CN110516668A - A kind of honey adulteration detection method and device based on high light spectrum image-forming technology - Google Patents
A kind of honey adulteration detection method and device based on high light spectrum image-forming technology Download PDFInfo
- Publication number
- CN110516668A CN110516668A CN201910758654.3A CN201910758654A CN110516668A CN 110516668 A CN110516668 A CN 110516668A CN 201910758654 A CN201910758654 A CN 201910758654A CN 110516668 A CN110516668 A CN 110516668A
- Authority
- CN
- China
- Prior art keywords
- honey
- spectrum image
- data
- residual error
- spectral
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/68—Food, e.g. fruit or vegetables
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Pathology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Immunology (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Chemical & Material Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a kind of honey adulteration detection method and device based on high light spectrum image-forming technology, wherein honey adulteration detection method, the high spectrum image of the honey sample to be detected including acquiring heterogeneity ratio, and carry out black and white correction；The area-of-interest of high spectrum image after extracting black and white correction, calculates the averaged spectrum of area-of-interest；Averaged spectrum is filtered, the high-spectral data after obtaining removal noise；Three dimensions of obtained high-spectral data are normalized；Data enhancing is carried out to the high-spectral data after normalized, and is divided into training set and test set；Residual error network model is established, using training set debugging model, determines that final residual error network model carries out the detection of honey adulteration situation using detection of adulterations model using the classifying quality of test set test model as detection of adulterations model.The present invention solves that prior art detection method detection process is complicated, time-consuming, and belongs to the problem of damaging detection.
Description
Technical field
The application belongs to technical field of food detection, and in particular to a kind of honey adulteration inspection based on high light spectrum image-forming technology
Survey method and apparatus.
Background technique
Honey is that the nectar that honeybee is adopted from the flowering plant's flowers passes through crude sweet made of sufficiently brewing in honeycomb
Substance.The monosaccharide contained in honey can easily be absorbed by the body, and it is edible to be especially suitable for women, children and the middle-aged and the old.
Honey has effects that enrich blood, qi-restoratives, removing toxic substances, beauty face-whitening-nourishing, prolong life.The nutritive value of honey is high, contains fructose, Portugal
The various saccharides such as grape sugar can provide more energy to human body, in addition to this, multivitamin is also rich in, to the skin of people
Beneficial, there are also anti-aging functions.
Honey industry development in China's is rapid in recent years, and yield occupies first of the world, and the price of some unifloal honeys has and is in
The trend ramped.But honey adulteration and fraud event also constantly occur therewith, this not only influences domestic honey market
It develops in a healthy way, has an effect on the public praise in outlet honey market.
The technology of traditional detection honey adulteration mainly has: Carbon isotope detection technique, enzyme assay method, rotation
Light method, differential scanning calorimetry, gas chromatography, mass spectrography, high performance liquid chromatography etc..These methods are complicated for operation, consumption
Honey, time-consuming, and is not suitable for lossless online quick detection, is unfavorable for promoting.
Therefore, simple, quick, the lossless honey adulteration detection technique of one kind is researched and developed to have great importance.
Summary of the invention
The application's is designed to provide a kind of honey adulteration detection method and device based on high light spectrum image-forming technology, with
It solves the problem of that prior art detection method detection process is complicated, time-consuming and belongs to damage detection.
To achieve the above object, the technical solution that the application is taken are as follows:
A kind of honey adulteration detection method based on high light spectrum image-forming technology, it is described to be based on for the non-destructive testing of honey
The honey adulteration detection method of high light spectrum image-forming technology, comprising:
Step S1, the high spectrum image of the honey sample to be detected of heterogeneity ratio is acquired, and to bloom collected
Spectrogram picture carries out black and white correction；
Step S2, the area-of-interest of the high spectrum image after extracting black and white correction, calculates all pictures in area-of-interest
Averaged spectrum of the average value of the spectral reflectivity of element as honey sample to be detected；
Step S3, the averaged spectrum is filtered using Savitzky-Golay filtering method, is removed
The high-spectral data with smooth spectral information after noise；
Step S4, three dimensions of high-spectral data obtained in step S3 are normalized；
Step S5, data enhancing is carried out to the high-spectral data after normalized, and is divided into training set and test set,
Sample in training set and test set is marked；
Step S6, residual error network model is established, using the training set debugging model, determines final residual error network model
The detection of adulterations mould is utilized using the classifying quality of test set test detection of adulterations model as detection of adulterations model
The detection of type progress honey adulteration situation；
The residual error network model is successively passed through the convolutional layer of a 3*3, first by the input terminal of data to output end
Block, second block, third block, the 4th block, the first full articulamentum and the second full articulamentum；Described
The convolutional layer for the 3*3 that one block includes sequentially connected 8 residual error modules and a step-length is 2, described second
The convolutional layer for the 3*3 that block includes sequentially connected 8 residual error modules and a step-length is 2, the third block packet
The convolutional layer for the 3*3 that sequentially connected 16 residual error modules and a step-length are 2 is included, the 4th block includes successively
8 residual error modules of connection.
