CN111999257B - Optical fiber spectrum sorting device and method for male and female hatching egg embryos during hatching - Google Patents
Optical fiber spectrum sorting device and method for male and female hatching egg embryos during hatching Download PDFInfo
- Publication number
- CN111999257B CN111999257B CN202010913816.9A CN202010913816A CN111999257B CN 111999257 B CN111999257 B CN 111999257B CN 202010913816 A CN202010913816 A CN 202010913816A CN 111999257 B CN111999257 B CN 111999257B
- Authority
- CN
- China
- Prior art keywords
- male
- hatching
- female
- optical fiber
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
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
- 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
-
- 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
- G06N3/047—Probabilistic or stochastic networks
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Abstract
The invention discloses an optical fiber spectrum sorting device for hatching egg male and female embryos during incubation, which comprises: the bottom of the light chamber is provided with a focusing lens, and hatching eggs during incubation are placed above the focusing lens; the condensing lens is connected with a halogen light source through a second optical fiber; the halogen light source is connected with a voltage stabilizing module, and the voltage stabilizing module is connected with a light source adjusting knob and a lithium battery; a collimating lens is arranged above the light chamber, the collimating lens and the condensing lens are arranged in a vertically corresponding manner, the collimating lens is connected with a fiber optic spectrometer through an optical fiber, and the fiber optic spectrometer is connected with an ARM processor; the ARM processor is connected with a lithium battery and a touch display screen; the sorting device disclosed by the invention can be used for carrying out rapid nondestructive testing on male and female embryonated eggs in hatching eggs during hatching.
Description
Technical Field
The invention relates to the technical field of agricultural product grading detection and livestock and poultry breeding and hatching, in particular to an optical fiber spectrum sorting device and method for male and female hatching eggs during hatching.
Background
China is a big country for producing poultry eggs and poultry meat, billions of young poultry need to be hatched every year to meet the domestic poultry breeding demand, female individuals tend to be bred in the poultry egg production process, and as male poultry grow fast and have high meat production rate, farms and farmers prefer male individuals in poultry meat production. If the nondestructive detection of the male and female embryonic eggs can be realized during hatching egg hatching, the hatching industry can hatch the female birds and the male birds as required, and huge economic benefits can be generated. At present, the sex of hatching eggs cannot be detected under the condition of no damage by using machines or manpower in China, and the non-destructive detection of male and female embryonated eggs during hatching of the hatching eggs is a problem which needs to be solved urgently in the hatching industry.
At present, scholars at home and abroad make a lot of researches on the detection of hatching egg gender, and main research means comprise machine vision, spectrum, damage detection and the like. The nondestructive testing aspect comprises: the male and female identification is carried out by utilizing a machine vision technology (congratulating aspiration, etc., 2018, reported in agricultural engineering) according to the blood line characteristics of the chicken embryo incubated for 4d, the accuracy rate is 83.33%, but the model detection speed is poor due to excessive extracted image characteristics, and the model identification method is difficult to apply to actual production. The gender of the chick embryo incubated for 10 days is judged by using a hyperspectral imaging technology (Panyiqing, 2016, agricultural engineering report) with the judgment precision of 82.86%, but the hyperspectral imaging method is high in cost, slow in detection speed and not suitable for large-scale application in actual production. Ultraviolet-visible spectrum (congratulating, etc. 2019, spectroscopy and spectral analysis) is utilized to find that the chick embryos can be subjected to male and female discrimination when being incubated for 7d, the discrimination accuracy is 87.14%, but the established model has the risk of overfitting and can be applied to actual production to be further checked. And (3) destructive detection: sex determination was performed using differences in estrone sulfate content in allantoic fluid from chick embryos incubated 9d (Weiss-mann et al, 2013, Theriogenology). The sex of the chick embryos incubated for 5-7d was identified by PCR (Turkyilmaz et al, 2010, British Poultry Science). The detection precision of the destructive detection means is high, but the operation is complex, the detection cost is high, and the speed is slow.
Therefore, there is a need for a device and a method for sorting male and female information of hatching eggs during hatching, which are fast, stable and efficient.
Disclosure of Invention
The invention aims to provide an optical fiber spectrum sorting device and method for male and female hatching egg embryos during hatching, so as to solve the problems in the prior art.
