042984A (en) | Method and system for evaluating grade of medicinal material | |
| Zhu et al. | Determination and visualization of fat contents in salmon fillets based on visible and near-infrared hyperspectral imagery | |
| CN111562273A (en) | Hyperspectrum-based fish water jet descaling slight damage visualization method | |
| CN116297236A (en) | Method and device for identifying vitality of single corn seeds based on hyperspectrum | |
| Pothula et al. | Milled industrial beet color kinetics and total soluble solid contents by image analysis | |
| CN107860746A (en) | A kind of Yunnan local chicken muscle freeze-dried powder lysine content near infrared detection method |
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 |