Preferably, the honey sample to be detected of the heterogeneity ratio, comprising:
What mixed liquor, honey and the rice syrup that pure honey, honey and rice syrup is mixed with 4:1 were mixed with 3:2
It is mixed liquor that mixed liquor, honey and the rice syrup that mixed liquor, honey and rice syrup are mixed with 2:3 are mixed with 1:4, pure
Rice syrup.
Preferably, the area-of-interest is the center of the high spectrum image after being corrected using black and white as the center of circle, with 50 pictures
Element is the border circular areas of radius.
Preferably, the high-spectral data to after normalized carries out data enhancing, comprising:
According to amount of movement random movement in the spatial domain of high-spectral data, new high spectrum image, the movement are obtained
The maximum value of amount is 4% of pixel quantity in each Spatial Dimension；
The new high spectrum image that random movement generates is overturn, new high spectrum image, the overturning packet are obtained
Include flip horizontal and flip vertical.
Preferably, the residual error module include the convolutional layer of the 1*1 set gradually, 3*3 convolutional layer with
And the convolutional layer of a 1*1；
If the spectral band number of the output data of residual error module is identical as the spectral band number of input data, number will be inputted
According to the final output data being added with output data as residual error module；If the spectral band number of the output data of residual error module with
The spectral band number of input data is different, then first obtains input data and output number after the processing of the convolutional layer of a 1*1
It is added as residual error module most according to the intermediate data with same spectra wave band number, then by the intermediate data with output data
Whole output data.
The application also provides a kind of honey adulteration detection device based on high light spectrum image-forming technology, the lossless inspection for honey
It surveys, the honey adulteration detection device based on high light spectrum image-forming technology, including memory and processor, the memory storage
There is computer program, the processor realizes the honey based on high light spectrum image-forming technology when executing the computer program
The step of adulteration detection method.
Honey adulteration detection method and device provided by the present application based on high light spectrum image-forming technology, using Savitzky-
Golay method pre-processes spectrum, establishes residual error network, and use high light spectrum image-forming technology, and high spectrum image includes one
A two-dimensional Spatial Dimension and an one-dimensional spectral Dimensions, can provide spatial information and spectral information, to EO-1 hyperion simultaneously
Three dimensions (space-optical spectrum dimension) of image carry out feature extraction and classify, to obtain more comprehensive feature, so that inspection
Survey result it is more accurate, have many advantages, such as without damage, it is pollution-free, automate, rapidly and efficiently.
Detailed description of the invention
Fig. 1 is the flow diagram of the honey adulteration detection method based on high light spectrum image-forming technology of the application；
Fig. 2 is a kind of example structure schematic diagram of the residual error network model of the application；
Fig. 3 is the structural schematic diagram of first block of the residual error network model of the application；
Fig. 4 is the structural schematic diagram of second block of the residual error network model of the application；
Fig. 5 is the structural schematic diagram of the third block of the residual error network model of the application；
Fig. 6 is the structural schematic diagram of the 4th block of the residual error network model of the application；
Fig. 7 is a kind of example structure schematic diagram of the residual error module of the application；
Fig. 8 is another example structure schematic diagram of the residual error module of the application.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application carries out clear, complete
Site preparation description, it is clear that the described embodiments are only a part but not all of the embodiments of the present application.Based on this
Embodiment in application, every other reality obtained by those of ordinary skill in the art without making creative efforts
Example is applied, shall fall in the protection scope of this application.
Unless otherwise defined, all technical and scientific terms used herein and the technical field for belonging to the application
The normally understood meaning of technical staff is identical.The term used in the description of the present application is intended merely to description tool herein
The purpose of the embodiment of body is not to be to limit the application.
It should be understood that there is no stringent sequences to limit for the execution of each step unless expressly stating otherwise in the application
System, these steps can execute in other order.Moreover, at least part step may include multiple sub-steps or multiple
Stage, these sub-steps or stage are not necessarily to execute completion in synchronization, but can execute at different times,
These sub-steps perhaps the stage execution sequence be also not necessarily successively carry out but can be with other steps or other steps
Sub-step or at least part in stage execute in turn or alternately.