In order to achieve the purpose, the invention provides the following scheme: the invention provides an optical fiber spectrum sorting device for male and female hatching egg embryos during incubation, which comprises a light chamber and a control box, wherein a focusing lens, an objective table and a collimating lens are sequentially arranged in the light chamber from top to bottom, the focusing lens is arranged at the top of the light chamber, the objective table is positioned in the middle of the light chamber and fixedly connected to the inner wall of the light chamber, the focusing lens penetrates through the bottom of the light chamber and is arranged at the bottom of the light chamber, hatching eggs are placed on the objective table, and the focusing lens, the collimating lens and the hatching eggs are positioned on the same vertical line;
the control box is internally provided with a voltage stabilizing module, a halogen light source, a lithium battery, an ARM processor and a fiber optic spectrometer, a light source adjusting knob is fixedly connected to the outer side wall of the control box and electrically connected with the voltage stabilizing module, the voltage stabilizing module is fixed to one inner side wall of the control box, the fiber optic spectrometer is arranged on one side, away from the light source adjusting knob, of the voltage stabilizing module, the voltage stabilizing module is respectively and electrically connected with the halogen light source and the lithium battery, the halogen light source is installed on the voltage stabilizing module, the lithium battery is fixed to the other side wall of the control box, a touch display screen is fixedly installed on the front side wall of the control box, the ARM processor is arranged on the touch display screen, and the ARM processor is respectively and electrically connected with the touch display screen, the lithium battery and the fiber optic spectrometer;
the bottom of the focusing lens is connected with the control box through a second optical fiber, and the bottom of the collimating lens is connected with the control box through a first optical fiber.
Preferably, the light chamber includes casing and side door, the side door articulate on the casing, fixed mounting has a rack on the inner wall of casing top, the through-hole has been seted up to the bottom of rack, the mounting hole has been seted up at the casing top, collimating lens passes through the mounting hole and installs in the rack, the round hole has been seted up on the objective table, the hatching egg is placed in the round hole.
Preferably, one end of the first optical fiber penetrates through the control box and is fixedly connected to the optical fiber spectrometer, and the other end of the first optical fiber is fixedly connected to the bottom of the collimating lens; one end of the second optical fiber penetrates through the control box and is inserted on the side wall of the control box, the second optical fiber is arranged corresponding to the halogen light source, and the other end of the second optical fiber is fixedly connected to the bottom of the focusing lens.
The invention also provides an optical fiber spectrum sorting method of the hatching egg male and female embryos during incubation, which comprises the following steps:
s1, building an optical fiber spectrum sorting device for hatching egg male and female embryos during hatching;
s2, collecting the spectrum information of the hatching eggs during the incubation period by adopting the optical fiber spectrum sorting device built in the step S1;
s3, processing the spectral information of the hatching eggs during the hatching period collected in the step S2 to obtain a two-dimensional spectral information matrix;
s4, constructing a hatching egg male and female information identification network based on deep learning and the two-dimensional spectral feature matrix obtained in the step S3, and training the hatching egg male and female information identification network;
and S5, inputting the two-dimensional spectral feature matrix constructed in the step S3 into the hatching egg male and female information recognition model trained in the step S4, completing recognition of female embryonated eggs and male embryonated eggs, and displaying recognition information on a touch display screen.
Preferably, the specific implementation steps in step S3 are as follows:
s3.1, correcting and removing noise from the spectral information collected in the step S2 to obtain a sample set;
s3.2, performing labeling processing on the sample set obtained in the step S3.1, and dividing the sample set into a training set, a development set and a test set according to a preset proportion;
s3.3, finding out a set of characteristic wavelength points capable of reflecting the gender information of the hatching eggs from the samples of the training set in the step S3.2, and realizing the dimension reduction of the spectral information;
and S3.4, constructing a two-dimensional spectral feature matrix based on the characteristic wavelength points selected from the training set samples in the step S3.3.
Preferably, in step S4, a hatching egg (7) sex information identification model is established by using a 5-layer convolutional neural network and a two-dimensional spectral information matrix, the 5-layer convolutional neural network includes 1 input layer, 3 convolutional layers, and 1 fully-connected layer, and the 5-layer convolutional neural network is designed as follows:
s4.1, input layers (Inputs): converting the characteristic wavelength points selected from the training set samples in the step S3.3 into a two-dimensional spectrum matrix as the input of a network;
s4.2, convolutional layer 1(conv 1): the number of convolution kernels is not less than 60; after convolution, the input layer in the step S4.1 is activated by using a ReLU function;
s4.3, convolutional layer 2(conv 2): the number of convolution kernels is not less than 190, after the output in the step S4.2 is subjected to convolution, LRN operation and ReLU activation are carried out, and the characteristic matrix is output to the convolution layer 3;
s4.4, convolutional layer 3(conv 3): the number of convolution kernels is not less than 380, and the ReLU and LRN operations are added to the output of the convolution layer 2 in the step S4.3 and then the output is output to the full connection layer;
s4.5, full connectivity layer (FC): the number of the neurons is more than 500, and the output of the convolution layer 3 in the step S4.4 is leveled up to be more than 500 weight values;
s4.6, Output layer (Output): and (5) respectively obtaining the score coefficients of the female embryonated egg and the male embryonated egg by the weight values output by the full connection layer in the step S4.5 through a softmax function.
