CN108647675A - A kind of egg type discrimination method and device - Google Patents
A kind of egg type discrimination method and device Download PDFInfo
- Publication number
- CN108647675A CN108647675A CN201810444505.5A CN201810444505A CN108647675A CN 108647675 A CN108647675 A CN 108647675A CN 201810444505 A CN201810444505 A CN 201810444505A CN 108647675 A CN108647675 A CN 108647675A
- Authority
- CN
- China
- Prior art keywords
- egg
- spectrum
- model
- picture
- default
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/10—Image acquisition
- G06V10/12—Details of acquisition arrangements; Constructional details thereof
- G06V10/14—Optical characteristics of the device performing the acquisition or on the illumination arrangements
- G06V10/143—Sensing or illuminating at different wavelengths
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a kind of egg type discrimination method and devices, belong to deep learning and high light spectrum image-forming technology field.The method includes:The spectrum picture of egg is acquired by high light spectrum image-forming technology;The spectrum picture is pre-processed;Differentiate that model obtains egg type identification result using default egg type.The present invention can fast implement the whole process of the acquisition of egg spectrum picture, processing and identification result, the needs that ordinary populace consumer differentiates egg type in the market can be met, the technical issues of to solve pendent quick discriminating egg type in the prior art, and this method has stronger life practicability, can carry out large-scale promotion and application.
Description
Technical field
The present invention relates to deep learning and high light spectrum image-forming technology field, more particularly to a kind of egg type discrimination method
And device.
Background technology
Egg increasingly attracts attention as the most common food of family, quality.Currently on the market sell egg product
Kind is various, has Countryside Egg, convolvulate asiabell root, hen egg, Countryside Egg, foreign egg, wretch, red heart egg etc., allows people to have more visitors or business than one can attend to, at all
It can not judge egg quality.In fact, these eggs can substantially be divided into two major classes:Countryside Egg and feed egg, convolvulate asiabell root, hen
Though these have nuance for egg, Countryside Egg, Countryside Egg type is belonged to, is all the egg that natural environment descends free-ranging chicken to give birth to;Foreign chicken
Egg, wretch, red heart egg etc. belong to feed egg, in the breeding process mainly based on feeding material.
Although its nutritional ingredient is with feed egg almost without difference, Countryside Egg is more liked by consumer on the market,
Market price higher.Therefore this also leads to occur in the market the Countryside Eggs of many fake and forged commodity, for example, red heart egg be exactly by
It is added to pigment fuel in feed for nursing, artificially produces red yolk to pretend to be Countryside Egg.For ordinary consumer, lead to
The true and false or the quality for crossing identifying by naked eye these eggs are unlikely, and consumer can be helped by being badly in need of a kind of easy-to-use method
Quickly differentiate egg type.
Patent CN201710155034.1 discloses a kind of Countryside Egg based on fiber spectrum or foreign egg kind detection dress
It sets and its method, passes through the detection that tunable light source, fiber-optic probe instrument, spectrometer and computer etc. carry out egg kind, this side
Method is that comparison is traditional and be suitable for production line mass detection, it is clear that is unsuitable for that there are the quick discriminatings under ordinary consumer scene;
And of this sort detection method in the prior art generally acquires spectroscopic data, nothing by spectrometer in the form of putting individually
By be in operability or in efficiency cannot all meet ordinary consumer market egg classification is quickly differentiated it is urgent
Demand.
Invention content
In order to solve problems in the prior art, an embodiment of the present invention provides a kind of egg type discrimination method and devices.
The technical solution is as follows:
In a first aspect, a kind of egg type discrimination method, including:The spectrogram of egg is acquired by high light spectrum image-forming technology
Picture;The spectrum picture is pre-processed;Differentiate that model obtains egg type identification result using default egg type.
With reference to first aspect, in the first possible implementation, the light of egg is acquired by high light spectrum image-forming technology
Spectrogram picture, including:
By high light spectrum image-forming technology several spectrum pictures are acquired along the equator of egg and/or two extreme directions.
The possible realization method of with reference to first aspect the first, second may in realization method, by EO-1 hyperion at
As the spectrum picture of technology acquisition egg, including:
Between wavelength 400-1000nm under conditions of at least 100 wave bands, between the crosscutting circumference of egg equatorial direction
Several spectrum pictures are acquired every predetermined angular, and a width spectrum picture is acquired respectively in two extreme directions of egg.
With reference to first aspect, in the third possible realization method, the spectrum picture is pre-processed, including:
Egg target area is intercepted respectively to the spectrum picture, the immediate vicinity in the target area randomly selects pre-
The target point of fixed number amount takes the corresponding spectrum of the target point to eliminate noise.
The third possible realization method with reference to first aspect is interacted in the 4th kind of possible realization method by user
Mode, cluster or image segmentation algorithm carry out egg target area interception, are returned using smooth, multiplicative scatter correction and/or standard
One, which changes method, eliminates noise.
With reference to first aspect and the first to two kind of any possible realization method of first aspect, at the five to seven kind
In possible realization method, using before presetting egg type discriminating model acquisition egg type identification result, the method is also
Including:
Training obtains the default egg type and differentiates model.
Any possible realization method of the five to seven kind with reference to first aspect, in the eight to ten kind of possible realization method
In, training obtains the default egg type and differentiates model, including:
The spectrum picture of egg sample is acquired by high light spectrum image-forming technology;
The spectrum picture of the egg sample is pre-processed;
It is trained using predetermined depth confidence network, obtains the default egg type and differentiate model.
Any possible realization method of the eight to ten kind with reference to first aspect may be realized at the 11st to 13 kind
In mode, by high light spectrum image-forming technology acquire egg sample spectrum picture the step of, adopted by high light spectrum image-forming technology
Collect and completed in the spectrum picture step of egg, alternatively, individually completing.
Any possible realization method of the eight to ten kind with reference to first aspect may be realized at the 14th to 16 kind
In mode, pretreated step is carried out to the spectrum picture, the spectrum picture completed in pre-treatment step, or
Person individually completes.