A kind of honey adulteration detection method based on high light spectrum image-forming technology is provided in one of the embodiments, is used for
The non-destructive testing of honey.The technical program be based on high light spectrum image-forming technology, high light spectrum image-forming technology electromagnetic spectrum it is ultraviolet, can
Light-exposed, near-infrared and mid infrared region are imaged target area with tens of to hundreds of continuous and subdivision spectral band simultaneously.
While obtaining earth's surface image information, its spectral information is also obtained, the image of acquisition has the advantages that image in conjunction with spectrum.
As shown in Figure 1, the honey adulteration detection method based on high light spectrum image-forming technology, comprising:
Step S1, the high spectrum image of the honey sample to be detected of heterogeneity ratio is acquired, and to bloom collected
Spectrogram picture carries out black and white correction.
The honey sample to be detected of heterogeneity ratio employed in the present embodiment, comprising: pure honey, honey and
Mixed liquor, honey and the rice syrup that mixed liquor, honey and the rice syrup that rice syrup is mixed with 4:1 are mixed with 3:2 are with 2:3
Mixed liquor that mixed mixed liquor, honey and rice syrup is mixed with 1:4, pure rice syrup this 6 kinds of classifications, and it is each
Classification is equipped with different identifiers.
Honey therein with rice syrup is that supermarket or other approach are commercially available, and assume the honey being commercially available with
Rice syrup is pure honey and rice syrup, and the present embodiment mixes rice syrup in honey and tests institute taking human as production
The adulterated honey sample needed, every group of honey sample take 80 samples.
It should be noted that enumerated above is only honey sample used in the present embodiment, in other embodiments,
Honey sample can also be the pure honey directly bought, adulterated honey, or with the high-fructose corn syrup of arbitrary proportion, sweet tea
Dish syrup, fruit grape slurry etc. mix honey, adulterated honey needed for the experiment artificially made.
When carrying out high spectrum image acquisition operation, the Hyperspectral imager that the present embodiment uses closes spectrum for the double benefits in Sichuan
Image- λ-V10E-PS the Hyperspectral imager that Science and Technology Ltd. provides, chief component is imaging spectrometer
(Imperx IPX-2M30, Zolix), CCD camera, the halogen lamp of four 150w, electronic control translation stage and a calculating unit
At.Wherein the spectral region of imaging spectrometer acquisition is 383.70-1032.70nm, resolution ratio 2.73nm.High-spectral data is adopted
Integrate software as SpecView.
Entire collection process carries out in camera bellows, generates shadow to collected high spectrum image to prevent the light in environment
It rings.Parameter before acquiring high spectrum image are as follows: time for exposure 9ms, platform traveling time 1.57cm/s, the halogen lamp of four 150w
Angle with platform is 50 degree.
When carrying out high spectrum image acquisition, four halogen lamp in Hyperspectral imager are opened 30 minutes and are preheated, with
This stabilization to ensure light source.First by blank be placed on honey sample same distance and lighting position, and be full of camera one
The acquisition range of frame carries out whiteboard data acquisition, then successively carries out high spectrum image acquisition to 480 honey samples, finally
Light source is closed, and covers lens cap and carries out the acquisition of dark background data.After these data have been acquired, to collected honey
High spectrum image carries out black and white correction process.
In one embodiment, the black and white updating formula that black and white correction process uses are as follows:
Wherein: R is the high spectrum image after correction, and I is the original high spectrum image of honey sample, and W is blank diffusing reflection
Image, B are dark background image.
Step S2, the area-of-interest of the high spectrum image after extracting black and white correction, calculates all pictures in area-of-interest
Averaged spectrum of the average value of the spectral reflectivity of element as honey sample to be detected.
When the present embodiment extracts region of interest, the unified center for extracting the high spectrum image after correcting using black and white as the center of circle,
It is the border circular areas of radius as area-of-interest (ROI) using 50 pixels.Under normal conditions, honey can go out after placing a period of time
Existing crystalline polamer, in order to enable the detection method of the present embodiment to be suitable for detect the honey of crystallization and nodeless mesh, therefore this implementation
Example is using the center of high spectrum image as the center of circle of area-of-interest, in order to get lotion and the non-lotion bee in the case of crystallization
Honey, to extract perfect sample data, to improve the accuracy of subsequent detection of adulterations.