Preferably, the female embryonated egg in the step S3.2 is labeled as 0, the male embryonated egg is labeled as 1, and the ratio of the training set, the development set and the test set in the step S3.2 is 7:2: 1.
Preferably, the identification of the female embryonated egg and the male embryonated egg is completed according to the score coefficients of the female embryonated egg and the male embryonated egg obtained in the step S4.6, and finally 0 or 1 is output, and the female embryonated egg or the male embryonated egg is displayed on the touch display screen.
Preferably, the characteristic wavelength point set selected in step S3.4 is a one-dimensional spectral characteristic matrix, and the one-dimensional spectral characteristic matrix is converted into a two-dimensional spectral information matrix by multiplying the one-dimensional spectral characteristic matrix by a transpose matrix of the one-dimensional spectral characteristic matrix.
Preferably, in step S4, a hatching egg male and female information identification model is established by using a 5-layer convolutional neural network and a two-dimensional spectral information matrix.
Preferably, in step S4, the male and female information recognition model of the hatching egg is trained through a training set, and the training times is 15000-25000 times, preferably 20000 times.
The invention discloses the following technical effects:
(1) the invention utilizes the optical fiber spectrum to collect the spectrum of the hatching egg during the incubation period, the optical fiber spectrum is sensitive to the difference of the internal substances of the female embryonated egg and the male embryonated egg during the incubation period, and the spectral information of the hatching egg can be effectively obtained; meanwhile, the method constructs the female and male embryo and egg identification model based on deep learning, and can realize rapid and accurate nondestructive detection on the gender information of the hatching eggs during the incubation period;
(2) according to the invention, the collected hatching egg spectral information during the incubation period is subjected to effective spectral information interception to remove noise data, and the wavelength points capable of reflecting the difference between female embryonated eggs and male embryonated eggs during the incubation period are obtained through selection of the characteristic wavelength points, so that the calculated amount is effectively reduced, and the identification accuracy of the female embryonated eggs and the male embryonated eggs is improved; the spectral characteristics of hatching eggs during incubation can be more accurately represented by converting the one-dimensional spectral characteristic matrix of the hatching eggs during incubation into the two-dimensional spectral characteristic matrix, and the accuracy of identification of female embryonated eggs and male embryonated eggs is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a schematic view of the structure of a sorting apparatus according to the present invention;
FIG. 2 is a schematic view of the structure of the control box of the present invention;
FIG. 3 is a flow chart of a sorting method of the present invention;
FIG. 4 is a schematic diagram of a 5-layer convolutional neural network design in the sorting method of the present invention;
wherein, 1-an optical chamber; 2-a control box; 3-a shell; 4-side door; 5-a focusing lens; 6-an objective table; 7-hatching eggs; 8-a collimating lens; 9-a first optical fiber; 10-a second optical fiber; 11-light source adjusting knob; 12-a voltage stabilizing module; 13-a lithium battery; 14-ARM processor; 15-fiber optic spectrometer; 16-a touch display screen; 17-heat dissipation holes; 18-a fan; 19-a halogen light source; and 20-placing the rack.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Referring to fig. 1-2, the invention provides an optical fiber spectrum sorting device for hatching egg male and female embryos during incubation, which comprises a light chamber 1 and a control box 2, wherein the light chamber 1 needs to ensure that the collection process of spectrum information is not influenced by external illumination, so that the light chamber 1 adopts a stainless steel box body, and black paint is sprayed in the box body and a rectangular stainless steel box body can be adopted; the control box 2 is made of acrylic plates.