The the 8th to 16 kind of any possible realization method with reference to first aspect, in the 17th to 25 kind of possibility
It in realization method, is trained using predetermined depth confidence network, obtains the default egg type and differentiate model, including:
The predetermined depth confidence network is made of 1 one-dimensional convolutional layer and P limited Boltzmann machines, one-dimensional convolutional layer
Include the one-dimensional convolution sum pondization composition of S 1*K, the data of convolutional layer output are limited as the input for being limited Boltzmann machine
Boltzmann machine internal structure includes a visible layer and a hidden layer, in training process, trains up the limited of last layer
The limited Boltzmann machine of training current layer after Boltzmann machine, until last layer, finally obtains default egg type and differentiate
Model M, wherein P >=3, S >=4, K >=5.
Second aspect provides a kind of egg type identification device, including:
Spectrum picture acquisition module, the spectrum picture for acquiring egg by high light spectrum image-forming technology;
Preprocessing module, for being pre-processed to the spectrum picture;
Identification result acquisition module, for differentiating that model obtains egg type identification result using default egg type.
In conjunction with second aspect, in the first possible realization method of second aspect, the spectrum picture acquisition module is used
In:
By high light spectrum image-forming technology several spectrum pictures are acquired along the equator of egg and/or two extreme directions.
In conjunction with the first possible realization method of second aspect, in second of possible realization method, the spectrogram
As acquisition module is used for:
Between wavelength 400-1000nm under conditions of at least 100 wave bands, between the crosscutting circumference of egg equatorial direction
Several spectrum pictures are acquired every predetermined angular, and a width spectrum picture is acquired respectively in two extreme directions of egg.
In conjunction with second aspect, in the third possible realization method, the preprocessing module is used for:
Egg target area is intercepted respectively to the spectrum picture, the immediate vicinity in the target area randomly selects pre-
The target point of fixed number amount takes the corresponding spectrum of the target point to eliminate noise.
In conjunction with the third possible realization method of second aspect, in the 4th kind of possible realization method, the pretreatment mould
Block carries out egg target area interception by user's interactive mode, cluster or image segmentation algorithm, utilizes smooth, polynary scattering school
Just and/or standard method for normalizing eliminates noise.
In conjunction with second aspect, in the 5th kind of possible realization method, described device further includes:
Model training module obtains the default egg type for training and differentiates model.
In conjunction with the 5th kind of possible realization method of second aspect, in the 6th kind of possible realization method, the model training
Module is used for:
It is trained using predetermined depth confidence network, obtains the default egg type and differentiate model.
In conjunction with the 6th kind of possible realization method of second aspect, in the 7th kind of possible realization method,
The predetermined depth confidence network is made of 1 one-dimensional convolutional layer and P limited Boltzmann machines, one-dimensional convolutional layer
Include the one-dimensional convolution sum pondization composition of S 1*K, the data of convolutional layer output are limited as the input for being limited Boltzmann machine
Boltzmann machine internal structure includes a visible layer and a hidden layer, in training process, trains up the limited of last layer
The limited Boltzmann machine of training current layer after Boltzmann machine, until last layer, finally obtains default egg type and differentiate
Model M, wherein P, S, K are the integer more than 0.
The advantageous effect that technical solution provided in an embodiment of the present invention is brought is:
In conclusion egg type discrimination method provided in an embodiment of the present invention and device, have compared with the existing technology
Following advantageous effect:
1, the spectrum picture of egg is easily acquired by high light spectrum image-forming technology, it is only necessary to by easily acquiring egg
High spectrum image then differentiated, export identification result, can quickly differentiate egg type;
2, judged, will not be had any impact to the quality and structure of egg have using the diffusing transmission spectrum of light source
There is the advantages of Undamaged determination;
3, the discriminating model that the embodiment of the present invention utilizes is to learn to obtain by deep neural network, to external environment robust
Property it is higher, even if under complex environment, still have good distinguishing ability;
4, egg classification discrimination method and device provided in an embodiment of the present invention, discrimination process is easy to operate, has relatively strong
Practicability, large-scale promotion and application can be carried out.
Description of the drawings
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for
For those of ordinary skill in the art, without creative efforts, other are can also be obtained according to these attached drawings
Attached drawing.
Fig. 1 is the egg type discrimination method flow chart that the embodiment of the present invention 1 provides;
Fig. 2 implements the set-up mode exemplary plot of spectrum picture acquisition step;
Fig. 3 is the flow diagram of type determining program in discrimination process;
Fig. 4 is the egg type discrimination method flow chart that the embodiment of the present invention 2 provides;
Fig. 5 is the egg type discrimination method flow diagram that the embodiment of the present invention 3 provides;
Fig. 6 is the egg type identification device structural schematic diagram that the embodiment of the present invention 4 provides.
Specific implementation mode
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached in the embodiment of the present invention
Figure, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only this
Invention a part of the embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art exist
The every other embodiment obtained under the premise of creative work is not made, shall fall within the protection scope of the present invention.
Egg type discrimination method and device provided in an embodiment of the present invention, are easily acquired by high light spectrum image-forming technology
The spectrum picture of egg differentiates that model is counted by the pretreatment to spectrum picture using trained egg type in advance
It calculates, to obtain egg type identification result, output is as a result, above-mentioned high light spectrum image-forming technology, than more typical, may be used
The spectrum camera module of Fabry-Perot interferometers design based on MEMS and simple light supply apparatus, corresponding height
Spectroscopic data will be transmitted to progress egg type judgement in built-in discriminating model, so that it may easily collect egg to be identified
High spectrum image, potential principle is analyzed egg inside constituent using the transmittance of light, due to inhomogeneity
Type egg internal chemical ingredient is different, and different degrees of variation can also occur for absorption spectrum.Therefore, client can be by integrated
Or it is loaded with the terminal of above-mentioned spectrum camera module, fast implement the acquisition of egg spectrum picture, processing and identification result
Whole process can thus meet the needs that ordinary populace consumer differentiates egg type in the market, existing to solve
There is the technical issues of pendent quick discriminating egg type in technology, and this method has stronger life practicability,
It can carry out large-scale promotion and application.
With reference to specific embodiment and attached drawing, egg type discrimination method provided in an embodiment of the present invention and device are made
It further illustrates.
