CN116930176A - Method and system for detecting sex of early embryo of chicken hatching eggs based on hyperspectral technology - Google Patents

Abstract

The invention discloses a prediction method and a prediction system for detecting the sex of early embryo of chicken hatching eggs based on a hyperspectral imaging system. The prediction method comprises the following steps: collecting hyperspectral transmission images of the big head parts of the 0-14d chicken hatching eggs during incubation by using a hyperspectral imaging system, and then correcting and extracting spectral data information; establishing a 0-14d chicken hatching egg embryo sex discrimination model, comparing model discrimination rates under different incubation days, and determining the optimal detection days; the optimal hyperspectral pretreatment method of the chicken hatching egg embryo is selected through analyzing four different pretreatment algorithms; on the basis, screening characteristic wave bands, and constructing an optimal model for judging the sex of the chicken hatching egg embryos; and finally, identifying the sex of the chicken hatching egg embryo by using the constructed model. The invention can realize nondestructive and effective detection of chicken hatching egg embryos, has early discrimination time and high detection speed, thereby providing important basis for judging the sex of chicken hatching eggs in a hatchery, solving the problem of animal ethics and reducing the economic loss of the hatchery.

Description

A method and system for detecting the gender of early embryos of chicken hatching eggs based on hyperspectral technology

Technical field

The invention relates to the field of hyperspectral non-destructive detection, and specifically relates to a method and system for detecting the gender of early embryos of chicken hatching eggs based on hyperspectral technology.

Background technique

my country's poultry egg production ranks first in the world. As a branch of poultry eggs, the identification of embryonic sex is an important aspect that affects later production efficiency. For example, in broiler production, it is hoped to hatch male chicks, which can produce more meat due to their high feed conversion rate and faster growth rate. However, in the production of laying hens, more female chicks are expected to be hatched, and male chicks are usually disposed of right after birth. Every year, more than 7 billion day-old male chicks are suffocated or soaked in carbon dioxide and culled globally, causing serious animal ethics issues and resulting in significant economic losses. Therefore, early sexing of chicken embryos during incubation is a crucial issue from a commercial and ethical (animal welfare) perspective. If the gender of the embryo can be identified during embryonic development, it will not only solve the animal welfare problem, but also save the cost of hatching and the cost of disposal of waste chicks.

At present, my country's poultry and egg industry usually performs manual testing on the hatching of chicken eggs, especially the gender detection after the chicks are hatched. The main methods include the anus turning method, the feather speed method and the feather color method. The anal turning method requires professionals to perform detection, which is time-consuming and labor-intensive, and can easily cause certain damage to the chicks; the feather color method and the feather speed method are only for the detection of some proprietary species, and have great limitations; hyperspectral image detection technology It has the advantages of simultaneous detection of spectra and images and non-destructive testing, and obtains images and spectral information of the sample. At present, hyperspectral technology has been widely used in quality inspection of fruits, vegetables and meat. However, studies using hyperspectral technology to detect the gender of chicken egg embryos are rare.

Contents of the invention

In order to solve the above problems, the present invention provides a method and system for detecting the gender of early embryos of chicken hatching eggs based on hyperspectral technology, which can quickly, accurately and non-destructively detect the gender of chicken hatching egg embryos to meet the needs of online detection for gender identification of chicken hatching eggs. .

In order to achieve the above objects, the present invention provides the following solutions:

The present invention proposes a method and system for detecting the gender of early embryos of chicken hatching eggs based on hyperspectral technology, which includes the following steps:

Steps: Select chicken eggs of the same size and with no cracks in the shell for incubation;

Step 2: Collect hyperspectral transmission images of the large heads of 0-14d hen eggs; correct the collected hyperspectral images to obtain standard images;

Step 3: Extract the spectral information of the hyperspectral image and establish a gender discrimination model for 0-14d chicken egg embryos; determine the optimal number of days for detection;

Step four is based on step three. By analyzing four different pre-processing algorithms, the best hyperspectral pre-processing method for chicken egg embryos was selected;

Step 5: Based on Step 4, further screen the characteristic wavelengths, optimize the model, and establish an embryonic gender discrimination model for hyperspectral data corresponding to the characteristic wavelengths;

Step 6: Use the established hatching egg embryo gender discrimination model to detect the hatching egg embryo gender.

The incubation temperature of the incubator is 37.8°C and the relative humidity is 65%; the eggs are automatically turned over every two hours.