A focusing lens 5, an objective table 6 and a collimating lens 8 are sequentially arranged in the light chamber 1 from top to bottom, the focusing lens 5 is arranged at the top of the light chamber 1, the objective table 6 is positioned in the middle of the light chamber 1 and fixedly connected to the inner wall of the light chamber 1, wherein in order to enable the egg 7 to be placed more stably, a circular hole is formed in the objective table 6, and the egg 7 is placed in the circular hole during sorting; the focusing lens 5 penetrates through the bottom of the light chamber 1 and is arranged at the bottom of the light chamber 1, an egg 7 is placed on the objective table 6, the focusing lens 5, the collimating lens 8 and the egg 7 are positioned on the same vertical line, wherein the collimating lens 8 is used for receiving the transmission spectrum of the egg 7, and the collimating lens 8 can be a 84UV type collimating lens;
a voltage stabilizing module 12, a halogen light source 19, a lithium battery 13, an ARM processor 14 and an optical fiber spectrometer 15 are arranged in the control box 2, wherein the optical fiber spectrometer 15 is used for processing the transmission spectrum acquired by the collimating lens 8 and transmitting the processed spectrum information to the ARM processor 14, the optical fiber spectrometer 15 is a Maya2000Pro type optical fiber spectrometer, a light source adjusting knob 11 is fixedly connected to the outer side wall of the control box 2, the light source adjusting knob 11 is electrically connected with the voltage stabilizing module 12, the voltage stabilizing module 12 is fixed to one inner side wall of the control box 2, the optical fiber spectrometer 15 is arranged on one side, away from the light source adjusting knob 11, of the voltage stabilizing module 12, the voltage stabilizing module 12 is respectively electrically connected with the halogen light source 19 and the lithium battery 13, the halogen light source 19 is installed on the voltage stabilizing module 12, wherein the halogen light source 19 is used to provide the light energy required for detection of the hatching egg 7 during the incubation period; the lithium battery 13 is fixed on the other side wall of the control box 2, wherein the lithium battery 13 is used for supplying power to the voltage stabilizing module 12 and the ARM processor 14; a touch display screen 16 is fixedly installed on the front side wall of the control box 2, an ARM processor 14 is arranged on the touch display screen 16, the ARM processor 14 is electrically connected with the touch display screen 16, the lithium battery 13 and the optical fiber spectrometer 15 respectively, the touch display screen 16 is used for displaying processing information of the ARM processor 14 and providing a man-machine interaction interface, and meanwhile, the ARM processor 14 is used for identifying the hatching eggs 7 as female embryonated eggs or male embryonated eggs according to the obtained spectral information and transmitting an identification result to the touch display screen 16; the ARM processor 14 is an ARM11 processor with S3C6410 as a core; the peripheral circuit comprises a USB interface circuit which is mainly used for communicating with the spectrometer; the ARM processor 14 is internally embedded with a WinCE operating system, and the operating system is internally embedded with spectrum information processing software; the voltage stabilizing module 12 and the light source adjusting knob 11 are used for power supply and brightness adjustment of the halogen light source 19, the light source adjusting knob 11 adjusts the output power of the voltage stabilizing module, and the lithium battery 13 provides power for the voltage stabilizing module 12; in addition, in order to effectively dissipate heat of the control box 2, the side wall of the control box 2 is provided with a heat dissipation hole 17, a fan 18 is buckled in the heat dissipation hole 17, an output shaft of a motor (not shown in the figure) is fixedly connected to a central shaft of the fan 18, when the heat is small, the heat can be directly dissipated from the heat dissipation hole 17, when the heat is large, the motor needs to be started, and the motor drives the fan 18 to replace the air in the atmosphere and the heat flow in the control box 2, so that the heat dissipation effect is achieved.
The bottom of the focusing lens 5 is connected with the control box 2 through a second optical fiber 10, and the bottom of the collimating lens 8 is connected with the control box 2 through a first optical fiber 9; wherein the second optical fiber 10 is used for transmitting the light energy of the halogen light source 19 to the condenser lens, the condenser lens is used for focusing the light energy of the halogen light source 19, and the first optical fiber 9 is used for transmitting the transmission spectrum collected by the collimating lens 8 to the fiber spectrometer 15.
In order to facilitate the installation or replacement of components in the sorting device, the optical chamber 1 is further optimized to include a housing 3 and a side door 4, and the side door 4 is hinged to the housing 3; a placing frame 20 is fixedly installed on the inner wall of the top of the shell 3, wherein the placing frame 20 is a through hole formed in the bottom of the placing frame 20, a mounting hole is formed in the top of the shell 3, and the collimating lens 8 is installed in the placing frame 20 through the mounting hole.
In a further optimized scheme, one end of the first optical fiber 9 penetrates through the control box 2 and is fixedly connected to the optical fiber spectrometer 15, and the other end of the first optical fiber is fixedly connected to the bottom of the collimating lens 8; one end of the second optical fiber 10 penetrates through the control box 2 and is inserted on the side wall of the control box 2, and is arranged corresponding to the halogen light source 19, and the other end is fixedly connected to the bottom of the focusing lens 5, wherein the second optical fiber 10 is a quartz optical fiber.
The specific working flow of the sorting device of the invention is as follows:
firstly, parameters of all detection instruments in the sorting device are adjusted by a light source adjusting knob 11; setting parameters such as dark current, reference current, smoothing times, smoothing width, integration time and the like of the fiber spectrometer 15;
secondly, vertically placing the hatching eggs 7 in a circular hole above a condensing lens during hatching; the method comprises the steps of closing a side door 4 of an optical chamber 1 to ensure that visible light interference does not exist in the optical chamber 1, collecting the spectral information of hatching eggs 7 by using spectral processing software embedded in an ARM processor 14, processing the spectral information and converting the spectral information into a two-dimensional spectral information matrix through characteristic wavelength selection operation, guiding the spectral information matrix into a trained identification network to obtain distinguishing information, transmitting the distinguishing information to a touch display screen 16 by the ARM processor 14, displaying female embryonated eggs or male embryonated eggs in the touch display screen 16, and completing portable sorting of the female embryonated eggs and male embryonated eggs in the hatching eggs 7 during hatching.