Embodiment 1
Fig. 1 is the egg type discrimination method flow chart that the embodiment of the present invention 1 provides, as shown in Figure 1, the embodiment of the present invention
The egg type discrimination method of offer, includes the following steps:
101, the spectrum picture of egg is acquired by high light spectrum image-forming technology.
Specifically, acquiring several spectrum pictures along the equator of egg and/or two extreme directions by high light spectrum image-forming technology.Into
One step preferably, between wavelength 400-1000nm under conditions of at least 100 wave bands, with the crosscutting circumference of egg equatorial direction
Interval predetermined angular acquires several spectrum pictures, and acquires a width spectrum picture respectively in two extreme directions of egg.Here chicken
Egg is egg to be identified.
Illustratively, the Fabry-Perot based on MEMS specifically can be used in high light spectrum image-forming technology here
The spectra collection module (such as spectrum camera module of Unispectral companies on the market) that interferometer design is realized, carries out chicken
The collecting work of egg high spectrum image, this spectra collection module mainly have it is small, light-weight, low in energy consumption, reliability is high,
And the advantages that being easily integrated, acquisition spectrum range includes visible light and part short wavelength-NIR light, and spectral wavelength ranges are general
Between 400-1000nm, and it is generally assembled into the device of such as mobile phone terminal, especially using integrated or be mounted with
The mobile phone terminal of spectra collection module is stated, can realize the quick acquisition spectrum in egg type authentication schemes of the embodiment of the present invention
Image step.
Fig. 2 implements the set-up mode exemplary plot of spectrum picture acquisition step.
The mode that the acquisition step is embodied is as follows:
As shown in Fig. 2, by transmitted light source 1 and the terminal comprising spectra collection module 21 or spectrum camera 2, it is respectively placed in and waits for
Differentiate egg 3 both ends, light source 1 be close to egg 3 placement, acquisition position may be selected along egg 3 equatorial direction and egg 3 two
Extreme direction specifically acquires several spectrum pictures, and in egg 3 with the crosscutting circle spacing predetermined angular of 3 equatorial direction of egg
Two extreme directions acquire a width spectrum picture respectively, what spectrum camera 2 took be by egg internal composition absorption after light
Line.Predetermined angular is preferably 90 °, and the red laser having compared with high penetrating power can be used in light source 1, and image resolution ratio is not less than
500*500, at least 100 wave bands between wavelength 400-1000nm.At least 100 wave bands are general between wavelength 400-1000nm
Needs can be met, if wave-length coverage selection is smaller, wave band quantity can also select smaller range, certainly for figure
As rate respectively, the design parameter range selection of wavelength, wave band quantity, can be carried out choosing according to the concrete condition in practice real
It applies, the embodiment of the present invention does not limit it especially.
102, spectrum picture is pre-processed.
Specifically, intercepting egg target area respectively to spectrum picture, the immediate vicinity in target area randomly selects pre-
The target point of fixed number amount takes the corresponding spectrum of target point to eliminate noise.It is further preferred that passing through user's interactive mode, cluster
Or image segmentation algorithm carries out egg target area interception, is disappeared using smooth, multiplicative scatter correction and/or standard method for normalizing
Except noise.
Illustratively, egg region is intercepted respectively to collected spectrum picture, interception way can be user interaction side
Formula can also automatically be realized with cluster or image segmentation algorithm.Preferably, poly- using Kmeans is carried out in Lab color spaces
Egg region and background area are polymerized to two classes by the mode of class respectively, form binarization segmentation as a result, center class regional center
Target area center as egg.It, can be more simply and effectively using cluster mode because egg shell color is more stable
Egg region is extracted, man-machine interactively is avoided.N number of point is randomly choosed in the egg target area immediate vicinity of interception, takes its correspondence
Spectrum eliminate noise.Preferably, the value of N is not less than 300.Due to carrying out feature learning using deep neural network, need to reach
To certain sample size, usual value gets 300 and can guarantee the diversity of spectrum samples, while being also avoided that a small amount of exceptions
The interference of sample.Spectrum eliminates noise and smooth, multiplicative scatter correction or standard normalization may be used or wherein combine two-by-two
Method improve signal-to-noise ratio to eliminate noise caused by scattering of impurity etc. in light intensity attenuation, egg.Preferably, using spectrum
Smoothing method eliminates noise, smooth in such a way that 5 points in continuum are calculated with average value.Smoothing method can be eliminated
Abnormal under single wave band, under normal circumstances, the influence curve of spectrum smooth enough under continuous 5 wave bands can allow spectrum bent too much
Line dropout detailed information is unable to reach smooth purpose very little.When specific to carrying out type test discriminating, pre- by above-mentioned image
After processing and before being differentiated, in egg target area, immediate vicinity randomly chooses T test point, and the spectrum of test point is made
For the characteristic spectrum of egg, the input of follow-up type identification is carried out, it is preferable that setting T=99.Since test quantity very little may
It will appear the reliable phenomenon of unstable result, more computing resources can be consumed too much, therefore it is one fine that usually T values, which are 99,
Compromise, can guarantee result it is reliable and stable and calculate response quickly.
It should be noted that in interception area process and elimination noise process, implemented using above-mentioned processing method
Concrete mode, the embodiment of the present invention do not limit it especially.
103, differentiate that model obtains egg type identification result using default egg type.
Specifically, default egg type here differentiates model, any possible egg class in the prior art may be used
Type differentiates model to implement used in this programme, and preferably, and model training method instruction provided in an embodiment of the present invention may be used
It gets default egg type and differentiates model.Therefore, differentiating that model obtains egg type and differentiates knot using default egg type
Further include the steps that training obtains default egg type discriminating model before fruit.
Specifically, training, which obtains default egg type, differentiates that model includes the following steps:
The spectrum picture of egg sample is acquired by high light spectrum image-forming technology;
Sample spectra image is pre-processed;
It is trained using predetermined depth confidence network, obtains and preset egg type discriminating model.