The hyperspectral image collection method of hatching eggs is line scanning;

The image resolution is 440*804, the exposure time is 72ms, the acquisition speed is 1.5mm/s, and the light source intensity is 255W.

Spectral data were extracted and analyzed using MatlabR2010b software. Select the grayscale image of the bulk part of the chicken hatching egg at a wavelength of 689nm as the mask feature image, perform binary threshold segmentation and background removal, count the total number of pixels and the sum of effective spectral values of the sample, and extract the average spectrum of each sample value, and the relative transmittance is obtained after normalization for subsequent data processing and analysis.

Discriminant model partial least squares-discriminant analysis (PLS-DA) and support vector machine (SVM) optimization model

Preprocessing methods include automatic standardization (Autoscale), variable standardization (standard normalizedvariate, SNV), multiplicative scatter correction (MSC) and the first derivative (The first derivative, 1-st).

Feature wavelength screening methods include continuous projection algorithm (SPA) and competitive adaptive reweighting algorithm (CARS).

The invention also provides a system for detecting the gender of early embryos of chicken hatching eggs based on hyperspectral technology, including: an image acquisition module: used to collect hyperspectral images of the large head of chicken hatching eggs; an image correction module: used to eliminate uneven light source intensity, The impact on spectral information caused by camera dark current; spectral information extraction module: used to extract hyperspectral data of each pixel in the large head of hen eggs; preprocessing method processing module: used to eliminate or weaken the noise and baseline in the original spectrum The influence of redundant information such as drift, background color and absorption peak overlap on spectral information, to obtain more effective spectral information and high-precision and stable models; characteristic wavelength extraction module; used to extract characteristic wavelength optimization models in hyperspectral data. Training module: used to train partial least squares discriminant analysis and support vector machine models with hyperspectral data to obtain an identification model for the gender of chicken hatching egg embryos; Discrimination module: used for the identification model of the gender of chicken hatching egg embryos to identify chicken hatching eggs The gender of the embryo is determined.

According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

The present invention applies hyperspectral imaging technology to the detection of the gender of early embryos in hatching chicken eggs, and combines it with computer algorithms to achieve fast, accurate, and non-destructive detection of the gender of chicken hatching embryos, and provides information for early embryo gender detection of chicken hatching eggs in hatcheries. help.

Description of the drawings

Figure 1 is a flow chart of a method for detecting the gender of early embryos in chicken hatching eggs based on hyperspectral technology.

Figure 2 shows the average hyperspectral data extraction process of hyperspectral images of the bulk part of chicken hatching eggs.

Figure 3 shows the characteristic wavelengths extracted from hyperspectral data of the bulk part of chicken hatching eggs using the SPA algorithm.

Figure 4 shows the variable screening process of the CARS algorithm in the 400-1000nm band range.

Detailed ways

The technical solution of the present invention will be further described in detail with reference to the following specific examples.

The purpose of the present invention is to provide a method and system for detecting the gender of early embryos of chicken hatching eggs based on hyperspectral technology, which can quickly, accurately and non-destructively detect the gender of chicken hatching egg embryos to meet the needs of online detection for gender identification of chicken hatching eggs.

In order to make the above objects, features and advantages of the present invention more obvious and understandable, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

As shown in Figure 1, the present invention provides a method for detecting the gender of early embryos of chicken hatching eggs based on hyperspectral technology, including the following steps:

Step 1: Select chicken eggs of the same size and with no cracks in the shell for incubation;

Step 2: Collect hyperspectral transmission images of the large heads of 0-14d hen eggs; correct the collected hyperspectral images to obtain standard images;

Step 3: Extract the spectral information of the hyperspectral image and establish a gender discrimination model for 0-14d chicken egg embryos; determine the optimal number of days for detection;

Step four is based on step three. By analyzing four different pre-processing algorithms, the best hyperspectral pre-processing method for chicken egg embryos was selected;

Step 5: Based on Step 4, further screen the characteristic wavelengths, optimize the model, and establish an embryonic gender discrimination model for hyperspectral data corresponding to the characteristic wavelengths;

Step 6: Use the established chicken egg embryo gender discrimination model to detect the gender of the egg embryo.

Among them, step one specifically includes:

Select chicken eggs that are uniform in size and have no cracks in the shell. After wiping, disinfecting and drying with 75% alcohol, place them in an incubator to start incubation. The temperature of the incubator is 37.8°C and the relative humidity is 65%. The incubator automatically turns the eggs every two hours.