The present embodiment using 50 pixels as radius is set according to most of container for placing honey in the market, at present city
For placing, the container shapes of honey are different or even partial containers are connecing paracentral part in convergence state, in order to protect on field
Demonstrate,proving area-of-interest can be complete border circular areas in the case where directly carrying out detection of adulterations using honey original container, therefore this
It is 50 pixels that radius is arranged in embodiment, to realize the non-destructive testing of honey.
It is easily understood that the range of area-of-interest is set according to actual high spectrum image, such as place bee
The container of honey is too small, can suitably reduce the range of area-of-interest；If the container for placing honey is excessive, can suitably increase interested
The range in region.Repeat the area-of-interest of identical operation to the high spectrum image for obtaining whole test samples.
The average value of the spectral reflectivity of all pixels point in the rounded interested area is calculated as the flat of each sample
Equal spectrum, to obtain the averaged spectrum of each sample.
Step S3, the averaged spectrum is filtered using Savitzky-Golay filtering method, is removed
The high-spectral data with smooth spectral information after noise.
Since the quality of honey causes collected high spectrum image that can generate some small mirror-reflections, this will be led
Same sample collected spectrum within the same time is caused small noise occur, in order to avoid noise is to the shadow of subsequent operation
It rings, noise is eliminated by Savitzky-Golay filtering method in the present embodiment.
In one embodiment, a kind of calculation formula of Savitzky-Golay filtering method is provided are as follows:
Wherein, hiIt is smoothing factor.Each measured value multiplied by the purpose of smoothing factor is reduced as far as possible smoothly to useful
The influence of information improves the disadvantage for smoothly removing hot-tempered algorithm,It can be based on the principle of least square, acquired with fitting of a polynomial.
The smooth matrix operator of Savitzky-Golay convolution solves as follows:
If the width of filter window is n=2m+1, each measurement point is x=(- m ,-m+1 ..., 0 ..., m-1, m), using k-
1 order polynomial is fitted the data point in window, multinomial are as follows:
Y=a0+a1x+a2x2+…+akxk-1,k<n
Wherein, y is value of the measurement point x after over-fitting, a0,a1,…,ak-1For fitting parameter.Then it obtains as n
Equation constitutes k member system of linear equations, passes through least square method you and determining fitting parameter A=(a0, a1..., ak-1)T.Thus may be used
Obtain k member system of linear equations:
It is expressed in matrix as:
Y(2m+1)×1=X(2m+1)×k·Ak×1+E(2m+1)×1
Wherein E is the residual error of least square fitting.
The least square solution of AAre as follows:
The filter value of YAre as follows:
Wherein B=X (XT·X)-1·XT.The present embodiment is smooth to each sample progress Savitzky-Golay convolution,
So that obtained spectrum is more smooth, influence of the noise to experiment is reduced, and the shape of spectral preservation and width are constant, to rear
Continuous modeling and detection has great importance.
It should be noted that Savitzky-Golay filtering method provided in this embodiment is existing method, part is thin
Section is no longer repeated.Since Savitzky-Golay filtering method can reduce the smoothly influence to useful information, therefore use
Savitzky-Golay filtering method eliminates noise, can not only eliminate the high spectrum image due to caused by honey quality and has and make an uproar
The problem of sound, can also retain the original granular sensation of honey and micro-bubble in high spectrum image, to improve the accurate of adulterated judgement
Property, there is more preferably filter effect compared to other filtering modes such as Gabor filtering, bilateral filterings.
Step S4, three dimensions of high-spectral data obtained in step S3 are normalized.
Normalized can convert the data in the same order of magnitude for the data of three dimensions of high-spectral data, with
Convenient for the processing to data.
In one embodiment, the normalization formula provided are as follows:
Wherein, xI, maxAnd xI, minSpace or spectrum tie up maximum value and minimum value in x, x respectively in high-spectral dataiFor
Pixel or spectroscopic data before normalization, xI, normFor the pixel or spectroscopic data after normalization.
Step S5, data enhancing is carried out to the high-spectral data after normalized, and is divided into training set and test set,
And the sample in training set and test set is marked.
Since sample data volume is bigger, subsequent obtained model is more accurate, therefore after the present embodiment is to normalized
Data carry out data enhancing processing.
In one embodiment, data enhancement operations include: and are moved at random in the spatial domain of high-spectral data according to amount of movement
It is dynamic, new high spectrum image is obtained, the maximum value of the amount of movement is 4% of pixel quantity in each Spatial Dimension；It will be random
The mobile new high spectrum image generated is overturn, and new high spectrum image is obtained, and the overturning includes flip horizontal and hangs down
Straight overturning.