Wherein, the step of acquiring the spectrum picture of egg sample by high light spectrum image-forming technology, can be in above-mentioned 101 step
In complete together, be equally all the acquisition of egg spectrum picture, only acquiring the egg object that is directed to has egg to be identified
With for trained egg sample point, alternatively, the step for can also individually be completed with an independent step, this independence
After step can both state the spectrum picture that 101 steps acquire egg by high light spectrum image-forming technology on the implementation, immediately really
Apply, can also 102 steps completion after, corresponding timing node is complete in the discriminating Type model training process before 103 steps
At.Specific implementation process and above-mentioned 101 step that the spectrum picture of egg sample is acquired above by high light spectrum image-forming technology are real
It applies that process is identical, refers to the descriptive text to 101 steps, details are not described herein.
Furthermore it is preferred that the step for acquiring the spectrum picture of egg sample by high light spectrum image-forming technology is in implementation
Before, further include egg sample preparation process, specifically embodiment can be:
It collects N1 Countryside Egg and N2 feed egg composition training set is used for training pattern.Select the production time 30 days with
Interior egg, the removing surface of egg is clean, before gathered data, egg is placed on a period of time under isoperibol.
Pretreated step is carried out to sample spectra image, can together be completed in above-mentioned 102 step, alternatively, with one
A independent step is individually completed, this independent process both can state step 102 on the implementation and be pre-processed to spectrum picture
Later, and then implement, it can also be after the completion of 102 steps, phase in the discriminating Type model training process before 103 steps
Timing node is answered to complete.It is above-mentioned that the pretreated specific implementation process of sample spectra image progress was implemented with above-mentioned 102 step
Cheng Xiangtong refers to the descriptive text to 102 steps, and details are not described herein.
It is trained using predetermined depth confidence network, obtains and preset egg type discriminating model, the step implementation process
It may be used such as under type:
Wherein predetermined depth confidence network is made of 1 one-dimensional convolutional layer and P limited Boltzmann machines (RBM), one-dimensional
Convolutional layer includes the one-dimensional convolution sum pondization composition of S 1*K, input of the data that convolutional layer exports as RBM, RBM internal junctions
Structure includes a visible layer and a hidden layer, in training process, trains up the RBM of training current layer after the RBM of last layer,
Until last layer, finally obtains default egg type and differentiates model M, wherein P, S, K are the integer more than 0, it is preferable that P
>=3, S >=4, K >=5.It can ensure that depth network can fully learn into spectroscopic data potential key feature in this way.
After being trained by above-mentioned training process, model M, which stores, to be differentiated to default egg type, by the model with
Its corresponding evaluator preserves jointly, to be called when type judges.
Then, differentiate that model is differentiated using above-mentioned trained default egg type, obtain egg type and differentiate knot
Fruit.
Specifically, carrying out type identification analysis to unknown sample according to the model that training is established, above-mentioned pretreatment will be passed through
Egg spectroscopic data to be identified be input to the training stage preservation egg type differentiate model M in, obtained by model prediction
The judgement of spectroscopic data type, the spectroscopic data type which is contained by it determine, if most spectrum
Data are identified as Countryside Egg type, then the egg in the width image is considered as then Countryside Egg, is otherwise considered as feed egg.
Fig. 3 is the flow diagram of type determining program in discrimination process, as shown in figure 3, pretreated T spectroscopic data, respectively
It is input in model M, judges whether to be Countryside Egg, it is assumed that have n in T spectroscopic data judging resultpA Countryside Egg type spectrum
Data (i.e. npFor Countryside Egg spectrum number), nnA feed eggs type spectroscopic data (i.e. nnFor Countryside Egg spectrum number), T=np+
nn, if that np>nnIt is Countryside Egg to be considered as the egg, otherwise it is assumed that being feed egg.It is handled in this manner, can keep away
Exempt to generate error in judgement since outside noise interferes, makes model that there is better robustness.
Type identification result, i.e. Countryside Egg or feed egg are provided respectively for every width spectrum picture.The above-mentioned type judges
After process is completed, identification result is exported.As a result output can be JSON formats, be made of object key-value pair,
(“imageName”:" xxxx " (for example, " newrawfile20180403165909 "), " eggType ":“free_range/
Feed "), this group of data can be used for other routine calls, show result.
Embodiment 2
Fig. 4 is the egg type discrimination method flow chart that the embodiment of the present invention 2 provides, as shown in figure 4, the embodiment of the present invention
The egg type discrimination method of offer, includes the following steps:
201, the spectrum picture of egg is acquired by high light spectrum image-forming technology, wherein egg includes for trained egg sample
Product and egg to be identified.
Before acquisition, egg to be identified is got out, and carry out egg sample preparation:
Egg of the production time within 30 days is selected, the removing surface of egg is clean, before gathered data, egg is put
Set a period of time under isoperibol.
Here the Fabry-Perot interferometers design based on MEMS specifically can be used in high light spectrum image-forming technology
The spectra collection module (such as spectrum camera module of Unispectral companies on the market) of realization carries out egg high-spectrum
The collecting work of picture, this spectra collection module mainly have that small, light-weight, low in energy consumption, reliability is high, and are easy to collect
At the advantages that, acquisition spectrum range include visible light and part short wavelength-NIR light, spectral wavelength ranges are generally in 400-
Between 1000nm, and it is generally assembled into the device of such as mobile phone terminal, especially using integrated or be mounted with above-mentioned spectrum
The mobile phone terminal of acquisition module can realize the quick acquisition spectrum picture step in egg type authentication schemes of the embodiment of the present invention
Suddenly.
The mode that the acquisition step is embodied is as follows:
Fig. 2 is returned, transmitted light source and the terminal comprising spectra collection module or spectrum camera are respectively placed in be identified
The both ends of egg, light source are close to egg placement, and two extreme directions along equatorial direction and egg may be selected in acquisition position, specifically,
Several spectrum pictures are acquired with the crosscutting circle spacing predetermined angular of egg equatorial direction, and are adopted respectively in two extreme directions of egg
Collect a width spectrum picture, what spectrum camera took is the light after the absorption of egg internal composition.Predetermined angular is preferably
It it is 90 °, light source can be used with the red laser compared with high penetrating power, and image resolution ratio is not less than 500*500, in wavelength 400-
At least 100 wave bands between 1000nm.At least 100 wave bands can generally meet needs between wavelength 400-1000nm, if
If wave-length coverage selection is smaller, wave band quantity can also select smaller range, certainly for image rate, wavelength, wave respectively
The design parameter range of segment number selects, and selection implementation can be carried out according to the concrete condition in practice, the embodiment of the present invention is not
It is especially limited.