Among them, step two specifically includes:

(1) The hen eggs are placed with the big head upright and the small head down. The light source is located directly below the sample, and the camera lens is located directly above the sample, 30.0cm away from the sample.

(2) After preliminary experiments, it was finally determined that the image resolution is 440*804, the exposure time is 72ms, the acquisition speed is 1.5mm/s, and the light source intensity is 255W.

(3) Collection of hyperspectral images: Use a hyperspectral imager to collect hyperspectral transmission images of the large head of chicken eggs using line scanning.

(4) Correction of hyperspectral images: Calibrate the hyperspectral images and use the following formula to correct the hyperspectral images of the large heads of chicken hatching eggs collected.

In the formula: R is the corrected hyperspectral image; R0 is the original image; B is the all-black transmission image: W is the all-white transmission image.

Among them, step three specifically includes:

(1) Extraction of spectral information from hyperspectral images: Use MatlabR2010b software to extract spectral data and analyze it. Select the grayscale image of the bulk part of the chicken hatching egg at a wavelength of 689nm as the mask feature image, perform binary threshold segmentation and background removal, count the total number of pixels and the sum of effective spectral values of the sample, and extract the average spectrum of each sample value, and the relative transmittance is obtained after normalization for subsequent data processing and analysis (Figure 2).

(2) Establish a gender discrimination model for 0-14d chicken hatching egg embryos: the discrimination model includes partial least squares-discriminant analysis (PLS-DA) and support vector machine (SVM) optimization models. When building the model, all data sets are divided into training sets and test sets, with a ratio of 2:1. The model is evaluated through the discrimination accuracy of the modeling set and prediction set. Establish PLS-DA and SVM embryo sex identification models for chicken hatching eggs from 0 to 14 days incubation, and compare the model identification accuracy under different incubation days. The results show that the accuracy of the embryo sex identification model based on PLS-DA and SVM reaches the highest on the 9th day. 80.00% and 82.50% respectively.

Among them, step four specifically includes:

By analyzing four different preprocessing algorithms, they are Autoscale, standard normalized variate (SNV), multiplicative scatter correction (MSC) and the first derivative (The first dcrivative, 1-st). . The result is that the PLS-DA model established after SNV preprocessing is optimal.

Among them, step five specifically includes:

(1) Extract characteristic wavelengths based on SPA algorithm: SPA is a forward loop algorithm for selecting characteristic bands. It is mainly used to eliminate collinearity between original spectral data, screen out characteristic bands, prevent data from crossing and overlapping, and simplify the model. . According to the RMSE value obtained by the SPA algorithm as the basis for selecting characteristic variables, in the 400-1000nm band, a total of 9 characteristic wavelengths were screened out: 400, 416, 700, 730, 808, 853, 903, 923, 935nm ( image 3).

(2) Extract characteristic wavelengths based on the CARS algorithm: The CARS algorithm obtains a series of wavelength variable subsets through multiple repeated screenings, and the optimal characteristic wavelength combination obtained has the smallest cross-validation root mean square error (RMSECV). Figure 4 shows the process of screening characteristic variables by the CARS algorithm in the 400-1000nm band. The number of Monte Carlo sampling is 50. As the number of sampling increases, the number of selected characteristic wavelengths decreases, and the RMSECV first decreases and then slowly increases. At the 30th sampling time, RMSECV is the smallest, indicating that wavelength variables unrelated to male and female information have been eliminated in the first 29 sampling times, while key variables related to male and female information may have been eliminated in the last 21 sampling times. When RMSECV is the smallest, we get The 18 characteristic wavelengths are 663, 668, 669, 671, 674, 676, 679, 698, 700, 701, 703, 704, 705, 707, 708, 710, 711, 713nm.

(3) Extract the hyperspectral data corresponding to the characteristic wavelengths extracted by the SPA and CARS algorithms, input them into the PLSDA model respectively, and obtain the early embryo gender detection model of chicken hatching eggs after training. The results show that the accuracy rates of the modeling set and prediction set based on the characteristic wavelength prediction model of the SPA algorithm are 91.25% and 82.50% respectively; the accuracy rates of the modeling set and prediction set based on the characteristic wavelength prediction model of the CARS algorithm are 81.25% and 72.50% respectively. . The results show that rapid, accurate, and non-destructive detection of the embryonic gender of chicken hatching eggs can be achieved by using hyperspectral data from the bulk head of chicken hatching eggs to extract characteristic wavelengths and constructing an early embryonic gender detection model for chicken hatching eggs.