Referring to fig. 3-4, the invention further provides a method for sorting hatching egg 7 male and female embryos by optical fiber spectrum during incubation, and a method for sorting hatching egg 7 male and female embryos by optical fiber spectrum during incubation, which comprises the following steps:
s1, building an optical fiber spectrum sorting device for hatching egg 7 male and female embryos during hatching;
s2, collecting the spectrum information of the hatching eggs 7 during the incubation period by adopting the optical fiber spectrum sorting device built in the S1; wherein the collected spectrum wavelength range is 200-1000nm, and the sampling interval is 0.5 nm;
s3, processing the spectral information of the hatching egg 7 during the hatching period collected in S2 to obtain a two-dimensional spectral information matrix;
s4, constructing a hatching egg 7 male and female information recognition network based on the deep learning and the two-dimensional spectral feature matrix obtained in S3.4, and training the hatching egg 7 male and female information recognition network; the hatching egg 7 male and female information recognition model can be established by adopting a 5-layer convolutional neural network and a two-dimensional spectral information matrix, the hatching egg 7 male and female information recognition model is trained through a training set, the training times are 20000, and the training is finished and then the training is stored as a pb file and is exported; testing the hatching egg 7 male and female information identification model through a development set and a test set to determine the applicability of the hatching egg 7 male and female information identification model; the analysis of the 5-layer convolutional neural network is as follows:
the 5-layer convolutional neural network comprises 1 input layer, 3 convolutional layers and 1 full-connection layer, and the 5-layer convolutional neural network is designed as follows:
s4.1, input layers (Inputs): converting the characteristic wavelength points selected from the training set samples in the step S3.3 into a two-dimensional spectrum matrix as the input of a network;
s4.2, convolutional layer 1(conv 1): the number of convolution kernels is not less than 60; activating the input layer in the step S4.1 by using a ReLU function after convolution;
s4.3, convolutional layer 2(conv 2): the number of convolution kernels is not less than 190, after the output in the step S4.2 is subjected to convolution, LRN operation and ReLU activation are carried out, and the characteristic matrix is output to the convolution layer 3;
s4.4, convolutional layer 3(conv 3): the number of convolution kernels is not less than 380, and the ReLU and LRN operations are added to the output of the convolution layer 2 in the step S4.3 and then the output is output to the full connection layer;
s4.5, full connectivity layer (FC): the number of the neurons is more than 500, and the output of the convolution layer 3 in the step S4.4 is leveled up to be more than 500 weight values;
s4.6, Output layer (Output): respectively obtaining score coefficients of female embryonated eggs and male embryonated eggs by the weight values output by the full connection layer in the step S4.5 through a softmax function;
s5, inputting the two-dimensional spectral feature matrix constructed in the S3 into the hatching egg 7 male and female information recognition model trained in the S4, completing recognition of female embryonated eggs and male embryonated eggs, and displaying recognition information on the touch display screen 16; specifically calling pb files stored in S4, inputting a hatching egg 7 two-dimensional spectral feature matrix during hatching, and completing identification of female embryonated eggs and male embryonated eggs; and (4) according to the score coefficients of the female embryonated eggs and the male embryonated eggs obtained in the step (S4.6), identifying the female embryonated eggs and the male embryonated eggs, finally outputting 0 or 1, and displaying the female embryonated eggs or the male embryonated eggs on the touch display screen.
Further optimizing the scheme, in order to ensure the accuracy adopted in the above scheme, the specific implementation steps in the step S3 are as follows:
s3.1, correcting and removing noise from the spectral information collected in the S2 to obtain a sample set; the spectrum information collected in S2 is corrected by using MSC multi-element scattering correction, so that errors caused by environment and operation can be removed; carrying out noise removal on the corrected spectral information by using an SG smoothing filtering algorithm;
s3.2, performing labeling processing on the sample set obtained in the S3.1, and dividing the sample set into a training set, a development set and a test set according to a preset proportion; labeling the female embryonated egg in the step S3.2 as 0 and the male embryonated egg as 1, wherein the ratio of the training set, the development set and the test set in the step S3.2 is 7:2: 1;
s3.3, finding out a set of characteristic wavelength points capable of reflecting gender information of the hatching eggs 7 from the samples of the training set in the S3.2, and realizing spectral information dimension reduction; the wavelength range of the hatching egg 7 sex information is 200-1100nm, 1800 characteristic wavelength points are contained, if the data dimension is too large, network training is not facilitated, and meanwhile, the method for selecting the characteristic wavelength points of the denoised spectrum of the training set sample can select a set of the characteristic wavelength points capable of reflecting the hatching egg 7 sex information by using an information-free variable elimination method in combination with a continuous projection algorithm; in addition, the selection of characteristic wavelength points is realized by combining a genetic algorithm GA with a continuous projection algorithm spa, and 1800-dimensional original spectral data can be reduced to be within 50 dimensions.