It is worth noting that, step 201 process can also pass through its other party other than the mode described in above-mentioned steps
Formula realizes that the process, the embodiment of the present invention are not limited specific mode.
202, spectrum picture is pre-processed.
Specifically, intercepting egg target area respectively to spectrum picture, the immediate vicinity in target area randomly selects pre-
The target point of fixed number amount takes the corresponding spectrum of target point to eliminate noise.It is further preferred that passing through user's interactive mode, cluster
Or image segmentation algorithm carries out egg target area interception, is disappeared using smooth, multiplicative scatter correction and/or standard method for normalizing
Except noise.
Illustratively, egg region is intercepted respectively to collected spectrum picture, interception way can be user interaction side
Formula can also automatically be realized with cluster or image segmentation algorithm.Preferably, poly- using Kmeans is carried out in Lab color spaces
Egg region and background area are polymerized to two classes by the mode of class respectively, form binarization segmentation as a result, center class regional center
Target area center as egg.It, can be more simply and effectively using cluster mode because egg shell color is more stable
Egg region is extracted, man-machine interactively is avoided.N number of point is randomly choosed in the egg target area immediate vicinity of interception, takes its correspondence
Spectrum eliminate noise.Preferably, the value of N is not less than 300.Due to carrying out feature learning using deep neural network, need to reach
To certain sample size, usual value gets 300 and can guarantee the diversity of spectrum samples, while being also avoided that a small amount of exceptions
The interference of sample.Spectrum eliminates noise and smooth, multiplicative scatter correction or standard normalization may be used or wherein combine two-by-two
Method improve signal-to-noise ratio to eliminate noise caused by scattering of impurity etc. in light intensity attenuation, egg.Preferably, using spectrum
Smoothing method eliminates noise, smooth in such a way that 5 points in continuum are calculated with average value.Smoothing method can be eliminated
Abnormal under single wave band, under normal circumstances, the influence curve of spectrum smooth enough under continuous 5 wave bands can allow spectrum bent too much
Line dropout detailed information is unable to reach smooth purpose very little.When specific to carrying out type test discriminating, pre- by above-mentioned image
After processing and before being differentiated, in egg target area, immediate vicinity randomly chooses T test point, and the spectrum of test point is made
For the characteristic spectrum of egg, the input of follow-up type identification is carried out, it is preferable that setting T=99.Since test quantity very little may
It will appear the reliable phenomenon of unstable result, more computing resources can be consumed too much, therefore it is one fine that usually T values, which are 99,
Compromise, can guarantee result it is reliable and stable and calculate response quickly.It should be noted that in interception area process and disappearing
Except noise process, the concrete mode implemented using above-mentioned processing method, the embodiment of the present invention does not limit it especially.
It is worth noting that, step 202 process can also pass through its other party other than the mode described in above-mentioned steps
Formula realizes that the process, the embodiment of the present invention are not limited specific mode.
203, training obtains default egg type discriminating model.
It is trained using predetermined depth confidence network, obtains and preset egg type discriminating model, the step implementation process
It may be used such as under type:
Wherein predetermined depth confidence network is made of 1 one-dimensional convolutional layer and P limited Boltzmann machines, one-dimensional convolutional layer
Include the one-dimensional convolution sum pondization composition of S 1*K, the data of convolutional layer output are limited as the input for being limited Boltzmann machine
Boltzmann machine internal structure includes a visible layer and a hidden layer, in training process, trains up the limited of last layer
The limited Boltzmann machine of training current layer after Boltzmann machine, until last layer, finally obtains default egg type and differentiate
Model M, wherein P, S, K are the integer more than 0, it is preferable that P >=3, S >=4, K >=5.
After being trained by above-mentioned training process, model M, which stores, to be differentiated to default egg type, by the model with
Its corresponding evaluator preserves jointly, to be called when type judges.
It is worth noting that, step 203 process can also pass through its other party other than the mode described in above-mentioned steps
Formula realizes that the process, the embodiment of the present invention are not limited specific mode.
204, differentiate that model obtains egg type identification result using default egg type.
Specifically, carrying out type identification analysis to unknown sample according to the model that training is established, above-mentioned pretreatment will be passed through
Egg spectroscopic data to be identified be input to the training stage preservation egg type differentiate model M in, obtained by model prediction
The judgement of spectroscopic data type, the spectroscopic data type which is contained by it determine, if most spectrum
Data are identified as Countryside Egg type, then the egg in the width image is considered as then Countryside Egg, is otherwise considered as feed egg.
Fig. 3 is the flow diagram of type determining program in discrimination process, as shown in figure 3, pretreated T spectroscopic data, respectively
It is input in model M, judges whether to be Countryside Egg, it is assumed that have n in T spectroscopic data judging resultpA Countryside Egg type spectrum
Data, nnA feed eggs type spectroscopic data, T=np+nn, if that np>nnIt is Countryside Egg to be considered as the egg, otherwise it is assumed that
It is feed egg.It is handled in this manner, error in judgement can be generated to avoid being interfered due to outside noise, model is made to have more
Good robustness.
Type identification result, i.e. Countryside Egg or feed egg are provided respectively for every width spectrum picture.The above-mentioned type judges
After process is completed, identification result is exported.As a result output can be JSON formats, be made of object key-value pair,
(“imageName”:" xxxx " (for example, " newrawfile20180403165909 "), " eggType ":“free_range/
Feed "), this group of data can be used for other routine calls, show result.
It is worth noting that, step 204 process can also pass through its other party other than the mode described in above-mentioned steps
Formula realizes that the process, the embodiment of the present invention are not limited specific mode.
Embodiment 3
Fig. 5 is the egg type discrimination method flow diagram that the embodiment of the present invention 3 provides, as shown in figure 5, the present invention is real
The egg type discrimination method for applying example offer, includes the following steps:
301, the spectrum picture of egg sample is acquired by high light spectrum image-forming technology.