Movement is carried out to high spectrum image, overturns processing realization data enhancing, it is different that multiple are obtained by an original image
The image of sample, so that data sample amount is increased to 6 times of original sample number.
Wherein, itself for mobile and turning operation, movement in the prior art can be used or turning operation is realized,
Such as number of patent application is the random translation averaged spectrum brightness of image mentioned in 201710595555.9 document, for another example specially
Benefit is no longer repeated herein application No. is the Image Reversal mentioned in 201810330536.8 document.
After data enhancing, obtained all sample datas are split to obtain training set and test set, data point
Cut ratio depending on actual needs, the ratio in the present embodiment with 4:1 is divided into training set and test set, and to each sample
It is marked, the input data as residual error network model.Classification mark mainly is carried out to sample when being marked in the present embodiment
Note.
Step S6, residual error network model is established, using the training set debugging model, determines final residual error network model
The detection of adulterations mould is utilized using the classifying quality of test set test detection of adulterations model as detection of adulterations model
The detection of type progress honey adulteration situation.
In one embodiment, used residual error network model, comprising:
As shown in Fig. 2, residual error network model successively passed through by the input terminal of data to output end a 3*3 convolutional layer,
First block, second block, third block, the 4th block, the first full articulamentum and the second full articulamentum.
Wherein, as illustrated in figures 3-6, the volume for the 3*3 that first block includes 8 residual error modules and a step-length is 2
Lamination, the convolutional layer for the 3*3 that second block includes 8 residual error modules and a step-length is 2, third block includes 16
The convolutional layer for the 3*3 that a residual error module and a step-length are 2, the 4th block include 8 residual error modules.
The residual error network model of the present embodiment can preferably realize the function of Classification and Identification, certainly in the feelings of meet demand
Under condition, the residual error network model that the application uses can be existing residual error network model, such as ResNet residual error network.
Wherein, residual error module is a fixed module, and residual error module employed in the present embodiment includes setting gradually
The convolutional layer of a 1*1, the convolutional layer of 3*3 and a 1*1 convolutional layer.
Also, the data output of residual error module meets the following conditions: as shown in fig. 7, if the output data of residual error module
Spectral band number is identical as the spectral band number of input data, then is added input data as residual error module with output data
Final output data；As shown in figure 8, if the spectral band number of the output data of residual error module and the spectral band number of input data
Input data is then first obtained having same spectra wave band number with output data by difference after the processing of the convolutional layer of a 1*1
Intermediate data, then the intermediate data is added with output data to the final output data as residual error module.
It should be noted that residual error module involved in four block is with the spectrum wave of output data in Fig. 3~6
Number of segment form identical with the spectral band number of input data is indicated, and the representation is not as to the final of residual error module
The limitation of output data, i.e., the final output data of residual error module involved in four block are all satisfied above-mentioned residual in Fig. 3~6
The data output condition of difference module.
In two full articulamentums, the first full articulamentum is followed by a swish activation primitive, the activation of the second full articulamentum
Function is softmax.Loss function used in network is cross entropy loss function.Final network output category accuracy rate, is called together
Rate and f-score are returned, f-score is the comprehensive value model of accuracy rate and recall rate, is the weighting of accuracy rate and recall rate
Harmonic average.
After the continuous debugging model of training set, finally obtained detection of adulterations model is to the accurate of training set Classification and Identification
Rate can reach 100%, reach 97.024% to the accuracy rate of test set Classification and Identification, recall rate 0.9702, and f-score is
0.9700.With preferably recognition effect, mixed using high-spectral data of the detection of adulterations model to honey to be detected
Obtained detection of adulterations model, is used for the popularization and application of honey adulteration detection by vacation detection.If the classification tested using test set
Effect is undesirable, and it is more excellent with new sample debugging acquisition can to re-use training set debugging model, or reacquisition training set
Detection of adulterations model.
In another embodiment, a kind of honey adulteration detection device based on high light spectrum image-forming technology is also provided, is based on
The honey adulteration detection device of high light spectrum image-forming technology includes:
Image capture module, the high spectrum image of the honey sample to be detected for acquiring heterogeneity ratio, and to institute
The high spectrum image of acquisition carries out black and white correction；
First computing module calculates region of interest for extracting the area-of-interest of the high spectrum image after black and white corrects
Averaged spectrum of the average value of the spectral reflectivity of all pixels as honey sample to be detected in domain；
Second computing module, for being filtered place to the averaged spectrum using Savitzky-Golay filtering method
Reason, the high-spectral data with smooth spectral information after obtaining removal noise；
Third computing module is normalized for three dimensions to obtained high-spectral data；
4th computing module for carrying out data enhancing to the high-spectral data after normalized, and is divided into training
Collection and test set, are marked the sample in training set and test set；
Model building module, using the training set debugging model, determines finally residual for establishing residual error network model
Poor network model is as detection of adulterations model, using the classifying quality of test set test detection of adulterations model, using described
The detection of detection of adulterations model progress honey adulteration situation.