And S3.4, constructing a two-dimensional spectral characteristic matrix based on the characteristic wavelength points selected from the training set sample in the S3.3, wherein the selected characteristic wavelength point set is a one-dimensional spectral characteristic matrix, and converting the one-dimensional spectral characteristic matrix into a two-dimensional spectral information matrix by multiplying the one-dimensional spectral characteristic matrix by a transpose matrix of the one-dimensional spectral characteristic matrix.
In order to further verify the effectiveness of the sorting method, in this embodiment, a total of 345 Jinyun sheldrake hatching eggs purchased from the Shendan breeding duck breeding base in Hubei province are selected, wherein the ratio of female embryo eggs to male embryo eggs is about 1:1, the training set is 242, the development set is 72, and the test set is 31, the detection accuracy of the training set is 93.36%, the detection accuracy of the development set is 93.12%, and the accuracy of the verification set is 93.83%.
In the description of the present invention, it is to be understood that the terms "longitudinal", "lateral", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", and the like, indicate orientations or positional relationships based on those shown in the drawings, are merely for convenience of description of the present invention, and do not indicate or imply that the referenced devices or elements must have a particular orientation, be constructed and operated in a particular orientation, and thus, are not to be construed as limiting the present invention.
The above-described embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solutions of the present invention can be made by those skilled in the art without departing from the spirit of the present invention, and the technical solutions of the present invention are within the scope of the present invention defined by the claims.
Claims (6)
1. An optical fiber spectrum sorting method for hatching egg male and female embryos during hatching is characterized by comprising the following steps: the method comprises the following steps:
s1, building an optical fiber spectrum sorting device for male and female embryos of hatching eggs (7) during hatching;
s2, collecting the spectrum information of the hatching eggs (7) during the incubation period by adopting the optical fiber spectrum sorting device built in the S1;
s3, processing the spectral information of the hatching eggs (7) during the hatching period collected in the step S2 to obtain a two-dimensional spectral information matrix;
s4, building a hatching egg (7) male and female information identification network model based on the deep learning and the two-dimensional spectral feature matrix obtained in the step S3, and training the hatching egg (7) male and female information identification network model;
s5, inputting the two-dimensional spectral feature matrix constructed in the step S3 into the female and male information identification network model of the hatching egg (7) trained in the step S4, completing identification of female embryonated eggs and male embryonated eggs, and displaying identification information on a touch display screen (16);
the specific implementation steps in step S3 are as follows: s3.1, correcting and removing noise from the spectral information collected in the step S2 to obtain a sample set; the spectrum information collected in S2 is corrected by using MSC multi-element scattering correction, and errors caused by environment and operation are removed; carrying out noise removal on the corrected spectral information by using an SG smoothing filtering algorithm;
s3.2, performing labeling processing on the sample set obtained in the step S3.1, and dividing the sample set into a training set, a development set and a test set according to a preset proportion; the ratio of the training set, the development set and the test set in the step S3.2 is 7:2: 1;
s3.3, finding out a set of characteristic wavelength points capable of reflecting gender information of the hatching eggs (7) from the samples of the training set in the step S3.2, and realizing the dimension reduction of the spectral information; the wavelength range of the hatching egg 7 sex information is 200-1100nm, 1800 characteristic wavelength points are contained, if the data dimension is too large, the network training is not facilitated, and meanwhile, the characteristic wavelength point selection method for the spectrum after the training set sample is denoised selects a set of characteristic wavelength points capable of reflecting the hatching egg 7 sex information by using an information-free variable elimination method in combination with a continuous projection algorithm; in addition, the selection of characteristic wavelength points is realized by combining a genetic algorithm GA with a continuous projection algorithm spa, and 1800-dimensional original spectral data are reduced to be within 50 dimensions;
s3.4, constructing a two-dimensional spectral characteristic matrix based on the characteristic wavelength points selected from the training set sample in the step S3.3, wherein the selected characteristic wavelength point set is a one-dimensional spectral characteristic matrix, and converting the one-dimensional spectral characteristic matrix into a two-dimensional spectral information matrix by multiplying the one-dimensional spectral characteristic matrix by a transpose matrix of the one-dimensional spectral characteristic matrix;