Here the spectrum picture gatherer process for the egg sample that gatherer process is related to 201 steps in embodiment 2 is implemented
Mode is identical, specifically refers to its description, details are not described herein.
It is worth noting that, step 301 process can also pass through its other party other than the mode described in above-mentioned steps
Formula realizes that the process, the embodiment of the present invention are not limited specific mode.
302, the spectrum picture of egg sample is pre-processed.
Here gatherer process carries out preprocessing process embodiment party with what 202 steps in embodiment 2 were related to spectrum picture
Formula is identical, specifically refers to its description, details are not described herein.
It is worth noting that, step 302 process can also pass through its other party other than the mode described in above-mentioned steps
Formula realizes that the process, the embodiment of the present invention are not limited specific mode.
303, training obtains default egg type discriminating model.
Here the training of gatherer process and 203 steps in embodiment 2 obtains default egg type and differentiates that model process is real
It is identical to apply mode, specifically refers to its description, details are not described herein.
It is worth noting that, step 303 process can also pass through its other party other than the mode described in above-mentioned steps
Formula realizes that the process, the embodiment of the present invention are not limited specific mode.
304, the spectrum picture of egg to be identified is acquired by high light spectrum image-forming technology.
Here the spectrum picture gatherer process for the egg sample that gatherer process is related to 201 steps in embodiment 2 is implemented
Mode is identical, specifically refers to its description, details are not described herein.
It is worth noting that, step 304 process can also pass through its other party other than the mode described in above-mentioned steps
Formula realizes that the process, the embodiment of the present invention are not limited specific mode.
305, the spectrum picture of egg to be identified is pre-processed.
Here gatherer process carries out preprocessing process embodiment party with what 202 steps in embodiment 2 were related to spectrum picture
Formula is identical, specifically refers to its description, details are not described herein.
It is worth noting that, step 305 process can also pass through its other party other than the mode described in above-mentioned steps
Formula realizes that the process, the embodiment of the present invention are not limited specific mode.
306, differentiate that model obtains egg type identification result using default egg type.
Here the utilization of gatherer process and 204 steps in embodiment 2 presets egg type and differentiates that model obtains egg class
Type identification result process embodiment is identical, specifically refers to its description, details are not described herein.
It is worth noting that, step 306 process can also pass through its other party other than the mode described in above-mentioned steps
Formula realizes that the process, the embodiment of the present invention are not limited specific mode.
Embodiment 4
Fig. 6 is the egg type identification device structural schematic diagram that the embodiment of the present invention 4 provides, as shown in fig. 6, the present invention is real
The egg type identification device for applying example offer includes that spectrum picture acquisition module, preprocessing module and identification result obtain mould
Block.
Specifically, spectrum picture acquisition module is used to acquire the spectrum picture of egg by high light spectrum image-forming technology;Pre- place
Module is managed, for being pre-processed to spectrum picture;Identification result acquisition module, for differentiating model using default egg type
Obtain egg type identification result.Here egg includes the egg sample and egg to be identified for model training.
Preferably, spectrum picture acquisition module is used for:
By high light spectrum image-forming technology several spectrum pictures are acquired along the equator of egg and/or two extreme directions.It is further excellent
Selection of land, spectrum picture acquisition module are used for:
Between wavelength 400-1000nm under conditions of at least 100 wave bands, between the crosscutting circumference of egg equatorial direction
Several spectrum pictures are acquired every predetermined angular, and a width spectrum picture is acquired respectively in two extreme directions of egg.
Preferably, preprocessing module is used for:Intercept egg target area respectively to spectrum picture, the center in target area
The target point for nearby randomly selecting predetermined quantity takes the corresponding spectrum of target point to eliminate noise.It is further preferred that pretreatment mould
Block carries out egg target area interception by user's interactive mode, cluster or image segmentation algorithm, utilizes smooth, polynary scattering school
Just and/or standard method for normalizing eliminates noise.
Furthermore it is preferred that above-mentioned apparatus further includes:
Model training module obtains default egg type for training and differentiates model.
Preferably, model training module is used for:
It is trained using predetermined depth confidence network, obtains and preset egg type discriminating model.
It is further preferred that predetermined depth confidence network is by 1 one-dimensional convolutional layer and P limited Boltzmann machines (RBM)
Composition, one-dimensional convolutional layer include that the one-dimensional convolution sum pondization of S 1*K forms, input of the data that convolutional layer exports as RBM,
RBM internal structures include a visible layer and a hidden layer, in training process, after training up the RBM of last layer training work as
The RBM of front layer, until last layer, finally obtains default egg type and differentiate model M, wherein P, S, K are whole more than 0
Number.
Additionally preferably, above-mentioned apparatus further includes:
Model memory module, for differentiating that model stores to trained default egg type;
Identification result output module, the egg types results output for obtaining identification result acquisition module, in order to
Other programs are applied or are directly output to user and obtain information interface.
It should be noted that:The egg type identification device that above-described embodiment provides is in triggering egg type authentication service
When, only the example of the division of the above functional modules, in practical application, above-mentioned function can be divided as needed
With by different function module completions, i.e., the internal structure of device is divided into different function modules, to complete above description
All or part of function.In addition, the egg type identification device and egg type discrimination method that above-described embodiment provides are real
It applies example and belongs to same design, specific implementation process refers to embodiment of the method, and which is not described herein again.
The alternative embodiment that any combination forms the present invention may be used, herein no longer in above-mentioned all optional technical solutions
It repeats one by one.
In conclusion egg type discrimination method provided in an embodiment of the present invention and device, have compared with the existing technology
Following advantageous effect:
1, the spectrum picture of egg is easily acquired by high light spectrum image-forming technology, it is only necessary to by easily acquiring egg
High spectrum image then differentiated, export identification result, can quickly differentiate egg type;
2, judged, will not be had any impact to the quality and structure of egg have using the diffusing transmission spectrum of light source
There is the advantages of Undamaged determination;
3, the discriminating model that the embodiment of the present invention utilizes is to learn to obtain by deep neural network, to external environment robust
Property it is higher, even if under complex environment, still have good distinguishing ability;
4, egg classification discrimination method and device provided in an embodiment of the present invention, discrimination process is easy to operate, has relatively strong
Practicability, large-scale promotion and application can be carried out.