Wherein residual error network model is successively passed through the convolutional layer of a 3*3, first by the input terminal of data to output end
Block, second block, third block, the 4th block, the first full articulamentum and the second full articulamentum；Described
The convolutional layer for the 3*3 that one block includes 8 residual error modules and a step-length is 2, second block include 8 residual
The convolutional layer for the 3*3 that difference module and a step-length are 2, the third block include 16 residual error modules and a step
The convolutional layer of a length of 2 3*3, the 4th block include 8 residual error modules.
About the honey adulteration detection device based on high light spectrum image-forming technology it is specific restriction may refer to above for
The restriction of honey adulteration detection method based on high light spectrum image-forming technology, details are not described herein.
In another embodiment, a kind of computer equipment, i.e., a kind of honey based on high light spectrum image-forming technology are also provided
Detection of adulterations device, computer equipment can be terminal, internal structure may include the processor connected by system bus,
Memory, network interface, display screen and input unit.Wherein, the processor of computer equipment calculates and controls energy for providing
Power.The memory of computer equipment includes non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with behaviour
Make system and computer program.The built-in storage is the operation of the operating system and computer program in non-volatile memory medium
Environment is provided.The network interface of computer equipment is used to communicate with external terminal by network connection.The computer program quilt
To realize the above-mentioned honey adulteration detection method based on high light spectrum image-forming technology when processor executes.The display screen of computer equipment
It can be liquid crystal display or electric ink display screen, the input unit of each equipment can be the touch covered on display screen
Layer, is also possible to the key being arranged on computer equipment shell, trace ball or Trackpad, can also be external keyboard, touch-control
Plate or mouse etc..
The honey adulteration detection device based on high light spectrum image-forming technology of the present embodiment, for the non-destructive testing of honey, institute
The honey adulteration detection device based on high light spectrum image-forming technology, including memory and processor are stated, the memory is stored with meter
Calculation machine program, the processor realize the honey adulteration based on high light spectrum image-forming technology when executing the computer program
The step of detection method.
About the honey adulteration detection device based on high light spectrum image-forming technology it is specific restriction may refer to above for
The restriction of honey adulteration detection method based on high light spectrum image-forming technology, details are not described herein.
Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality
It applies all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not present
Contradiction all should be considered as described in this specification.
The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the application
Range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.
Claims (6)
1. a kind of honey adulteration detection method based on high light spectrum image-forming technology, the non-destructive testing for honey, which is characterized in that
The honey adulteration detection method based on high light spectrum image-forming technology, comprising:
Step S1, the high spectrum image of the honey sample to be detected of heterogeneity ratio is acquired, and to high-spectrum collected
As carrying out black and white correction；
Step S2, the area-of-interest of the high spectrum image after extracting black and white correction calculates all pixels in area-of-interest
Averaged spectrum of the average value of spectral reflectivity as honey sample to be detected；
Step S3, the averaged spectrum is filtered using Savitzky-Golay filtering method, obtains removal noise
The high-spectral data with smooth spectral information afterwards；
Step S4, three dimensions of high-spectral data obtained in step S3 are normalized；
Step S5, data enhancing is carried out to the high-spectral data after normalized, and is divided into training set and test set, to instruction
The sample practiced in collection and test set is marked；
Step S6, residual error network model is established, using the training set debugging model, determines final residual error network model conduct
Detection of adulterations model, using the test set test detection of adulterations model classifying quality, using the detection of adulterations model into
The detection of row honey adulteration situation；
The residual error network model is successively passed through the convolutional layer of a 3*3, first by the input terminal of data to output end
Block, second block, third block, the 4th block, the first full articulamentum and the second full articulamentum；Described
The convolutional layer for the 3*3 that one block includes sequentially connected 8 residual error modules and a step-length is 2, described second
The convolutional layer for the 3*3 that block includes sequentially connected 8 residual error modules and a step-length is 2, the third block packet
The convolutional layer for the 3*3 that sequentially connected 16 residual error modules and a step-length are 2 is included, the 4th block includes successively
8 residual error modules of connection.