The invention is a method based on hyperspectral technology to detect the early embryonic gender of chicken eggs, thereby solving animal welfare problems and reducing hatchery economic losses. Traditional gender detection of chicken hatching eggs relies on manual means and empirical judgment, which is prone to misjudgment, and needs to be tested after incubation, which results in low detection efficiency; it also causes serious animal ethics issues. Therefore, the present invention applies hyperspectral imaging technology combined with computer algorithm to detect the gender of early embryos in chicken hatching eggs to improve the economic efficiency of the hatchery.

The above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art can still make modifications to the foregoing embodiments. Modifications are made to the recorded technical solutions, or equivalent substitutions are made to some of the technical features; however, these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions claimed by the present invention.

Claims (7)

1. A method for detecting the gender of early embryos of chicken hatching eggs based on hyperspectral technology, which is characterized by including the following steps: Step 1: Select chicken eggs of the same size and with no cracks in the shell for incubation; Step 2: Collect hyperspectral transmission images of the large heads of 0-14d hen eggs; correct the collected hyperspectral images to obtain standard images; Step 3: Extract the spectral information of the hyperspectral image and establish a gender discrimination model for 0-14d chicken egg embryos; determine the optimal number of days for detection; Step four is based on step three. By analyzing four different pre-processing algorithms, the best hyperspectral pre-processing method for chicken egg embryos was selected; Step 5: Based on Step 4, further screen the characteristic wavelengths, optimize the model, and establish an embryonic gender discrimination model for hyperspectral data corresponding to the characteristic wavelengths; Step 6: Use the established chicken egg embryo gender discrimination model to detect the gender of the egg embryo. 2. The method for identifying the sex of early embryos of chicken hatching eggs according to claim 1, characterized in that: the incubation temperature of the incubator in step one is 37.8°C and the relative humidity is 65%; the eggs are automatically turned over every two hours. 3. The method for identifying the sex of early embryos of chicken hatching eggs according to claim 1, characterized in that: in step two, the hyperspectral image acquisition method of chicken hatching eggs is a line scan method; the image resolution is 440*804, and the exposure time is 72ms. The collection speed is 1.5mm/s, and the light source intensity is 255W. 4. The method for identifying the sex of early embryos of chicken hatching eggs according to claim 1, characterized in that: in step three, MatlabR2010b software is used to extract and analyze spectral data. The grayscale image of the bulk part of the chicken hatching egg at a wavelength of 689nm is selected as the mask feature image. Binary threshold segmentation and background removal are performed. The average spectral value of each sample is extracted. After normalization, the relative transmittance is obtained for subsequent data. processing and analysis. The discriminant model includes partial least squares-discriminant analysis (PLS-DA) and support vector machine (SVM) optimization models. 5. The method for gender discrimination of early embryos of chicken hatching eggs as claimed in claim 1, characterized in that: the preprocessing methods in step four include automatic standardization (Autoscale), variable standardization (standard normalized variate, SNV), multivariate scattering correction ( multiplicative scatter correction (MSC) and the first derivative (The firstderivative, 1-st). 6. The early embryo gender discrimination method for chicken hatching eggs according to claim 1, characterized in that: the characteristic wavelength screening method in step five includes continuous projection algorithm (SPA) and competitive adaptive reweighting algorithm (CARS). 7. A system for detecting the gender of early embryos of chicken hatching eggs based on hyperspectral technology, which is characterized by including: an image acquisition module: used to collect hyperspectral images of the large head of chicken hatching eggs; an image correction module: used to eliminate uneven light source intensity , camera dark current, etc. on spectral information; spectral information extraction module: used to extract hyperspectral data of each pixel in the large head of hen eggs; preprocessing method processing module: used to eliminate or weaken the noise in the original spectrum, The impact of redundant information such as baseline drift, background color and absorption peak overlap on spectral information, to obtain more effective spectral information and high-precision and stable models; Feature wavelength extraction module: used to extract feature wavelength optimization models in hyperspectral data . Training module: Use hyperspectral data to train partial least squares discriminant analysis and support vector machine models to obtain a gender identification model for chicken hatching egg embryos; Discrimination module: Apply the described gender identification model for chicken hatching egg embryos to identify chicken hatching egg embryos Distinguish gender.