One of ordinary skill in the art will appreciate that realizing that all or part of step of above-described embodiment can pass through hardware
It completes, relevant hardware can also be instructed to complete by program, the program can be stored in a kind of computer-readable
In storage medium, storage medium mentioned above can be read-only memory, disk or CD etc..
It is with reference to according to the method for embodiment, equipment (system) and calculating in the embodiment of the present application in the embodiment of the present application
The flowchart and/or the block diagram of machine program product describes.It should be understood that can be realized by computer program instructions flow chart and/or
The combination of the flow and/or box in each flow and/or block and flowchart and/or the block diagram in block diagram.It can carry
For the processing of these computer program instructions to all-purpose computer, special purpose computer, Embedded Processor or other programmable datas
The processor of equipment is to generate a machine so that is executed by computer or the processor of other programmable data processing devices
Instruction generation refer to for realizing in one flow of flow chart or multiple flows and/or one box of block diagram or multiple boxes
The device of fixed function.
These computer program instructions, which may also be stored in, can guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works so that instruction generation stored in the computer readable memory includes referring to
Enable the manufacture of device, the command device realize in one flow of flow chart or multiple flows and/or one box of block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device so that count
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, in computer or
The instruction executed on other programmable devices is provided for realizing in one flow of flow chart or multiple flows and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Although the preferred embodiment in the embodiment of the present application has been described, once a person skilled in the art knows
Basic creative concept, then additional changes and modifications may be made to these embodiments.So appended claims are intended to explain
It is to include preferred embodiment and fall into all change and modification of range in the embodiment of the present application.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
God and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent replacement, improvement and so on should all be included in the protection scope of the present invention.
Claims (18)
1. a kind of egg type discrimination method, which is characterized in that including:
The spectrum picture of egg is acquired by high light spectrum image-forming technology;
The spectrum picture is pre-processed;
Differentiate that model obtains egg type identification result using default egg type.
2. according to the method described in claim 1, it is characterized in that, acquiring the spectrogram of egg by high light spectrum image-forming technology
Picture, including:
By high light spectrum image-forming technology several spectrum pictures are acquired along the equator of egg and/or two extreme directions.
3. according to the method described in claim 2, it is characterized in that, acquiring the spectrogram of egg by high light spectrum image-forming technology
Picture, including:
It is pre- with the crosscutting circle spacing of egg equatorial direction between wavelength 400-1000nm under conditions of at least 100 wave bands
Determine angle acquisition several spectrum pictures, and a width spectrum picture is acquired respectively in two extreme directions of egg.
4. according to the method described in claim 1, it is characterized in that, pre-processed to the spectrum picture, including:
Egg target area is intercepted respectively to the spectrum picture, the immediate vicinity in the target area randomly selects predetermined number
The target point of amount takes the corresponding spectrum of the target point to eliminate noise.
5. according to the method described in claim 4, it is characterized in that, passing through user's interactive mode, cluster or image segmentation algorithm
Egg target area interception is carried out, noise is eliminated using smooth, multiplicative scatter correction and/or standard method for normalizing.
6. the method according to claims 1 to 3, which is characterized in that differentiating that model obtains chicken using default egg type
Before eggs type identification result, the method further includes:
Training obtains the default egg type and differentiates model.
7. according to the method described in claim 6, it is characterized in that, training, which obtains the default egg type, differentiates model, packet
It includes:
The spectrum picture of egg sample is acquired by high light spectrum image-forming technology;
The spectrum picture of the egg sample is pre-processed;
It is trained using predetermined depth confidence network, obtains the default egg type and differentiate model.
8. the method according to the description of claim 7 is characterized in that acquiring the spectrum of egg sample by high light spectrum image-forming technology
The step of image, completes in the spectrum picture step for acquiring egg by high light spectrum image-forming technology, alternatively, individually completing.
9. the method according to the description of claim 7 is characterized in that pretreated step is carried out to the spectrum picture, right
The spectrum picture completed in pre-treatment step, alternatively, individually completing.
10. according to claim 7-9 any one of them methods, which is characterized in that instructed using predetermined depth confidence network
Practice, obtains the default egg type and differentiate model, including:
The predetermined depth confidence network is made of 1 one-dimensional convolutional layer and P limited Boltzmann machines, and one-dimensional convolutional layer includes
The data of the one-dimensional convolution sum pondization composition of S 1*K, convolutional layer output are limited Bohr as the input for being limited Boltzmann machine
Hereby graceful machine internal structure includes a visible layer and a hidden layer, in training process, trains up limited Bohr of last layer
Hereby the limited Boltzmann machine of training current layer finally obtains default egg type and differentiates model up to last layer after graceful machine
M, wherein P >=3, S >=4, K >=5.
11. a kind of egg type identification device, which is characterized in that including:
Spectrum picture acquisition module, the spectrum picture for acquiring egg by high light spectrum image-forming technology;
Preprocessing module, for being pre-processed to the spectrum picture;
Identification result acquisition module, for differentiating that model obtains egg type identification result using default egg type.
12. according to the devices described in claim 11, which is characterized in that the spectrum picture acquisition module is used for:
By high light spectrum image-forming technology several spectrum pictures are acquired along the equator of egg and/or two extreme directions.
13. device according to claim 12, which is characterized in that the spectrum picture acquisition module is used for:
It is pre- with the crosscutting circle spacing of egg equatorial direction between wavelength 400-1000nm under conditions of at least 100 wave bands
Determine angle acquisition several spectrum pictures, and a width spectrum picture is acquired respectively in two extreme directions of egg.
14. according to the devices described in claim 11, which is characterized in that the preprocessing module is used for:
Egg target area is intercepted respectively to the spectrum picture, the immediate vicinity in the target area randomly selects predetermined number
The target point of amount takes the corresponding spectrum of the target point to eliminate noise.
15. device according to claim 14, which is characterized in that the preprocessing module passes through user's interactive mode, poly-
Class or image segmentation algorithm carry out egg target area interception, utilize smooth, multiplicative scatter correction and/or standard method for normalizing
Eliminate noise.