2. the honey adulteration detection method based on high light spectrum image-forming technology as described in claim 1, which is characterized in that it is described not
The honey sample to be detected of congruent ratio, comprising:
The mixing that mixed liquor, honey and the rice syrup that pure honey, honey and rice syrup is mixed with 4:1 are mixed with 3:2
Mixed liquor that mixed liquor, honey and the rice syrup that liquid, honey and rice syrup are mixed with 2:3 are mixed with 1:4, pure rice
Syrup.
3. the honey adulteration detection method based on high light spectrum image-forming technology as described in claim 1, which is characterized in that the sense
Interest region is the center of the high spectrum image after being corrected using black and white as the center of circle, using 50 pixels as the border circular areas of radius.
4. the honey adulteration detection method based on high light spectrum image-forming technology as described in claim 1, which is characterized in that described right
High-spectral data after normalized carries out data enhancing, comprising:
According to amount of movement random movement in the spatial domain of high-spectral data, new high spectrum image is obtained, the amount of movement
Maximum value is 4% of pixel quantity in each Spatial Dimension；
The new high spectrum image that random movement generates is overturn, new high spectrum image is obtained, the overturning includes water
Flat overturning and flip vertical.
5. the honey adulteration detection method based on high light spectrum image-forming technology as described in claim 1, which is characterized in that described residual
Difference module includes the convolutional layer of the convolutional layer of the 1*1 set gradually, the convolutional layer of 3*3 and a 1*1；
If the spectral band number of the output data of residual error module is identical as the spectral band number of input data, by input data and
Output data is added the final output data as residual error module；If the spectral band number of the output data of residual error module and input
The spectral band number of data is different, then first obtains having with output data after the processing of the convolutional layer of a 1*1 by input data
There is the intermediate data of same spectra wave band number, then the intermediate data is added with output data as the final defeated of residual error module
Data out.
6. a kind of honey adulteration detection device based on high light spectrum image-forming technology, described based on height for the non-destructive testing of honey
The honey adulteration detection device of spectral imaging technology, including memory and processor, the memory are stored with computer program,
It is characterized in that, being realized when the processor execution computer program according to any one of claims 1 to 5 based on height
The step of honey adulteration detection method of spectral imaging technology.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910758654.3A CN110516668A (en) | 2019-08-16 | 2019-08-16 | A kind of honey adulteration detection method and device based on high light spectrum image-forming technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910758654.3A CN110516668A (en) | 2019-08-16 | 2019-08-16 | A kind of honey adulteration detection method and device based on high light spectrum image-forming technology |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110516668A true CN110516668A (en) | 2019-11-29 |
Family
ID=68626324
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910758654.3A Pending CN110516668A (en) | 2019-08-16 | 2019-08-16 | A kind of honey adulteration detection method and device based on high light spectrum image-forming technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110516668A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111008970A (en) * | 2019-12-11 | 2020-04-14 | 江南大学 | Hyperspectral image acquisition system-based method for detecting impurities in preserved vegetables |
| CN111175239A (en) * | 2020-01-19 | 2020-05-19 | 北京科技大学 | High-spectrum nondestructive testing and identifying system for imaging of colored drawing cultural relics under deep learning |
| CN112016392A (en) * | 2020-07-17 | 2020-12-01 | 浙江理工大学 | Hyperspectral image-based small sample detection method for soybean pest damage degree |
| CN113959961A (en) * | 2021-12-22 | 2022-01-21 | 广东省农业科学院动物科学研究所 | Hyperspectral image-based tannin additive anti-counterfeiting detection method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108398391A (en) * | 2018-01-23 | 2018-08-14 | 浙江理工大学 | A kind of detection method that the olive oil based on high light spectrum image-forming technology is adulterated |
| CN109540831A (en) * | 2019-01-25 | 2019-03-29 | 中国中医科学院中药研究所 | Fructus lycii variety ecotype method based on high light spectrum image-forming technology |
| CN110046575A (en) * | 2019-04-16 | 2019-07-23 | 浙江农林大学 | Based on the remote sensing images scene classification method for improving residual error network |
-
2019