in the step S4, a hatching egg (7) sex information identification model is established by using a 5-layer convolutional neural network and a two-dimensional spectral information matrix, the 5-layer convolutional neural network includes 1 input layer, 3 convolutional layers and 1 full-link layer, and the 5-layer convolutional neural network is designed as follows:
s4.1, input layers (Inputs): converting the characteristic wavelength points selected from the training set samples in the step S3.3 into a two-dimensional spectrum matrix as the input of a network;
s4.2, convolutional layer 1(conv 1): the number of convolution kernels is not less than 60; activating the input layer in the step S4.1 by using a ReLU function after convolution;
s4.3, convolutional layer 2(conv 2): the number of convolution kernels is not less than 190, after the output in the step S4.2 is subjected to convolution, LRN operation and ReLU activation are carried out, and the characteristic matrix is output to the convolution layer 3;
s4.4, convolutional layer 3(conv 3): the number of convolution kernels is not less than 380, and the ReLU and LRN operations are added to the output of the convolution layer 2 in the step S4.3 and then the output is output to the full connection layer;
s4.5, full connectivity layer (FC): the number of the neurons is more than 500, and the output of the convolution layer 3 in the step S4.4 is leveled up to be more than 500 weight values;
s4.6, Output layer (Output): respectively obtaining score coefficients of female embryonated eggs and male embryonated eggs by the weight values output by the full connection layer in the step S4.5 through a softmax function;
the optical fiber spectrum sorting device comprises an optical chamber (1) and a control box (2), wherein a focusing lens (5), an objective table (6) and a collimating lens (8) are sequentially arranged in the optical chamber (1) from top to bottom, the focusing lens (5) is arranged at the top of the optical chamber (1), the objective table (6) is located in the middle of the optical chamber (1) and fixedly connected to the inner wall of the optical chamber (1), the focusing lens (5) penetrates through the bottom of the optical chamber (1) and is arranged at the bottom of the optical chamber (1), hatching eggs (7) are placed on the objective table (6), and the focusing lens (5), the collimating lens (8) and the hatching eggs (7) are located on the same vertical line;
a voltage stabilizing module (12), a halogen light source (19), a lithium battery (13), an ARM processor (14) and a fiber optic spectrometer (15) are arranged in the control box (2), a light source adjusting knob (11) is fixedly connected to the outer side wall of the control box (2), the light source adjusting knob (11) is electrically connected with the voltage stabilizing module (12), the voltage stabilizing module (12) is fixed to the inner side wall of the control box (2), the fiber optic spectrometer (15) is arranged on one side, far away from the light source adjusting knob (11), of the voltage stabilizing module (12), the halogen light source (19) and the lithium battery (13) are respectively electrically connected with the voltage stabilizing module (12), the halogen light source (19) is installed on the voltage stabilizing module (12), the lithium battery (13) is fixed to the other side wall of the control box (2), a touch display screen (16) is fixedly installed on the front side wall of the control box (2), the touch display screen (16) is provided with an ARM processor (14), and the ARM processor (14) is respectively and electrically connected with the touch display screen (16), the lithium battery (13) and the optical fiber spectrometer (15);
the bottom of the focusing lens (5) is connected with the control box (2) through a second optical fiber (10), and the bottom of the collimating lens (8) is connected with the control box (2) through a first optical fiber (9).
2. The method for fiber optic spectroscopic sorting of hatching egg male and female embryos during incubation of claim 1, wherein: light room (1) includes casing (3) and side door (4), side door (4) articulate on casing (3), fixed mounting has a rack (20) on casing (3) top inner wall, the through-hole has been seted up to the bottom of rack (20), the mounting hole has been seted up at casing (3) top, collimating lens (8) are installed in rack (20) through the mounting hole, the round hole has been seted up on objective table (6), hatching egg (7) are placed in the round hole.
3. The method for fiber optic spectroscopic sorting of hatching egg male and female embryos during incubation of claim 1, wherein: one end of the first optical fiber (9) penetrates through the control box (2) and is fixedly connected to the optical fiber spectrometer (15), and the other end of the first optical fiber is fixedly connected to the bottom of the collimating lens (8); one end of the second optical fiber (10) penetrates through the control box (2) and is inserted on the side wall of the control box (2) and is arranged corresponding to the halogen light source (19), and the other end of the second optical fiber is fixedly connected to the bottom of the focusing lens (5).
4. The method for fiber optic spectroscopic sorting of hatching egg male and female embryos during incubation of claim 1, wherein: and (3) labeling the female embryonated egg in the step S3.2 as 0 and labeling the male embryonated egg as 1.
5. The method for fiber optic spectroscopic sorting of hatching egg male and female embryos during incubation of claim 1, wherein: and (4) according to the score coefficients of the female embryonated eggs and the male embryonated eggs obtained in the step (S4.6), identifying the female embryonated eggs and the male embryonated eggs, finally outputting 0 or 1, and displaying the female embryonated eggs or the male embryonated eggs on the touch display screen.