16. according to the method for claim 11, which is characterized in that described device further includes:
Model training module obtains the default egg type for training and differentiates model.
17. device according to claim 16, which is characterized in that the model training module is used for:
It is trained using predetermined depth confidence network, obtains the default egg type and differentiate model.
18. device according to claim 17, which is characterized in that
The predetermined depth confidence network is made of 1 one-dimensional convolutional layer and P limited Boltzmann machines, and one-dimensional convolutional layer includes
The data of the one-dimensional convolution sum pondization composition of S 1*K, convolutional layer output are limited Bohr as the input for being limited Boltzmann machine
Hereby graceful machine internal structure includes a visible layer and a hidden layer, in training process, trains up limited Bohr of last layer
Hereby the limited Boltzmann machine of training current layer finally obtains default egg type and differentiates model up to last layer after graceful machine
M, wherein P, S, K are the integer more than 0.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810444505.5A CN108647675A (en) | 2018-05-10 | 2018-05-10 | A kind of egg type discrimination method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810444505.5A CN108647675A (en) | 2018-05-10 | 2018-05-10 | A kind of egg type discrimination method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108647675A true CN108647675A (en) | 2018-10-12 |
Family
ID=63754239
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810444505.5A Pending CN108647675A (en) | 2018-05-10 | 2018-05-10 | A kind of egg type discrimination method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108647675A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109191461A (en) * | 2018-10-22 | 2019-01-11 | 广东工业大学 | A kind of Countryside Egg recognition methods and identification device based on machine vision technique |
CN109916901A (en) * | 2019-02-26 | 2019-06-21 | 中国农业科学院农产品加工研究所 | The method for quick identification of calm and peaceful black-bone chicken egg and hybridization black-bone chicken egg |
CN111751297A (en) * | 2020-07-06 | 2020-10-09 | 广东智芯光电科技有限公司 | Method and system for identifying quality of fresh product based on camera |
CN114916469A (en) * | 2022-05-27 | 2022-08-19 | 浙江大学 | Optical identification method for gender of breeding eggs of laying hens |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104198398A (en) * | 2014-08-07 | 2014-12-10 | 浙江大学 | Method for distinguishing donkey-hide gelatin based on hyperspectral imaging |
CN106290238A (en) * | 2016-08-09 | 2017-01-04 | 西北农林科技大学 | A kind of apple variety method for quick identification based on high light spectrum image-forming |
CN107423670A (en) * | 2017-04-20 | 2017-12-01 | 上海交通大学 | MODIS mist monitoring methods based on depth confidence network |
-
2018
- 2018-05-10 CN CN201810444505.5A patent/CN108647675A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104198398A (en) * | 2014-08-07 | 2014-12-10 | 浙江大学 | Method for distinguishing donkey-hide gelatin based on hyperspectral imaging |
CN106290238A (en) * | 2016-08-09 | 2017-01-04 | 西北农林科技大学 | A kind of apple variety method for quick identification based on high light spectrum image-forming |
CN107423670A (en) * | 2017-04-20 | 2017-12-01 | 上海交通大学 | MODIS mist monitoring methods based on depth confidence network |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109191461A (en) * | 2018-10-22 | 2019-01-11 | 广东工业大学 | A kind of Countryside Egg recognition methods and identification device based on machine vision technique |
CN109916901A (en) * | 2019-02-26 | 2019-06-21 | 中国农业科学院农产品加工研究所 | The method for quick identification of calm and peaceful black-bone chicken egg and hybridization black-bone chicken egg |
CN109916901B (en) * | 2019-02-26 | 2021-07-20 | 中国农业科学院农产品加工研究所 | Rapid identification method for Taihe black-bone chicken eggs and hybrid black-bone chicken eggs |
CN111751297A (en) * | 2020-07-06 | 2020-10-09 | 广东智芯光电科技有限公司 | Method and system for identifying quality of fresh product based on camera |
CN114916469A (en) * | 2022-05-27 | 2022-08-19 | 浙江大学 | Optical identification method for gender of breeding eggs of laying hens |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108647675A (en) | A kind of egg type discrimination method and device | |
Liu et al. | Data synthesis using deep feature enhanced generative adversarial networks for rolling bearing imbalanced fault diagnosis | |
CN104834933B (en) | A kind of detection method and device in saliency region | |
CN104215623B (en) | Laser Raman spectroscopy intelligence discrimination method and system towards conglomerate detection | |
CN107346434A (en) | A kind of plant pest detection method based on multiple features and SVMs | |
CN111027378B (en) | Pedestrian re-identification method, device, terminal and storage medium | |
CN108229485A (en) | For testing the method and apparatus of user interface | |
CN107688530A (en) | Method for testing software and device | |
CN102959354A (en) | Method and apparatus for for analyzing tree canopies with LiDAR data | |
CN109580534A (en) | Substance classes recognition methods, terahertz time-domain spectroscopy instrument system and terminal device | |
Bansal et al. | Green citrus detection using fast Fourier transform (FFT) leakage | |
CN107330731A (en) | It is a kind of to recognize that advertisement position clicks on abnormal method and apparatus | |
CN104062262A (en) | Crop seed variety authenticity identification method based on near infrared spectrum | |
CN107044959A (en) | Micro- multi-modal fusion spectral detection system | |
CN108579094A (en) | A kind of user interface detection method and relevant apparatus, system and storage medium | |
CN110188796A (en) | User identification method, device, storage medium and processor | |
Dai Pra et al. | Dynamic speckle image segmentation using self-organizing maps | |
CN107273890A (en) | Graphical verification code recognition methods and device for character combination | |
CN108171208A (en) | Information acquisition method and device | |
Ali et al. | Detecting fraudulent labeling of rice samples using computer vision and fuzzy knowledge | |
Mohamadzadeh Moghadam et al. | Nondestructive classification of saffron using color and textural analysis | |
CN107844970A (en) | Image-recognizing method and retrospect scale terminal | |
CN108875536A (en) | Pedestrian's analysis method, device, system and storage medium | |
Sujatha et al. | Innovation in agriculture industry by automated sorting of rice grains | |
Sharma et al. | Plant disease identification using deep learning: a systematic review |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181012 |