- 2019-08-16 CN CN201910758654.3A patent/CN110516668A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108398391A (en) * | 2018-01-23 | 2018-08-14 | 浙江理工大学 | A kind of detection method that the olive oil based on high light spectrum image-forming technology is adulterated |
| CN109540831A (en) * | 2019-01-25 | 2019-03-29 | 中国中医科学院中药研究所 | Fructus lycii variety ecotype method based on high light spectrum image-forming technology |
| CN110046575A (en) * | 2019-04-16 | 2019-07-23 | 浙江农林大学 | Based on the remote sensing images scene classification method for improving residual error network |
Non-Patent Citations (4)
| Title |
|---|
| KAIMING HE，ET AL.: "Deep residual learning for image recognition", 《2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| SAHAMEH SHAFIEE，ET AL.: "Detection of Honey Adulteration using Hyperspectral Imaging", 《IFAC-PAPERSONLINE》 * |
| 刘冰，等.: "联合空－谱信息的高光谱影像深度三维卷积网络分类", 《测绘学报》 * |
| 李鑫，周巍，段哲民著: "《微处理器系统级片上温度感知技术》", 31 March 2019, 西北工业大学出版社 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111008970A (en) * | 2019-12-11 | 2020-04-14 | 江南大学 | Hyperspectral image acquisition system-based method for detecting impurities in preserved vegetables |
| CN111175239A (en) * | 2020-01-19 | 2020-05-19 | 北京科技大学 | High-spectrum nondestructive testing and identifying system for imaging of colored drawing cultural relics under deep learning |
| CN111175239B (en) * | 2020-01-19 | 2021-01-15 | 北京科技大学 | High-spectrum nondestructive testing and identifying system for imaging of colored drawing cultural relics under deep learning |
| CN112016392A (en) * | 2020-07-17 | 2020-12-01 | 浙江理工大学 | Hyperspectral image-based small sample detection method for soybean pest damage degree |
| CN112016392B (en) * | 2020-07-17 | 2024-05-28 | 浙江理工大学 | Small sample detection method for soybean insect pest degree based on hyperspectral image |
| CN113959961A (en) * | 2021-12-22 | 2022-01-21 | 广东省农业科学院动物科学研究所 | Hyperspectral image-based tannin additive anti-counterfeiting detection method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110516668A (en) | A kind of honey adulteration detection method and device based on high light spectrum image-forming technology | |
| Li et al. | Apple quality identification and classification by image processing based on convolutional neural networks | |
| Vahadane et al. | Structure-preserving color normalization and sparse stain separation for histological images | |
| Yang et al. | No-reference hyperspectral image quality assessment via quality-sensitive features learning | |
| Li et al. | Wheat ear recognition based on RetinaNet and transfer learning | |
| He et al. | Determination of total polysaccharides and total flavonoids in Chrysanthemum morifolium using near-infrared hyperspectral imaging and multivariate analysis | |
| Lukinac et al. | Computer vision method in beer quality evaluation—A review | |
| Xu et al. | Multi-exposure image fusion techniques: A comprehensive review | |
| Shu et al. | Improving the estimation accuracy of SPAD values for maize leaves by removing UAV hyperspectral image backgrounds | |
| Zhai | Inversion of organic matter content in wetland soil based on Landsat 8 remote sensing image | |
| Liang et al. | Estimating crop LAI using spectral feature extraction and the hybrid inversion method | |
| CN108398391A (en) | A kind of detection method that the olive oil based on high light spectrum image-forming technology is adulterated | |
| Li et al. | Low-light image enhancement based on quasi-symmetric correction functions by fusion | |
| Pan et al. | Low-light image enhancement method based on retinex theory by improving illumination map | |
| Przybył et al. | Neural image analysis and electron microscopy to detect and describe selected quality factors of fruit and vegetable spray-dried powders—Case study: Chokeberry powder | |
| Rahman et al. | A smart system for low-light image enhancement with color constancy and detail manipulation in complex light environments | |
| Yin et al. | Hyperspectral UAV images at different altitudes for monitoring the leaf nitrogen content in cotton crops | |
| Yu et al. | GLAGC: Adaptive dual-gamma function for image illumination perception and correction in the wavelet domain | |
| Chen et al. | Application of multispectral camera in monitoring the quality parameters of fresh tea leaves | |
| Pham et al. | Hyperspectral imaging system with rotation platform for investigation of jujube skin defects | |
| Kumar et al. | Exploring subpixel learning algorithms for estimating global land cover fractions from satellite data using high performance computing | |
| Li et al. | Recent development and challenges in spectroscopy and machine vision technologies for crop nitrogen diagnosis: A review | |
| Mu et al. | Non‐destructive detection of blueberry skin pigments and intrinsic fruit qualities based on deep learning | |
| Luo et al. | Spatial frequency domain imaging system calibration, correction and application for pear surface damage detection | |
| He et al. | Remote sensing image haze removal based on superpixel |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191129 |
|
| RJ01 | Rejection of invention patent application after publication |