6. The method for fiber optic spectroscopic sorting of hatching egg male and female embryos during incubation of claim 1, wherein: in step S4, the male and female information recognition model of the hatching egg is trained through the training set, where the training times are 15000 times and 25000 times.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010913816.9A CN111999257B (en) | 2020-09-03 | 2020-09-03 | Optical fiber spectrum sorting device and method for male and female hatching egg embryos during hatching |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010913816.9A CN111999257B (en) | 2020-09-03 | 2020-09-03 | Optical fiber spectrum sorting device and method for male and female hatching egg embryos during hatching |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111999257A CN111999257A (en) | 2020-11-27 |
| CN111999257B true CN111999257B (en) | 2022-04-12 |
Family
ID=73464545
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010913816.9A Active CN111999257B (en) | 2020-09-03 | 2020-09-03 | Optical fiber spectrum sorting device and method for male and female hatching egg embryos during hatching |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111999257B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103487396A (en) * | 2013-09-20 | 2014-01-01 | 华东交通大学 | Near-infrared fruit sugar degree nondestructive detecting device with adjustable illumination parameters |
| CN109142248A (en) * | 2018-08-27 | 2019-01-04 | 华中农业大学 | Early chick embryo male and female know method for distinguishing |
-
2020
- 2020-09-03 CN CN202010913816.9A patent/CN111999257B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103487396A (en) * | 2013-09-20 | 2014-01-01 | 华东交通大学 | Near-infrared fruit sugar degree nondestructive detecting device with adjustable illumination parameters |
| CN109142248A (en) * | 2018-08-27 | 2019-01-04 | 华中农业大学 | Early chick embryo male and female know method for distinguishing |
Non-Patent Citations (4)
| Title |
|---|
| 卷积神经网络用于近红外光谱预测土壤含水率;王璨等;《光谱学与光谱分析》;20180131;第38卷(第1期);引言部分,1.1节，1.2节，2.3节 * |
| 基于深度学习的种鸭蛋孵化早期受精信息无损检测;李庆旭等;《农业机械学报》;20200131;第51卷(第1期);1.2，2.2，2.3节，第191页的2.2.2节，图1，图6，表1 * |
| 基于深度学习的鸡蛋胚胎分类方法研究;颜廷玉;《中国优秀硕士学位论文全文数据库 医药卫生科技辑》;20181015;第23，33-34页 * |
| 基于高光谱图像的鸡种蛋孵化早期胚胎性别鉴定;潘磊庆等;《农业工程学报》;20160131;第32卷(第1期);1.1-1.2节，184-185页，表1，图2 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111999257A (en) | 2020-11-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104316473A (en) | Gender determination method for chicken hatching egg incubation early embryo based on hyperspectral image | |
| CN101672839B (en) | Device and method for detecting hatching egg incubation quality based on computer vision | |
| CN109060143B (en) | Pig body temperature monitoring system and method based on thermal infrared technology | |
| CN105043998B (en) | One kind differentiates the haploid method of corn | |
| CN106546541A (en) | A kind of identifying device and method based on EO-1 hyperion transgenic corns seed | |
| CN204500970U (en) | For studying the experimental provision of batch toy anxiety/frightened behavior reaction | |
| CN204502024U (en) | For simulate severe psychiatric stress with the device of extensive experiment | |
| CN105866050A (en) | Low-cost lossless and fast detecting equipment for apple moldy core | |
| CN113283352B (en) | Organoid vitality evaluation method and system based on microscopic image | |
| CN106846462B (en) | insect recognition device and method based on three-dimensional simulation | |
| CN112595367A (en) | Rice root system property nondestructive measurement device based on intelligent robot | |
| CN114027052A (en) | Illumination regulation and control system for plant reproductive development | |
| CN114898405B (en) | Portable broiler chicken anomaly monitoring system based on edge calculation | |
| CN111999257B (en) | Optical fiber spectrum sorting device and method for male and female hatching egg embryos during hatching | |
| CN105352555A (en) | Portable detection device for quickly identifying storage time of fowl eggs, and application method thereof | |
| CN102788796B (en) | Nutrient diagnosis device and nutrient diagnosis method for nitrogen of crops based on multi-information integration of high spectral images and fluorescent images | |
| CN109406506B (en) | Shared self-testing health terminal and testing method | |
| CN115152357B (en) | Plant seed culture dish, germination checking device, cultivation device and germination method | |
| CN116297236A (en) | Method and device for identifying vitality of single corn seeds based on hyperspectrum | |
| CN109993076A (en) | A kind of white mouse behavior classification method based on deep learning | |
| CN111802281B (en) | Optical fiber spectrum grading detection device and method for fertilized eggs in pre-incubated breeding duck eggs | |
| CN115358973A (en) | Organoid ATP analysis method and system based on artificial intelligence | |
| CN112220449B (en) | Photoelectric chicken embryo survival detection device and method | |
| CN107389594B (en) | Terahertz light-based main stem terminal bud identification and positioning method and device | |
| CN109633143B (en) | System and method for detecting bone marrow microcirculation environment of patient after hematopoietic stem cell transplantation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |