Non-Destructive Identification Of Chicken Embryo Eggs In Early Hatching Based On Spectroscopy

In poultry breeding,hatching of chicken embryo eggs is a time consuming and energy intensive process that takes about 21 days.Chicken embryo eggs hatch under natural conditions,the ratio of male to female is roughly 1:1.Because the sex of chicken embryo eggs in embryo hatching period cannot be identified at present,it is generally observed after the chicks are shelled according to the sexual inheritance of the feathers and the genitals to observe the reproductive organs.Artificial methods to determine the sex of chicks and the elimination of male chicks not only cause serious waste of resources,but also the method of slaughtering of chicks such as mixing and beating does not conform to the principle of animal welfare farming.Therefore,gender identification of chicken embryo eggs,according to market demand to adjust the ratio of hatching of male and female embryo eggs,not only can save equipment to reduce incubation costs,has realistic industrial benefits,but also avoid social ethics issues.In this study,the chicken embryo egg of Jingfen No.1 was taken as the research object.The gender identification model of chicken embryo egg was studied by spectroscopy technique and pattern recognition method,so as to realize gender nondestructive identification in the early stage of chicken embryo egg hatching.The main research contents and conclusions of this topic are as follows:(1)The method of placement and the number of days of hatching.The transmission spectra of chicken embryo eggs from 0d-15 d in vertical vertical and horizontal placement were collected,and an extreme learning machine(ELM)and random forest(RF)model were established.The identification was the best when placed vertically and hatched to the 7th day,with an accuracy rate of 81.43%.(2)Spectral pretreatment.The transmission spectrum which was placed vertically and hatched to the 7th was preliminarily divided into five different wavelength ranges according to the nature of the light,and the ultraviolet/visible band(360 nm-780 nm)was selected as the effective band for the identification of chicken embryo eggs.The accuracy rate is 84.29%.The best preprocessing methodis Multiplicative Scatter Correction(MSC),which can effectively reduce the error caused by the longitudinal shift of the spectrum.(3)Characteristic wavelength screening.In order to select the characteristic wavelength points reflecting the differences between the male and the male embryos,two methods,such as Competitive Adaptive Reweighted Sampling(CARS)and Successive Projection Algorithm(SPA),were used to Spectral data in the visible light range was used to screen the characteristic wavelengths and establish an ELM model for comparative analysis.It was found that the SPA screening effect was better than CARS.The SPA screens obtained 9 characteristic wavelength variables of 366 nm,368 nm,373 nm,388 nm,398 nm,415 nm,417 nm,438 nm,and 458 nm,which greatly reduced the model input variables.(4)Model building and optimization.In order to determine the optimal parameters of the ELM model,the number of hidden layer neurons is optimized under three different activation functions,and three discriminant models such as ELM,CARS-ELM and SPA-ELM in the ultraviolet/visible band are obtained.The optimal parameter combination is used.The input weight and hidden layer threshold of ELM are optimized by GA algorithm.The accuracy of the three discriminant models is improved.The accuracy of the GA-ELM model with SPA filtering feature wavelength is 87.14%.The number of variables used by the SPA is only 2.14% of the ultraviolet/visible band,and the model is shorter in time.Therefore,the best discriminant model for the identification of chicken embryo eggs is the GA-ELM model with UV-visible wavelength-selected wavelengths in the SPA.The female recognition rate is 88.57%,the male recognition rate is 85.71%,and the average discrimination time for a single sample is 0.0801 ms.(5)Software development.Using the idea of mixed programming of Qt and MATLAB,the software design and functional module division were completed,and the chicken embryo egg sex detection software was developed to realize the non-destructive detection of chicken embryo egg sex.